On the Empirical Characterization of the Low Voltage Distribution Grid as a Transmission Medium for Narrowband Power Line Communications (9-500kHz)

Ph.D. Thesis On the Empirical Characterization of the Low Voltage Distribution Grid as a Transmission Medium for Narrowband Power Line Communications (9-500 kHz) Author: Jon González-Ramos Supervisors: Dr. Itziar Angulo Pita Dr. Igor Fernández Pérez 2025 Ph.D. Thesis On the Empirical Characterization of the Low Voltage Distribution Grid as a Transmission Medium for Narrowband Power Line Communications (9-500 kHz) Author: Jon González-Ramos Supervisors: Dr. Itziar Angulo Pita Dr. Igor Fernández Pérez 2025 (cc) 2025 Jon González-Ramos (cc by-nd 4.0) i ii Abstract Narrowband Power Line Communication (NB-PLC) technologies exploit the existing Low Voltage (LV) distribution grid for carrying data signals, covering the 3-500 kHz frequency band. Although it offers several advantages over other wired or wireless alternatives, the frequencyand time-dependent transmission medium can jeopardize the proper performance of NB-PLC systems. In recent years, a great deal of effort has been devoted by the scientific community to characterize the electrical grid as a communication channel. However, due to the relentless evolution of the power network, mainly related to the expected increase in the number of Electric Vehicles (EVs) and Distributed Energy Resources (DERs), the propagation medium is expected to have changed significantly. Therefore, a proper analysis of the current and future characteristics of the LV grid as a transmission medium is still needed. In this context, this Doctoral Thesis aims to address two main objectives. First, characterize the electrical grid as a communication channel in terms of the Non-Intentional Emissions (NIEs), the grid access impedance, and the attenuation in the 9-500 kHz frequency band and, second, evaluate NB-PLC technologies according to PRIME v1.4 standard under different channel conditions. Considering the high penetration of EVs expected in the coming years, this Doctoral Thesis specifically focuses on characterizing the electrical grid in the presence of EV Charging Processes (EVCPs). Concerning the conducted emissions, this Doctoral Thesis covers three main aspects. First, the emissions generated by EVCPs under isolated conditions, i.e., using a Line Impedance Stabilization Network (LISN), are qualitatively characterized in both the frequency and time domains. Second, a novel procedure for the quantitative evaluation of the NIEs generated by EVCPs is presented, in addition to analyzing their interaction and propagation through a controlled LV grid. Finally, the emissions registered during the charging of several EVs under isolated and on-line conditions (when measuring directly in the LV grid) are compared by applying the previously defined method. With regard to the grid impedance, this Doctoral Thesis deals with the sub-cycle, mean, and long-term grid impedance variations due to EVCPs by means of two measurement campaigns carried out in France and Austria. Regarding the evaluation of the performance of NB-PLC technologies under different channel conditions, this Doctoral Thesis covers three main areas of research. First, the influence of the spectral characteristics of the previously characterized conducted emissions generated by EVCPs is individually evaluated through laboratory trials. Similarly, second, the potential effects of impedance variations are addressed, considering both frequency-dependent and sub-cycle grid impedance variations. Finally, in order to evaluate the performance of PRIME v1.4 in a situation close to real grid conditions, this Doctoral Thesis investigates the potential degradation of the communications in a controlled LV grid in Austria. In conclusion, this Doctoral Thesis deals with the characterization of the LV distribution grid as a transmission channel and the evaluation of the performance of PRIME v1.4 under different channel conditions, in order to take advantage of the power network for NB-PLC. iii Laburpena NB-PLC teknologiak behe-tentsioko banaketa-sarea erabiltzen du 3 kHz eta 500 kHz bitarteko maiztasun-bandan datu-seinaleak garraiatzeko. Harizko edo hari gabeko beste aukera batzuen aldean hainbat abantaila dituen arren, maiztasunaren eta denboraren mendeko transmisio-bideak zaildu egin dezake NB-PLC sistemen funtzionamendu egokia. Azken urteotan, komunitate zientifikoak ahalegin handia egin du BT banaketa-sarea komunikazio-kanal gisa karakterizatzeko. Hala ere, sare elektrikoaren etengabeko bilakaera dela eta, batez ere ibilgailu elektrikoen eta banatutako energia baliabideen kopuruaren hazkundearekin lotuta, komunikazio-kanala nabarmen aldatuko dela espero da. Beraz, beharrezkoa da behe tentsio sarearen egungo eta etorkizuneko ezaugarriak behar bezala aztertzea, transmisio-bide gisa. Testuinguru horretan, Doktorego Tesi honek bi helburu nagusi jorratu nahi ditu. Lehenik eta behin, sare elektrikoa komunikazio-kanal gisa ezaugarritzea, nahigabeko emisioen, sarearen sarbide-inpedantziaren eta atenuazioaren terminoetan, 9-500 kHz-eko maiztasunbandan, eta, bigarrenik, NB-PLC teknologia ebaluatzea PRIME v1.4 estandarraren arabera, kanal-baldintza desberdinetan. Datozen urteetan ibilgailu elektrikoen barneratze handia espero dela kontuan hartuta, Doktorego Tesi honek sare elektrikoaren karakterizazioa ibilgailu elektrikoak kargatzeko prozesuetan oinarrituko du. Emisioei dagokienez, Doktorego Tesi honek hiru alderdi nagusi biltzen ditu. Lehenik eta behin, baldintza isolatuetan ibilgailu elektrikoak kargatzeko prozesuek sortutako emisioak kualitatiboki ezaugarritzen dira, hau da, LISN bat erabiliz, bai maiztasunaren eremuan, bai denboraren eremuan. Bigarrenik, prozedura berritzaile bat aurkezten da ibilgailu elektrikoak kargatzeko prozesuek sortutako nahigabeko emisioen ebaluazio kuantitatiborako, eta, horrez gain, haien interakzioa eta hedapena aztertzen dira behe tentsio sare kontrolatu baten bidez. Azkenik, baldintza isolatuetan eta behe tentsio sarean zuzenean neurtzean hainbat IEk kargatzean erregistratutako emisioak alderatzen dira, aurrez zehaztutako metodoa aplikatuz. Sarearen inpedantziari dagokionez, Doktorego Tesi honek ibilgailu elektrikoak kargatzeko prozesuek eragindako inpedantziaren azpizikloko, batezbesteko eta epe luzeko aldaketak jorratzen ditu. Horretarako, bi neurketa-kanpaina egiten dira Frantzian eta Austrian. NB-PLC teknologien errendimendua hainbat kanal-baldintzatan ebaluatzeari dagokionez, Doktorego Tesi honek hiru ikerketa-arlo nagusi biltzen ditu. Lehenik eta behin, aurrez karakterizatutako ibilgailu elektrikoak kargatzeko prozesuek sortutako emisioen ezaugarri espektralen eragina ebaluatzen da, laborategiko saiakuntzen bidez. Era berean, bigarrenik, inpedantzia-aldakuntzen efektuei helduko zaie, maiztasunaren mendeko inpedantziaaldakuntzak zein azpizikloko sareko inpedantzia-aldakuntzak kontuan hartuta. Azkenik, PRIME v1.4ren errendimendua sarearen baldintza errealetatik hurbil dagoen egoera batean ebaluatzeko, Doktorego Tesi honek komunikazioen degradazio potentziala ikertzen du behe tentsioko sare batean. Ondorioz, sare elektrikoak NB-PLCrako eskaintzen dituen aukerak maximizatzeko asmoz, Doktorego Tesi honek behe tentsioko banaketa-sarearen karakterizazioa transmisio-kanal gisa eta PRIME v1.4ren errendimenduaren ebaluazioa jorratzen ditu, kanal-baldintza desberdinetan. iv Resumen NB-PLC es una tecnología que utiliza la existente red de distribución de baja tensión para transportar señales de datos en la banda de frecuencias de 3 kHz a 500 kHz. Aunque ofrece diversas ventajas sobre otras alternativas cableadas o inalámbricas, el medio de transmisión, dependiente de la frecuencia y el tiempo, puede dificultar el correcto funcionamiento de los sistemas NB-PLC. En los últimos años, la comunidad científica ha dedicado un gran esfuerzo a caracterizar la red de distribución de baja tensión como canal de comunicación. Sin embargo, debido a la incesante evolución de la red eléctrica, principalmente relacionada con el aumento del número de vehículos eléctricos y recursos energéticos distribuidos, se espera que el medio de propagación haya cambiado significativamente. Por lo tanto, es necesario un análisis adecuado de las características actuales y futuras de la red de baja tensión como medio de transmisión. En este contexto, esta Tesis Doctoral pretende abordar dos objetivos principales. En primer lugar, caracterizar la red eléctrica como canal de comunicación en términos de las emisiones no intencionadas, la impedancia de acceso de la red y la atenuación, en la banda de frecuencias 9-500 kHz y, en segundo lugar, evaluar las tecnologías NB-PLC según el estándar PRIME v1.4 bajo diferentes condiciones de canal. Teniendo en cuenta la alta penetración de vehículos eléctricos que se espera en los próximos años, esta Tesis Doctoral se centra específicamente en la caracterización de la red eléctrica en presencia de procesos de carga de vehículos eléctricos. En cuanto a las emisiones conducidas, esta Tesis Doctoral cubre tres aspectos principales. En primer lugar, se caracterizan cualitativamente las emisiones generadas por los procesos de carga de vehículos eléctricos en condiciones aisladas, es decir, utilizando una LISN, tanto en el dominio de la frecuencia como en el del tiempo. En segundo lugar, se presenta un novedoso procedimiento para la evaluación cuantitativa de las emisiones no intencionadas generadas por los procesos de carga de vehículos eléctricos, además de analizar su interacción y propagación a través de una red de baja tensión controlada. Por último, se comparan las emisiones registradas durante la carga de varios vehículos eléctricos en condiciones aisladas y al medir directamente en la red de baja tensión, aplicando el método previamente definido. En cuanto a la impedancia de red, esta Tesis Doctoral aborda las variaciones sub-ciclo, media y a largo plazo de la impedancia debidas a los procesos de carga de vehículos eléctricos. Para ello, se realizan dos campañas de medida en Francia y Austria. En cuanto a la evaluación del rendimiento de las tecnologías NB-PLC en diferentes condiciones de canal, esta Tesis Doctoral abarca tres áreas principales de investigación. En primer lugar, se evalúa individualmente, mediante ensayos de laboratorio, la influencia de las características espectrales de las emisiones conducidas generadas por procesos de carga de vehículos eléctricos previamente caracterizadas. Asimismo, en segundo lugar, se abordan los efectos de las variaciones de impedancia, considerando tanto las variaciones de impedancia dependientes de la frecuencia como las variaciones sub-ciclo de impedancia. Finalmente, para evaluar el rendimiento de PRIME v1.4 en una situación cercana a las v condiciones reales de la red, esta Tesis Doctoral investiga la degradación potencial de las comunicaciones en una red de baja tensión controlada en Austria. En conclusión, con el fin de maximizar las oportunidades ofrecidas por la red eléctrica para NB-PLC, esta Tesis Doctoral aborda la caracterización de la red de distribución de baja tensión como canal de transmisión y la evaluación del rendimiento de PRIME v1.4 bajo diferentes condiciones de canal. vi Acknowledgments Desde que empecé a redactar el documento de tesis, llevo pensando en que este momento llegaría, el de escribir estos agradecimientos y recapitular todo lo que ha supuesto para mí esta etapa de casi tres años y medio. Espero haber encontrado las palabras que describan lo agradecido que estoy por el apoyo que he sentido durante este tiempo. En primer lugar, quiero dar las gracias al equipo de la línea Txispas. Trabajar con vosotros es un verdadero placer y espero seguir haciéndolo en los próximos años. David, gracias por tu cercanía, por siempre estar dispuesto a revisar un artículo, incluso en tus ratos libres los fines de semana, y por tus labores de gestión en la línea. Amaia, gracias por tu buen humor y tus bromas, esos ratos de cotilleo en los que uno se entera de todo y, sobre todo, por estar siempre abierta a echarme una mano. Quiero agradecer también a Alex, con quien he podido compartir este camino desde que empezamos en TSR. Gracias por todas esas conversaciones mañaneras sobre nuestro futuro (y esas conversaciones de Txisbec) y, por supuesto, por ayudarme con las infinitas gestiones que he tenido que hacer todos estos años. A Javier, mucha suerte en tu tesis en marcha, no tengo dudas de que saldrán grandes resultados de ella. A Mikel, gracias por todas esas conversaciones futbolísticas de los cafés y por todo el tiempo que has dedicado en las campañas de medidas sin las cuales esta tesis no existiría. Quiero agradecer también a Idurre, por su cercanía y por el buen trabajo de estos primeros años de investigación. Espero que podamos seguir trabajando juntos en el futuro. Y, por supuesto, a mis directores de tesis. A Igor, gracias por el buen rollo que siempre transmites, por estar siempre dispuesto a echar una mano y, cómo no, por ese maravillo sistema de medidas sin el cual esta tesis no sería posible. Gracias a tus clases de TC, entré hace más de cinco años en TSR y estoy hoy a punto de defender esta tesis doctoral. Y, por último, a Itziar, gracias por enseñarme tu pasión por la investigación y por el trabajo bien hecho. Sin tus revisiones minuciosas y llenas de color, este documento de tesis no sería como es. Pero, sobre todo, quiero darte las gracias por ser siempre un apoyo para mí, por aconsejarme y preocuparte por mí, y por guiarme en este camino como lo has hecho. Gracias por todas esas conversaciones sobre la vida y el futuro, me han ayudado mucho. Quiero dar las gracias también al resto de miembros de TSR. Especialmente, a Teresa, por todas esas horas de duro gimnasio y por esas infinitas conversaciones más allá del trabajo. Gracias por estar siempre ahí para mí, especialmente en los momentos más difíciles. Gracias también a mis dos tontos, Iñigo y Endika, por alegrarme los días en TSR. Sin vosotros mis cien nombres no existirían. Endika, gracias por todos los buenos momentos que hemos pasado juntos fuera y dentro de TSR, estoy seguro de que todavía nos quedan muchos por vivir. Iñigo, gracias por todas esas conversaciones sobre la vida de los últimos años. Espero que nos queden muchos palitos por tomar en los que recordar que “Hay un loco en Internet”. Quiero agradecer también a Eneko, por las risas de los cafés y todas esas conversaciones fuera de TSR. Espero que podamos seguir cotilleando y “criticando” juntos muchos años. A JoM, por estar siempre dispuesto a ayudar y alegrarnos los cafés con sus historias y sus memes. Gracias también por esos lunes de patxanga y, especialmente, por los kalimotxos terapéuticos previos, han sido un gran apoyo para mí. A Marta, por el buen rollo que siempre creas y por alegrar los cafés con esas pullas de las que nadie se puede librar . Al resto de doctorandos de TSR (Erick, Orlando, Dreyelian, Alejandro, vii Elizabet), mucha suerte en lo que os queda de tesis, estoy seguro de que haréis un trabajo excelente. Me gustaría agradecer también a Bernhard, por acogerme en su Universidad y por darme la oportunidad de conocer una ciudad como Viena. Gracias por tu ayuda en esa etapa de esta tesis, espero que sigamos colaborando en el futuro. Quiero dar también las gracias a mis amigos, que me han acompañado a lo largo de todos estos años. Primero, a mis telecos (Vicente, Adrián, Cristina, Jon Marcos, Ainhoa, Rada, Ixone, Xabi, Andrea, Laura, Izaskun, Andoni), por todos los buenos momentos que hemos pasado durante y después de la carrera. Sin vosotros, probablemente, no estaría donde estoy. Quiero dar también las gracias a Sendoa, por todos esos martes/miércoles de Champions y sábados/domingos de Premier, y esas infinitas conversaciones futbolísticas con pipas y pizzas. Gracias también a mis amigos de Santa Pola (Eli, David, Ignacio, Alex…) por alegrarme las vacaciones de verano y hacer que los meses de desconexión sean más divertidos. A mis madrileños (Paula, Beli, Rapo, Nerea, Bea, Karen y Fran) y vallisoletana (Elisa) favoritos, gracias por esos viajes por tierras palentinas, en los que comenzó una gran amistad. Estoy seguro de que todavía nos quedan muchos Villoldos (esperemos que sin sustos) por delante. Quiero agradecer también a Josu, por ser como un hermano para mí. Estoy seguro de que los años no cambiarán que sigamos comportándonos como niños de cinco (😉) años. Gracias a Marian y Fernan, por ser siempre un apoyo y hacerme sentir uno más de la familia. Por último, quiero agradecer a mi familia. A mis padres, Rosa y Gonzalo, por apoyarme en todas las decisiones que he tomado en la vida, habéis sido un pilar fundamental. Gracias también por transmitirme vuestros valores y por enseñarme que es necesario insistir para conseguir lo que uno busca. Sin vosotros, hoy no sería quien soy. Y a Malena, gracias por ser mi gran apoyo. Por aguantarme en mis días buenos y por soportarme y alegrar mis días malos. Gracias por estar siempre, por impulsarme a ser mejor. Creo que no tengo palabras para describir todo lo que significas para mí. Nunca te estaré lo suficientemente agradecido. xiv List of Tables Table I. Summary of the journal papers associated to the contributions of this Doctoral Thesis............................................................................................................................................... 24 Table II. Comparison between current NB-PLC protocols and standards [1], [56]. ........ 27 Table III. Comparison between current BB-PLC protocols and standards [47]. .............. 29 Table IV. Connected devices at each house (T1, T2, T3, T4) in the controlled LV grid at UASV. ............................................................................................................................................. 51 Table V. Summary of the journal papers associated to the contributions of this Doctoral Thesis............................................................................................................................................... 78 Table VI. Summary of the international conference papers associated to the contributions of this Doctoral Thesis. ................................................................................................................ 79 Table VII. Summary of the national conference papers associated to the contributions of this Doctoral Thesis. ..................................................................................................................... 79 xv Acronyms AC Alternate Current AMI Advanced Metering Infrastructure AMN Artificial Mains Network ARIB Association of Radio Industries and Businesses AWGN Additive White Gaussian Noise BB-PLC Broadband Power Line Communications BER Bit Error Rate BPF Band Pass Filter CDF Cumulative Density Function CENELEC European Committee for Electrotechnical Standardization CFR Channel Frequency Response CL Compatibility Level DBPSK_C Differential Binary Phase Shift Keying with Forward Error Correction DC Direct Current DQPSK_C Differential Quadrature Phase Shift Keying with Forward Error Correction DER Distributed Energy Resource DSL Digital Subscriber Lines DSO Distribution System Operator EDF Électricite de France EMC Electromagnetic Compatibility EUT Equipment Under Test EV Electric Vehicle EVCP Electric Vehicle Charging Process EVCS Electric Vehicle Charging Station FCC Federal Communications Commission FEC Forward Error Correction FER Frame Error Rate xvi FFT Fast Fourier Transform HDR High Data Rate JCR Journal Citation Reports LDR Low Data Rate ICES Industry Canada Equipment Standard for Digital Equipment LISN Line Impedance Stabilization Network LQI Link Quality Indicator LV Low Voltage MTL Multi-Conductor Transmission Line MV Medium Voltage NB-PLC Narrowband Power Line Communications NIEs Non-Intentional Emissions OFDM Orthogonal Frequency Division Multiplexing PFBL Percentage of the frequency bins exceeding the limits PLC Power Line Communications POC Point of Connection PQ Power Quality PWM Pulse Width Modulation QP Quasi-Peak R_DBPSK Robust Binary Quadrature Phase Shift Keying R_DQPSK Robust Differential Quadrature Phase Shift Keying RCFMFD Random Carrier Frequency Modulation Fixed Duty RF Radiofrequency RMS Root Mean Square RSSI Received Signal Strength Indicator RTT Round Trip Time SG Smart Grid SM Smart Meter SNR Signal to Noise Ratio xvii SoC State of Charge STFT Short-Term Fourier Transform SS Secondary Substation TL Transmission Line TSHV Total Supraharmonic Voltage 2TL Two-Conductor Transmission Line UASV University of Applied Sciences Vienna UNB-PLC Ultra Narrowband Power Line Communications USRP Universal Software Radio Peripheral V2G Vehicle to Grid 18 Chapter 1 Thesis synthesis 19 1. Overview 1.1. Introduction The Smart Grid (SG) concept is bringing a great revolution to the electricity network, where digital technologies are used in order to better match the supply and demand of electricity in real time while minimizing costs and maintaining the stability and reliability of the grid. The changes are more far-reaching in the distribution and customer domains, due to the fact that SGs allow consumers to be more involved in the operation of the power system [1]. The development of the SGs means the end of the classic production-distribution-consumption scheme of the electrical grid, evolving into a more distributed, flexible and efficient model. This new model paves the way for the integration of Distributed Energy Resources (DERs), the management of Electric Vehicle (EV) charging, and the implementation of energy demand response technologies [2], [3], [4], [5]. Moreover, it also contributes to the reduction of the losses and illegal usage of the transmission and distribution lines, as well as an easy operation of the grid and automated monitoring [3]. In this inexorable evolution of the electrical grid, high-performance communication technologies are essential to fulfill the strict requirements of new SG services [6]. For this reason, some Distribution System Operators (DSOs) have turned their gaze to Power Line Communications (PLC) as an option to implement the two-way communication between the utility provider and customers in the Low Voltage (LV) distribution section of the grid. Firstly, as the electrical grid is already deployed, coverage problems are prevented even in areas not covered by other communication technologies (neither wired nor wireless) [7]; and, secondly, additional infrastructure costs are avoided. Moreover, PLC technologies provide the DSOs the opportunity to have control of their own data transmission without relying on third parties [8]. In recent years, Narrowband PLC (NB-PLC) technologies, operating in the 3-500 kHz frequency band, have been widely deployed over the LV grid for applications such as telemetry, monitoring of the devices connected to the power grid, and signal quality analysis [9], [10], [11], [12]. However, since the electrical grid was not conceived for communications, the characteristics of the transmission medium have been proven to considerably degrade the performance of NB-PLC systems: noise due to Non-Intentional Emissions (NIEs), impedance variations, high attenuation/transmission losses and the timeand frequency-dependent channel response [7], [13], [14]. Firstly, with regard to the conducted emissions, the considerable increase in the number of electronic devices connected via power converters [15] has led to an increase in the amplitude of the NIEs in the frequency range of kHz [16], [17], [18], also known as supraharmonics [19]. These high-amplitude emissions, which are inherent to the operation of power electronics devices [19], in addition to degrading the quality of NB-PLC technologies [20], [21], can also cause Power Quality (PQ) issues, such as equipment malfunction, falsification of energy/smart meters, or overvoltage [22], [23], [24], [25], [26], [27]. The relevance of the effect of the NIEs on the communications depends directly on the amplitude, spectral form, and time-variant behavior of the emission. With the aim of analyzing the influence of NIEs on PLC technologies under laboratory conditions, the 20 ETSI TS 103 909 [28] defined in 2012 a set of reference noises representative of the LV grid. However, due to the large increase of DERs in recent years, the time and spectral patterns of the emissions present in the grid are expected to have varied considerably. Moreover, the effect of this kind of NIEs on PLC is still unknown. Secondly, the grid access impedance is another major aspect affecting NB-PLC, since it significantly conditions the signal propagation [29], [30]. It also plays a pivotal role when designing communication devices connected to the grid [31], as the transferred power of the transmitted signals depends on the impedance matching between the transmission device and the effective grid access impedance [32], [33]. Besides, very low impedance values may generate near short circuit situations, which lead to overcurrent for electronic devices directly connected to the grid [31]. Finally, short-term impedance variations, given within the fundamental period of the mains (20 ms) and already reported in [13], [34], could also have a negative effect on PLC. As the impedance of the electrical grid is affected by several factors, such as harmonics, interharmonics, variety of electronic devices, and the grid topology, among others, the need for carrying out extensive measurement campaigns all around the world, considering different grid configurations, remains latent [35], [36]. Moreover, due to the large number of existing topologies and loads, significant variations in the impedance can be observed from one country to another [35], turning this analysis even more complicated. Finally, the time [37] and frequency variations of the grid access impedance result into considerable changes in the Channel Frequency Response (CFR). The modulus of the CFR is related to the losses of the transmitted signal while propagating through the grid, and it can result in PLC signals being notably attenuated [29], [30], [38]. This high attenuation can significantly reduce the maximum range of the communications [39] and is, generally, due to the distance and sudden changes in loads [14]. Furthermore, these impedance variations do not only affect the PLC signal but also the NIEs present in the grid [40], [41]. In conclusion, in order to take advantage of the potential of the LV distribution grid as a communication infrastructure for PLC communications, it is necessary to characterize the electrical network in terms of grid access impedance, channel response, and NIEs. In this way, the industry would be able to adapt data coding, modulation, and multiplexing algorithms to the current characteristics of the LV grid. 21 1.2. Motivation The characterization of the LV distribution grid as a communication channel has been addressed by the scientific community for many years. However, considering that the relentless evolution of the power network is leading to considerable changes in the characteristics of the grid, new efforts in order to study the current and future characteristics of the propagation channel are necessary. Moreover, taking into account that the majority of the contributions published so far only focus on the supraharmonic range (9-150 kHz), further studies covering the complete NB-PLC frequency band (up to 500 kHz) are still necessary. There are several reasons to carry out this characterization by means of empirical trials. First, there is still no theoretical model of the LV distribution grid in the NB-PLC frequency range (9-500 kHz). Second, although laboratory trials allow for reproducibility in the measurements, they are performed under idealized conditions and, thus, they do not represent real grid conditions. Third, since the manufacturers do not publicly share the details of the implemented converters, the theoretical analysis of the emissions is not generally possible. In this context, this Doctoral Thesis aims at characterizing the electrical grid as a propagation medium in the 9-500 kHz frequency band in terms of grid access impedance, channel response, and NIEs, as well as empirically evaluating the performance of NB-PLC under different channel conditions. 22 2. Objectives This Doctoral Thesis has two main objectives (MOs): - MO1. Empirical characterization of the electrical grid as a communication channel in terms of the grid access impedance, channel response or attenuation, and NIEs in the 9-500 kHz frequency range. - MO2. Evaluation of NB-PLC technologies under different channel conditions. In order to achieve these two MOs, a number of specific objectives (SOs) have been defined. - SO1. Characterization of the conducted emissions in the electrical grid. The conducted emissions should be characterized both in the frequency and time domains in the 9-500 kHz frequency range. Considering the high penetration of EVs in coming years, this Doctoral Thesis should specifically address the conducted emissions generated by EV charging processes (EVCPs). This characterization should include NIEs generated by a wide range of EV models, charging currents, and States of Charges (SoCs). Moreover, with the aim of determining whether isolated measurements are a good representation of the real LV grid, different measurement conditions (isolated/on-line) should be taken into consideration. Finally, in order to fully determine the effects of the EVCPs on the amplitude of the emissions in the LV grid, the propagation and interaction of these NIEs should also be analyzed. - SO2. Characterization of the grid access impedance. The sub-cycle, mean, and long-term grid access impedance should be characterized in the 9-500 kHz frequency range. As in PO1, considering EVs as an example of modern devices to be widely deployed in the following years, the grid impedance in presence of EVCPs should be completely characterized. This characterization should include different commercial EV models, charging currents, and SoCs. - SO3. Evaluation of the potential influence of conducted emissions on NB-PLC. This Doctoral Thesis should analyze the potential influence of the conducted emissions on NB-PLC. For this purpose, NB-PLC should be evaluated in presence of the previously characterized NIEs generated by EVCPs. - SO4. Evaluation of the potential influence of impedance variations on NB-PLC. This Doctoral Thesis should evaluate the potential influence of impedance variations on NB-PLC, taking into consideration both frequency-dependent impedances static over time and sub-cycle impedance variations. In this text, the term characterization refers to the calculus of representative metrics and parameters of the signals and the identification of their patterns both in the frequency and time domains. A summary of the main and specific objectives of this Doctoral Thesis are presented in Fig. 1. 23 Fig. 1. Summary of the main objectives (MOs) and specific objectives (SOs) of this Doctoral Thesis. Finally, in order to relate the objectives of this Doctoral Thesis with the published papers that make up the compendium, in Table V, a summary of the journal papers associated to the contribution of this Doctoral Thesis are presented, together with the objective to which they are related. 30 Although different BB-PLC technologies have been developed for in-home and MV channels, there is still no BB-PLC system specifically designed to be deployed over the LV distribution grid. Considering the enhanced performance with respect to NB-PLC in terms of bandwidth, latency, and security requirements, BB-PLC is an alternative that is being considered by some DSOs for data transmission through the LV distribution grid, in order to fulfill the demanding requirements of new SG applications [76]. As the behavior of the outdoor channel is expected to be considerably different from the characteristics of the indoor environment, DSOs are taking into consideration two different options. First, the adaptation of the existing indoor technologies to the characteristics of the LV grid; and, second, the development of a new standard specifically designed for the outdoor transmission medium. In any case, the development of BB-PLC technologies would definitely be a solution for covering the needs of future SG applications, including the integration of DERs or EV charging management, among others [47]. 31 3.2. Characterization of the electrical grid as a transmission medium for NB-PLC technologies In this section, the most relevant aspects to be considered for the empirical characterization of the electrical grid as a transmission medium are addressed. Section 3.2.1 deals with the conducted emissions, while the grid access impedance and the attenuation/channel response are studied in section 3.2.2. Throughout the text, the term noise corresponds to an electromagnetic phenomenon not conveying information and which is combined with a wanted signal [77]. Since the emissions referred to in this Doctoral Thesis are always non-intentional, emissions and NIEs are used as synonyms and are defined as unwanted signals generated by the power electronics included in the circuitry of connected electronic devices. The terms interference and disturbance, in turn, imply a degradation of a system. In the case of an interference, the performance of a communications system is jeopardized, while a disturbance is related to the degradation of PQ, or the lifetime of a device, among others. Finally, the term distortion refers to an undesired change in the waveform of a signal that might lead to the appearance of new frequency components. 3.2.1. NIEs Main sources of NIEs Some recently published articles [78], [79], [80], [81], [82] show that photovoltaic inverters (PV), battery chargers, energy-efficient lighting, hydropower systems, wind turbines, or EV chargers, among others, are the main sources of the high-amplitude conducted emissions for the frequency band between 3 kHz and 500 kHz. In several field and laboratory trials, a frequency and time characterization of these emissions has been carried out, concluding that, as these emissions occur at frequencies assigned to NB-PLC, the quality of communications can be substantially affected and degraded [20], [50], [83], [84]. In some instances, as for the EV chargers and PV panels, the sources of disturbance are located near the smart meter, which might imply an additional challenge for the correct performance of NB-PLC [24]. Other non-intentional emitting devices, such as motors [79], lighting devices [85], [86], [87], or electronic amplifiers [19], introduce high-amplitude emissions with a wide range of spectral patterns. A significant time-dependent behavior has also been observed for the previously mentioned sources [79]. It should also be noted that high-amplitude impulsive disturbances are generated by this equipment when commuting between different states or working regimes [79], which may have an additional negative effect on communications. Normative framework Emission limits The maximum amplitude of the emissions generated by certain equipment connected to the electrical grid has already been specified by the International Special Committee on Radio Interference CISPR. CISPR15 (EN 55015) [85] defines the limits for the lighting equipment, whereas CISPR11 (EN 55011) [88] addresses the maximum amplitudes for 32 induction cooking devices. By contrast, no specific limits have been specified for many other sources of emissions, such as PVs, EVs, or hydropower systems. For these devices, the out-of-band limits defined for communications equipment in EN 50065-1 [89] might be considered as a conservative criterion [90]. This technical specification defines limits adapted to the frequency band of transmission (2-9 kHz, 9-150 kHz, or 150 kHz-30 MHz), and the type of detector that should be used in the measurements, Root Mean Square (RMS), Quasi Peak (QP) or Average [91], [92], [93]. These emission limits are defined for laboratory conditions and must be evaluated by means of a Line Impedance Stabilization Network (LISN). A LISN is a standard load impedance allowing the repeatability and comparability of EMI measurements, which also prevents the trials from being affected by external conducted emissions [94]. In the LV distribution grid, the Compatibility Levels (CLs) establish the maximum amplitudes of the emissions that cannot be exceeded at a specific electrical point, as the combination of the emissions generated by all the equipment connected to the network. The grid operator should ensure that at least in 95 % of the locations these limits are not exceeded. The Annex B of the IEC 61000-4-7 [95] specifies the limits for the frequency band from 2 kHz to 9 kHz in RMS values, whereas the IEC 61000-2-2 [96] defines the limits for the 9-150 kHz frequency range in QP values. Measurement methods In the bibliography, the term measurement method refers to the post-processing applied to the recorded signals to evaluate NIEs in the frequency domain. Up to now, no normative measurement method for the assessment of the disturbances in the LV distribution grid has been defined for frequencies above 9 kHz. As previously mentioned, the CLs in the 9-150 kHz frequency range that are included in IEC 61000-2-2 [96] are defined for QP values. For this reason, CISPR16 1-1 [91], a method based on a QP detector, is the method that might be used for the evaluation of the disturbances in grid measurements [97]. Nevertheless, this method presents several drawbacks: first, it is not intended for grid measurements, but rather for laboratory conditions; and, second, it has high complexity, computational burden, and memory requirements. Moreover, there is no standardized procedure for the quantitative characterization of the emissions in the frequency domain. The literature only considers the Total Supraharmonic Voltage (TSHV) [98], [99], [100], a parameter that gives an insight into the total amplitude of the emissions in the frequency band under analysis. However, this parameter is highly dependent on the measurement method used for the post-processing of the corresponding emission [101] and does not take into account the frequency distribution of the amplitude of the emissions, which plays an essential role in the proper design of PLC technologies. Moreover, the concepts of narrowband and broadband emission have not yet been defined and, therefore, there is no normalized procedure for their proper characterization. Finally, it should be mentioned that these measurement methods are only intended to characterize the emissions in the frequency domain, without considering their time-dependent behavior. Since these time variations can negatively affect PLC, the development of methods that address the time characterization in conjunction with the frequency characterization is necessary, for which the definition of standardized methodologies in the time domain is of upmost importance. A joint time-frequency 33 domain characterization of the emissions in the frequency range 9-150 kHz can be found in [102]. Measurement setups Existing standards specify measurement setups for the characterization of the emissions from equipment under test. CISPR 16-1-2 [103] defines a standardized measurement setup for the evaluation of the emissions based on a LISN from 9 kHz to 109 MHz. Similarly, IEC 61000-4-7 [95] specifies the use of an Artificial Mains Network (AMN) below 9 kHz. In this way, a controlled and isolated scenario is set and the measurements are not affected by external NIEs [104]. However, the measurement campaigns carried out so far lead to conclude that the reference impedances defined in [95], [103] are not a good representation of the actual grid access impedance values of the LV distribution grid [41]. For this reason, the characterization of the emissions should not only be based on measurements using AMN or LISN, where isolated effects can be evaluated, but also on field trials, so that representative disturbance values are obtained. In this context, some authors [105] have opted for conducting investigations related with NIEs at reconstructed facilities. This type of scenarios avoid the ideal conditions of laboratory measurements as well as the uncontrolled grid factors present in on-field trials, such as the variety of loads connected to the grid, the grid topology, the grid impedance, or the electrical cables, among others [78]. Propagation/interaction of the emissions According to the literature, the emissions generated by the devices connected to the grid propagate through the LV network, and may even be transferred to the MV grid over distances of several kilometers [106], [107], [108]. As a result, they may affect the operation of energy meters and PLC equipment [109], [110]. As individual devices have a greater impact on higher frequencies than on harmonics [40], the prediction of the NIEs in this frequency range should consider the whole installation and not only individual devices [111]. In accordance to [40], [41], [112], [113], [114], [115], the emissions generated by a certain device can be classified as primary and secondary emissions. The primary emission corresponds to the emission originated inside the device, whereas the secondary emission is generated by other electronic equipment or the grid itself, and propagated into the device. This propagation highly depends on the impedance of neighboring devices in relation to the impedance of the electrical grid [112], [116], [117]. Resonances are a key aspect in the propagation of the disturbances, as they imply increases in the emission at the switching frequency [41], [112], [118]. In [40], it is stated that a resonance results in an increase in the secondary emission, whereas the primary emission is attenuated. Regarding the interaction, some works have pointed out the existence of frequency beating and intermodulation effects. For instance, in [17], several simulations carried out in MATLAB led to conclude that the charging of EVs of the same type, due to slightly different switching frequencies (f1, f1’), implies an emission at |f1-f1’|. In general, the beating frequency is in the order of some Hz and does not affect PLC. Intermodulation 34 distortion in the frequency band up to 100 kHz, in turn, occurs due to the interaction of considerably different switching frequencies (f1, f2) and is in the order of tens of kHz [17]. Another similar analysis concerning intermodulation products can be found in [119]. In this article, the interaction between a PV panel and an EV is reported, evidencing the existence of intermodulation products of up to 4th order. The interaction between end-user and PLC equipment has also been studied in the literature. Five types of interactions between these devices are described in [120], concluding that the performance of communications is degraded and the lifetime of the end-user equipment is considerably reduced. Electric vehicles The rising number of EVs involves considerable challenges for the electricity grid. For example, EV Charging Stations (EVCSs) generate higher amplitude emissions than other electronic devices connected to the grid [121]. For this reason, there is great interest in the scientific community in analyzing and characterizing the disturbances generated by EVs during the charging process. Since the manufacturers of the EVCSs do not publicly share the details of the implemented converters, the theoretical analysis of the disturbances is not generally possible. For this reason, as stated in [122], the characterization of these emissions can only be carried out empirically, by means of extensive measurement campaigns. In recent years, some laboratory and field trials have been developed for the analysis of the disturbances generated by several EVCPs. For instance, in [105], the emissions introduced by a Bi-Directional Vehicle to Grid (V2G) EVCS are studied in a controlled grid scenario up to 150 kHz. This work concludes that narrowband emissions are present at the switching frequency and multiples of it, caused by the Pulse Width Modulation (PWM), whereas broadband emissions occur at higher frequencies due to the DC-DC converter when the Positive Temperature Coefficient heater is activated. Field trials presented in [123] show that the high amplitude emissions generated by an electric bus do not decrease with frequency in the band 2-150 kHz and that they could be as high as the CL defined in [96]. In [17], the propagation and interaction of the NIEs of four EVs are analyzed up to 100 kHz in an isolated grid scenario connected to the public electricity grid. As the grid allows switching to microgrid mode, measurements in both configurations have been considered, obtaining different results in some cases. Reference [124] analyzes the long-term supraharmonic emissions of three EVs in the time and frequency domains up to 100 kHz at three parking garages. The presented results lead to conclude that the power grid disturbance levels increase when the number of connected EVs rise. In [125], the disturbances generated by different types of chargers at five sites in China and Germany are characterized in the frequency domain. The article shows the dominant emission frequencies as well as their voltage amplitude up to 50 kHz. A similar analysis is presented in [126], [127], [128], where the switching frequencies of twelve, ten, and eight EVs, respectively, are identified. A huge variability in the switching frequency and its amplitude can be observed in all three cases. However, in on-site measurements, emissions are affected by several grid factors [129], including the time and frequency-dependent grid access impedance. Thus, in some works, such as [104], the EVCS is directly connected to a LISN in order to isolate the setup from 35 the LV grid, avoiding external noise and ensuring that only the emissions generated by the EVCPs are measured. Measurement campaigns for the characterization of the NIEs Several authors have performed on-field trials in different LV distribution grids all around the world considering certain sources of disturbances and grid topologies. As an example, in [130], the NIEs at industrial, residential, and rural areas in the Turkish LV grid are characterized up to 100 kHz, obtaining emission amplitudes around 90 dBµV, 100 dBµV, and 75 dBµV, respectively. A statistical characterization of the emissions in the LV grid based on probabilistic functions is presented in [25], covering the 2-150 kHz frequency band. Another statistical analysis can be found in [131], where a long-term characterization of the emissions in the LV grid in Qatar is presented considering more than 1.8 billion samples at three different locations over 10 days. The analysis is based on the stationarity, autocorrelation, and independence of the NIEs in the frequency range from 10 kHz to 490 kHz. A measurement campaign carried out in Spain [16], in turn, concludes that the propagation channel in rural scenarios is remarkably noisy in low frequencies, as the highest NIEs occur at frequencies lower than 150 kHz, mainly lower than 40 kHz. However, high-amplitude emissions can be found in the whole frequency range up to 500 kHz in urban environments. In [12], a similar characterization of the NIEs in the NB-PLC frequency band is gathered for the LV grid in France. This article points out that the noise is predominantly cyclostationary and concludes that the background noise is stable over periods of hours. More information about the measurement systems used for developing these measurement campaigns can be found in the journal paper (JP1), which can be found in Appendix A.1. 3.2.2. Grid access impedance Normative framework As the behavior of the grid access impedance in the frequency band assigned to PLC technologies is still unknown [132], the regulatory framework in this field is very limited. IEC 61000-4-7 defines a reference grid impedance in the frequency range 2-9 kHz, but some authors have extended its definition up to 150 kHz [105], [133], [134]. However, a measurement campaign carried out in the LV grid in Austria, Switzerland, Czech Republic, and Germany demonstrates that the reference grid impedance in IEC 61000-4-7 overestimates even the highest values of the impedances measured in the actual distribution grid [133]. CISPR 16-1-2 describes three different reference impedances for defining specific LISNs depending on the frequency range (9-150 kHz, 150 kHz-30 MHz, or 150 kHz-108 MHz) [103]. According to [134], the CISPR 16-1-2 reference impedance defined in the frequency range 9-150 kHz also exceeds the values of the impedances measured in indoor environments in Germany, United Kingdom, and Spain. 36 Measurement campaigns for the characterization of the impedance Only a few measurement campaigns to characterize the grid impedance have been performed all around the world. In [130], the grid access impedance in the LV grid in Turkey is characterized up to 100 kHz, considering residential, rural, and industrial scenarios. The paper concludes that, regardless of the area, impedance values below 10 Ω are measured in the whole frequency band. Another measurement campaign covering the same frequency range in Austria, Switzerland, Czech Republic, and Germany is presented in [133], showing that the impedance increases with frequency and that remains below the IEC 61000-4-7 reference impedance in the whole frequency band. In [132], phase to neutral impedance measurements are carried out at an EV charging plaza at the university and four residential household installations in The Netherlands in the 9-150 kHz frequency range. This article concludes, first, that, due to the capacitor included in their circuitry, both EVs and PVs imply a low-impedance path for communications signals, and, second, that household appliances have a considerable effect on the grid impedance. Moreover, the importance of the topology on the impedance is also highlighted. Although most of the published studies focus, exclusively, on the 9-150 kHz frequency band, some publications provide information up to 500 kHz. For example, in [36], the access impedance from 35 kHz to 500 kHz at three Secondary Substations (SSs) and at a set of access points of the LV distribution grid in the Basque Country (Spain) has been characterized. For this purpose, two urban and a rural scenario have been considered. The study provides an insight into the great differences between the impedances measured at each electrical point and concludes that the urban distribution can be modeled as a particular scenario of short cable sections and numerous homes. In [135], a similar characterization in the frequency band 30-500 kHz is performed, measuring the grid access impedance in an urban/suburban area in China with low-rise apartment buildings. However, these results cannot be extrapolated to any distribution network, since the performance of the electrical grid at the frequencies associated to PLC changes drastically from one country to another, mainly due to the diversity of loads and grid topologies [136], [137], [138]. More information about the measurement systems used for developing these measurement campaigns can be found in the journal paper (JP1), which can be found in Appendix A.1. 3.2.3. Channel response As previously mentioned, the time and frequency variations of the grid impedance imply changes in the channel response [37]. Since the modulus of the CFR corresponds to the attenuation of the transmitted PLC signal, communications can suffer from high attenuation [29], [30], [38]. This high attenuation is one of the main factors affecting the performance of NB-PLC, since it can reduce the range of the communications [39]. The characterization of the PLC channel is carried out according to two approaches: the top-down approach, where a model is obtained from a large number of field trials, and the bottom-up approach, based on modeling the channel by applying Transmission Line (TL) theory [139]. 37 Regarding the top-down approach, the indoor PLC channel has been extensively analyzed in terms of average channel attenuation or channel gain, delay spread, coherence bandwidth, and channel capacity in [136], [137], [140], [141], [142], [143], [144], [145], [146], [147], [148], [149], [150]. In [151], the characterization of the indoor PLC channel based on multipath phenomenon can be found. Concerning the characterization of the outdoor scenario, for instance, in [152], the grid access impedance, the channel response (in terms of delay spread, channel bandwidth, and attenuation), and the NIEs in different urban, semiurban, and rural scenarios in the LV distribution grid are analyzed, as well as the achievable data rates. The results presented in this paper only cover the CENELEC A NB-PLC frequency band (3-95 kHz). A similar analysis in the frequency range assigned to BB-PLC (1.7-100 MHz) is gathered in [153], where the Brazilian outdoor scenario is characterized considering the average channel attenuation, root mean squared delay spread, coherence bandwidth, coherence time, and the achievable data rate. In reference [154], a characterization of the outdoor channel considering the multipath signal propagation theory is presented. This model is verified by a set of channel responses obtained by means of field trials. Since in the bottom-up approach the network elements are mathematically modeled, a thorough knowledge of the electrical grid (topology, cable characteristics, load impedances, etc.) is needed. This approach considers the two-conductor TL (2TL) theory [155], [156], [157], [158], [159], [160], [161], used for power networks connected with two-conductor transmission lines, and the multi-conductor TL (MTL) theory [162], [163], [164], [165], [166], [167], [168], [169], which is a generalization of the 2TL approach. The TL theory has also been applied by some authors for the characterization of the outdoor PLC channel [159], [164], [165], [170], [171]. In reference [172], both the LV aerial and underground cable distribution lines are model by means of the MTL theory. A combination of the top-down and bottom-up approaches can be found in [173], [174], where a statistical analysis of the NB-PLC channel in the LV grid in Pakistan is presented up to 150 kHz and 500 kHz, respectively. Measurement campaigns for the characterization of the channel response The huge variety of causes that involve attenuations in PLC signals and the dependency of the behavior of the electrical grid on the geographical area [136], [137], [138], mainly because of the heterogeneity of grid topologies and connected loads, makes the prediction of the PLC channel response a complex study that should be addressed. In [130], for example, the attenuation between the SS and the SMs in the LV grid in Turkey up to 1 MHz is analyzed in residential, industrial, and rural areas, revealing attenuations around 20 dB (distances from 50 m to 150 m), 30 dB (distances from 80 m to 270 m), and 40 dB (distances from 150 m to 500 m) at each scenario, respectively. Reference [175] considers a set of representative grid topologies, concluding, as in [176], that an increase in distance implies higher transmission losses. However, it is also stated that there is not a linear trend between the attenuation and the distance between communications equipment, due to the great impact of the number of branches in tree like topologies. In [177], apart from measuring the attenuation in the frequency band associated to NB-PLC, a characterization of the transmission losses measured from 500 kHz to 10 MHz in the LV grid in China is presented. A comparison between the mean transmission losses measured in urban and 38 rural residential areas is carried out, concluding that the attenuation with respect to frequency in the urban scenarios is flatter than what is measured in rural areas. More information about the measurement systems used for developing these measurement campaigns can be found in the journal paper (JP1), which can be found in Appendix A.1. 3.2.4. Conclusions and open issues Since EVs will be massively deployed in coming years [178] and have been identified as one of the main sources of emissions in the grid [121], this Doctoral Thesis will focus on evaluating the conducted emissions generated by EVCPs, as well as their impact on the grid impedance. In any case, considering that some of these distributed energy resources are also based on inverters, as it is the case of an EVCP, it is expected that at least part of the conclusions drawn in this thesis will remain valid in those cases. Similarly, it is expected that the results related to propagation and interaction of emissions from EVCPs will also apply to other sources of NIEs of a similar nature in this frequency band. Considering the state of the art presented above, a set of research gaps have been identified regarding the characterization of the conducted emissions, which are detailed below together with the main contributions of this Doctoral Thesis - Since there are neither standardized methods nor metrics for the evaluation and quantification of the NIEs in the grid, this Doctoral Thesis will define a novel procedure for the evaluation of the emissions in both the frequency and time domains. This procedure will not only analyze the amplitude of the conducted emissions, but also their spectral distribution and time-dependent behavior. - The majority of the existing studies only cover the supraharmonic (9-150 kHz) frequency band. For this reason, this Doctoral Thesis will evaluate the conducted emissions covering the whole NB-PLC frequency band (9-500 kHz). - Since most studies do not evaluate the emissions for periods longer than few minutes, this Doctoral Thesis will characterize the NIEs during 600 s, as later justified in section 4.2.2. The Doctoral Thesis will define a specific method for the quantification of the time variability within the recording time. - The literature analyzes the propagation and interaction of the emissions based on a primary/secondary approach. However, this Doctoral Thesis will address the propagation and interaction of the NIEs based on synchronized measurements at different electrical points in the LV distribution grid. - Although the literature points out that the emissions generated by EVCPs under isolated and on-line conditions are not similar, there is a lack of studies in which the NIEs generated by the same EVCPs under both isolated and LV grid conditions are compared. For this reason, this Doctoral Thesis will analyze the NIEs due to a wide range of EV models and charging currents under both measurement conditions. In order to cover these research gaps, the following criteria has been applied: - The emission recordings will be performed by means of a measurement system developed by the research team, which has been proven to provide accurate values in the frequency band of interest. 39 - Since no specific emission limits have been defined for EVCPs, this Doctoral Thesis will use the PLC out-of-band emission limits defined in EN 50065-1 [89] for comparison purposes. - Concerning the measurement methods for the spectral characterization of the emissions, the CISPR 16-1-1 standard [91] will be selected. Concerning the characterization of the grid impedance, the following research avenues have been determined, as well as the contributions to be made in this investigation area: - As previously mentioned for the conducted emissions, existing studies only cover the supraharmonic range. This Doctoral Thesis will evaluate the grid impedance in the whole NB-PLC frequency range (20-500 kHz). - The literature only addresses the characterization of the mean grid impedance in the presence of EVCPs. For this reason, this Doctoral Thesis will present a complete characterization of the impedance, considering sub-cycle (20 ms), mean, and long-term variations (some hours). - This Doctoral Thesis will characterize the variations of the CFR due to impedance variations (frequency-dependent and sub-cycle impedance variations), in order to relate both parameters. With the aim of covering these research gaps, the following criteria have been determined: - The characterization of the grid impedance in the presence of EVCPs is performed in controlled LV grids, since they show two main advantages. First, since they are isolated from the public LV grid, external factors that may affect the impedance measurements are avoided; and, second, the resemblance to real network conditions is maintained, as they are recreated based on the characteristics of real public LV grids. - The impedance and CFR measurements will be conducted by means of ad-hoc measurement systems designed by the research team, which have been proven to provide accurate values in the frequency band of interest. 46 61000-4-19:2014 [212] standards. Moreover, its accuracy has been also evaluated in a controlled and reproducible laboratory scenario and compared with respect to other methods providing similar results [214]. Fig. 5. Measurement system for the assessment of the grid access impedance. 4.1.3. Attenuation The attenuation measurement system, shown in Fig. 6, calculates the transmission losses, in modulus and phase, between two electrical points, P1 and P2. Depending on the selected configuration, the mean or sub-cycle attenuation is provided. The attenuation is calculated as the difference between the level of the injected signal at P1 and the received signal at P2. This measurement system is composed of four voltage probes: three of them connected at point P1, and one at point P2. One of the active probes connected at P1 is responsible for injecting a sweep in the frequency range 5-530 kHz and for filtering the 50 Hz signal, while the two passive probes (one at P1 and one at P2) measure the transmitted and received voltage, respectively. The signal generator, controlled by means of a Matlab script, is responsible for transmitting bursts of 90 ms not synchronized with the mains frequency from 6 kHz to 507 kHz with 3 kHz steps. A gap of 10 ms between two consecutive bursts is used as a guard period. From each 90 ms burst, only 20 ms are evaluated, disregarding the first and the last 20 ms and synchronizing the start of those 20 ms to be analyzed with a zero crossing of the mains signal. This synchronization, performed as described for the impedance measurement system in section 4.1.2, is carried out by means of one of the active voltage probes connected at point A. This voltage probe is used for measuring the 50 Hz reference signal, in order to be able to analyze the attenuation in 20 ms intervals, synchronized to the mains period. 47 For the signal post-processing, a sliding Gaussian windowing with a 1/3000 s duration and an overlapping of 75 % between consecutive windows is applied, so that the frequency resolution is 3 kHz. Therefore, a time resolution or sampling period of the estimated channel frequency response of 83.33 µs is obtained. Fig. 6. Measurement system for the assessment of the channel frequency response for variable channel conditions. 48 4.2. Measurement campaigns 4.2.1. Measurement campaign for the characterization of the emissions generated by EVCPs under isolated conditions The on-site measurements were carried out on a commercial EVCS operating in mode 3 according to IEC 61851-1 [215]. All the measurements were performed in monophasic mode. With the aim of implementing a controlled measurement scenario, the setup was isolated from the LV grid by means of a LISN [216], a small transformer (Polylux PD400 [217] with a rated power of 4000 VA), and a band-pass filter (BPF) adjusted to the PLC frequency band. Both the LISN and the transformer operated in monophasic mode. The EVCS was directly connected to the Equipment Under Test (EUT) port of the LISN. The resulting emissions were measured from the Radiofrequency (RF) port with a digital oscilloscope, which was controlled by a laptop charging at a power station. As the measurements are obtained from the RF port, a correction factor should be applied in the post-processing stage to compensate the insertion loss of the LISN between the EUT and RF ports. Typical values of the insertion loss of the LISN as a function of frequency can be found in [216]. In Fig. 7, a representation of the measurement setup used for the recording of the NIEs using a LISN is shown. Fig. 7. Measurement scenario for the recording of the NIEs generated by EVs during their charging process using a LISN (isolated conditions). The measurement campaign considers the recording of the emissions generated by two EV models and two different charging currents. 4.2.2. Measurement campaign for the characterization, propagation, and interaction of the emissions generated by EVCPs in a controlled LV grid This measurement campaign was carried out in the “Concept Grid” laboratory of Électricite de France (EDF), a unique testing facility that goes beyond ideal conditions of laboratory trials, but at the same time, avoids uncontrolled background distortion that may substantially affect the results [105]. This testing scenario simulates a LV distribution grid composed of a Secondary Substation (SS) and five houses (H1, H2, H3, H4 and H5), with a three-phase installation, to which different electronic devices can be connected. In this study, three different EVCSs are analyzed. EVCS1 is installed at H2, EVCS2 at H3, 49 and EVCS3 at H5. There are three EV models available, and each EV model can only be charged at its corresponding EVCS. A representation of the measurement scenario is shown in Fig. 8, where the distance between the different houses and the location of the EVCSs are indicated. Fig. 8. Measurement scenario in the “Concept Grid” laboratory of EDF in Écuelles. In order to have a predominantly resistive load at the POC, domestic heaters were connected to each house. All the measurements were conducted in the same electrical phase (monophasic measurements). The disturbances were measured at the Point of Connection (POC) of each house to which the EVCS under study is installed. In order to analyze the propagation of the emissions generated by each EVCP individually, synchronized measurements were carried out at H2, H3, and H5 when a single EV was charging at its corresponding EVCS. The analysis of the interaction of the NIEs is based on measurements performed at H2, H3, and H5 when the three EVs are charging simultaneously. For all the EVs under study, recordings of 600 s were available. The common time frameworks used for PQ are 20 ms, 3 s, and 1 day. However, since a recording of 1 day would result in a large amount of data, implying an unmanageable file, and 3 s would not be enough time to characterize the expected time variations, the measurement time was configured to 600 s. This measurement time corresponds to a trade-off between the file size and the time needed to evaluate the variability of the NIEs. 4.2.3. Measurement campaign for the comparison of the emissions generated by EVCPs under isolated and on-line conditions The recording of the emissions generated by EVCPs under isolated conditions was performed as described in section 4.2.1. The recording of the emissions generated by EVCPs under on-line conditions was performed at a parking plaza, where the EVCS was directly connected to the LV 50 distribution grid, as shown in Fig. 9. For this purpose, the measurement system for the assessment of the NIEs described in section 4.1.1 was used. Fig. 9. Measurement scenario for the recording of the NIEs generated by EVs during their charging process in the LV grid (on-line conditions). This measurement campaign considers: - The recording of the emissions generated by seven commercial EV models with different charging currents under isolated conditions. In the case of EV1, EV3, and EV7, only two charging currents were possible (8A and 12A), whereas three charging currents were possible for EV2, EV4, EV5, and EV6 (8A, 12A, and 16A). - The recording of the emissions generated by five commercial EV models with different charging currents under on-conditions. In the case of EV10, only one charging current was possible (16 A), while five charging currents (8 A, 10 A, 12 A, 14 A, and 16 A) were possible for the remaining EV models. For all the EVs under study, recordings of 600 s were available, except for EV6 with a charging current of 8A, for which 150 s were recorded. For the comparison of the emissions generated by a specific EV model under isolated and on-line conditions, EV1 and EV10 (16 A), EV2 and EV8 (8 A, 12 A, and 16 A), EV3 and EV12 (12 A and 16 A), EV6 and EV11 (8 A, 12 A, and 16 A), and EV7 and EV9 (8 A, 12 A, and 16 A) are considered. 4.2.4. Measurement campaign for the characterization of the sub-cycle and mean grid impedance in presence of EVCPs The measurement campaign for the characterization of the sub-cycle and mean grid impedance was carried out in the same measurement scenario described in section 4.2.2. The study considers impedance measurements in the default situation (no EV is charging in the grid) and when each EV is charging. As in section 4.2.2, each EV model can only be charged at its corresponding EVCS. 4.2.5. Measurement campaign for the characterization of the long-term grid impedance variations in presence of EVCPs The measurement campaign for the evaluation of the long-term impedance variations in the presence of EVCPs was carried out in the controlled LV grid at the University of 51 Applied Sciences Vienna (UASV) shown in Fig. 10. This scenario is composed of a SS and four houses (T1, T2, T3, T4) at which different electronic devices can be connected. Monophasic measurements were conducted at one of the phases of the installation (L1), considering a star topology. Fig. 10. Representation of the controlled LV grid at the UASV (Austria) [105]. In this measurement campaign, the default situation corresponds to the configuration in which the devices gathered in Table IV are connected. Apart from the measurement location, the phase (L) to which each device is connected is also indicated in Table IV. L123 refers to a device connected to the three phases (tree-phased). Table IV. Connected devices at each house (T1, T2, T3, T4) in the controlled LV grid at UASV. Measurement location Connected devices T1 2 Smart Meters (SMs) (L123) T2 3 SMs (L123) Photovoltaic (PV) system (L2) T3 2 SMs (L123) PV system (L2) 1200 W load (L123) T4 3 SMs (L123) PV system (L123) Storage system (L123) In order to evaluate the long-term grid impedance variations, four EVs (EVA, EVB, EVC, and EVD) were individually charged at an EVCS connected to the POC of T2. The connection of the EVCS to the grid includes a long cable of 30 meters, which simulates the typical cable length in real charging situations. 52 4.2.6. Influence of conducted emissions generated by EVCPs on NB-PLC The laboratory setup for the evaluation of NB-PLC according to PRIME v1.4, which is shown in Fig. 11, is composed of two PL360G55CF-EK boards [218] supplied by a laptop, a variable attenuator controllable by software (Att 1 in Fig. 11), a manual variable attenuator (Att 2 in Fig. 11), and a Universal Software Radio Peripheral (USRP) [219], responsible for reproducing the recorded NIEs. The attenuator controllable by software introduces a maximum attenuation of 62.5 dB with steps of 0.25 dB, while the manual attenuator allows introducing attenuations up to 110 dB with increments of 1 dB and 10 dB. Fig. 11. Laboratory scenario for the evaluation of NB-PLC according to PRIME v1.4 standard. The measurement campaign includes the conducted emissions generated by the charging processes of the seven commercial EV models with different charging currents detailed in 4.2.3 under isolated conditions. As previously mentioned, in the case of EV1, EV3, and EV7, only two charging currents were possible (8A and 12A), whereas three charging currents were possible for EV2, EV4, EV5, and EV6 (8A, 12A, and 16A). For all the EVs under study, recordings of 600 s are available, except for EV6 with a charging current of 8A, for which 150 s were recorded. 4.2.7. Influence of impedance variations on NB-PLC In order to characterize the influence of impedance variations, a controlled laboratory scenario is defined and implemented (see Fig. 12). It is composed of three LISNs, responsible for offering a time-invariant standardized impedance at the EUT port, so that the measurements are not affected by the variations of the impedance of the grid. Besides, LISNs perform a low-pass filter function, with the aim of preventing unwanted grid noise from entering the EUT. In all the measurements, the characteristic impedance of the LISNs is set to 2 Ω, according to the PRIME specification [57]. A detailed characterization of the modulus of the impedance frequency response of the LISNs can be found in Fig. 13. 53 Fig. 12. Laboratory scenario. Loads under study are connected at either points A, B, or C [57]. Fig. 13. Measured impedance modulus of the LISN model for the frequency band of interest. In addition to the FER-SNR curves that evaluate the potential effect of impedance variations on the quality of the communications, in the case of the grid impedance, the attenuation due to the load under test is also calculated. This attenuation is obtained as the difference of the modulus of the channel frequency response when the corresponding load is connected to the setup and the modulus of the channel frequency response when no load is connected. For the evaluation of the influence of frequency-dependent impedance variations, a set of Electromagnetic Compatibility (EMC) filters are selected, while for the sub-cycle impedance variations a set of commercial devices are evaluated, as well as a synthetic load designed in the laboratory. NB-PLC according to PRIME v1.4 standard are established by means of two Microchip PL360G55CF-EK evaluation kits [218], in such a way that one is configured to act as a transmitter and the other acts as the receiver. The transmitter equipment is connected to the EUT port of the first LISN, whereas the receiver is connected to the EUT port of the third LISN. The second LISN allows connecting different loads on the channel to analyze 54 their effect on communications. This way, the influence of the load location on the attenuation difference suffered by the transmitted signal is performed by connecting the loads under study to the EUT port of each of the three LISNs alternately (point A, B, and C in Fig. 12). In order to ensure that the measurements are not affected by the internal noise of the receiver and the noise introduced by the load under test in the case of the active loads, AWGN is injected by means of a signal generator connected to the RF port of the third LISN. In this way, the noise of the measurement setup is flat for the whole frequency band of interest and masks the potential effect of the noise introduced by the different equipment connected to the setup. Therefore, taking into account that the noise of the system is given by a fixed level of AWGN, it is necessary to modify the power of the transmitted signal, so that the different SNR values that make up the FER-SNR curves are obtained. For that purpose, two attenuators are located between the second and the third LISNs: a 110 dB attenuator with 1 dB and 10 dB step sizes and a 1 dB attenuator with 0.1 dB step size. A 110 dB attenuator is also connected between the first and the second LISN, in order to ensure a sufficient level of attenuation to obtain the FER-SNR curves. Its level is fixed to 3 dB, except for the cases when it is not possible to get a FER of 5 % just by varying the variable attenuators located between the second and third LISNs. 4.2.8. Evaluation of NB-PLC in a reconstructed LV grid The evaluation of NB-PLC was carried out in the reconstructed LV grid at UASV described in section 4.2.5. NB-PLC according to PRIME v1.4 were evaluated by means of two PL360G55CF-EK boards, one acting as a transmitter and connected to the SS and the other operating as a receiver and connected to the corresponding house (H1, H2, H3, and H4). In order to evaluate the performance of PRIME v1.4, for each configuration, the mean value of the SNR for the transmitted frames is related to the FER. 55 5. Results and contributions This section summarizes the main results and contributions of this Doctoral Thesis. In section 5.1, the results related to the characterization of the grid are presented, while section 0 addresses the evaluation of NB-PLC under different channel conditions. 5.1. Characterization of the electrical grid as a transmission medium 5.1.1. Characterization of the conducted emissions Characterization of the conducted emissions generated by EVCPs under isolated conditions In this contribution, a first qualitative spectral and time characterization of the conducted emissions generated by a set of EVCPs using a LISN is presented. The study, covering the 9-500 kHz frequency band, considers the emissions generated by two EV models and two charging currents during 5 s. The results show that both the spectral pattern and amplitude of the emissions depend on the EV model and charging current under study. In some cases, high-amplitude tonal or narrowband emissions are recorded at specific frequencies, while, in other, emissions in the form of colored noise decreasing with frequency are reported. An example of these spectral characteristics can be found in Fig. 14. Fig. 14. QP values of the amplitude of the emissions generated by a certain EV model for charging currents of 9 A and 10 A. Regarding the time analysis, the spectrograms of the recordings show that a significant time variability is registered in the frequency band assigned to NB-PLC (see Fig. 15). In 62 Fig. 22. QP values of the amplitude of the emissions generated by a certain EVCP in each period of 50 s during the recording time (600 s) under on-line conditions. Except for the tonal emissions with oscillating central frequency generated by a specific EV model regardless of the charging current, the tonal and narrowband emissions generated by the remaining EVCPs follow the simplified equation of the model defined in the previous section. Since the FFT component at 100 Hz is the major contributor to the total variability in these cases, it can be concluded that the NIEs generated by these EVCPs vary within 10 ms, i.e., two times within the mains cycle. This might be due to the positive-negative symmetry of the power conversion stage. In the specific case of the tonal emissions with oscillating central frequency, the FFT analysis shows a wide range of spectral components spread over the whole FFT frequency band, being the FFT components around 41 Hz and its multiples (mainly 82 Hz and 123 Hz) the ones that present higher amplitude (see Fig. 23). This implies that these emissions follow a periodic pattern with a repetition rate of 24.4 ms. 63 Fig. 23. Modulus of the normalized FFT (dB) of the time samples corresponding to a tonal emission with oscillating central frequency. Finally, it should be mentioned that, only for three out of the twelve EVCPs under study, the total variability differs by more than 3 dB between both measurement configurations. In these cases, a higher variability is reported if a LISN is used. In summary, this is the first empirical study that presents a comparison between the emissions generated by EVCPs under isolated and on-line conditions considering a wide range of EV models and charging currents. The contribution reveals that, since the frequency and time characteristics of the emissions depend on the measurement conditions, this characterization cannot be only based on trials conducted using a LISN. Moreover, due to the different behavior of the emissions between the 9-150 kHz and 150-500 kHz frequency bands, this study highlights that it is not possible to directly extrapolate the understanding of the NIEs in the 9–150 kHz frequency range, typically addressed in the literature, to the 150–500 kHz band. The journal paper related to this contribution (JP3) can be found in Appendix A.3. 5.1.2. Characterization of the grid impedance Characterization of the sub-cycle, mean, and long-term impedance variations in the presence of EVCPs This contribution aims at analyzing the influence of EVCPs on the grid impedance in the 20-500 kHz frequency band. For this purpose, two measurement campaigns are carried out in controlled LV distribution grids in France and Austria. The first measurement campaign aims to evaluate the sub-cycle and mean impedance due to EVCPs, while the second analyzes the long-term impedance variations (some hours) under these circumstances. 64 Regarding the first measurement campaign, the results show impedance values lower than 18 Ω in the frequency band 20-500 kHz when measuring at three locations under study (see Fig. 24). Some resonances are observed at certain frequencies in the modulus, which imply abrupt changes in the phase. These resonances, also reported in [135], [220], could imply an increase in the amplitude of the emissions generated by a certain device [41] and, thus, could endanger the proper operation of PLC. The observed values go in line with the results presented in [36], [135] for urban scenarios in the LV grid in Spain and China, respectively. By contrast, they considerably differ from the reference impedance values included in the normative currently in force. As shown in Fig. 24, in the 20-500 kHz frequency range, the modulus of the impedances defined in CISPR 16-1-2 ranges from 7.3 Ω to 47.7 Ω. (a) (b) Fig. 24. Modulus and phase of the mean grid access impedance measured at the three locations under study when no EV is connected (default situation), together with the reference impedance defined in [103]. The study also analyzes the influence of an EVCP on the mean grid access impedance by comparing the impedance measured in the default situation (D) and when each EV is charging, as shown in Fig. 25. (a) (b) Fig. 25. Modulus and phase of the mean grid access impedance measured at three locations under study when no EV is connected (default situation) and when each EV is charging. 65 Fig. 25 shows differences of a few ohms between the modulus of the impedance in the default situation and when each EV is charging at the same location. Nevertheless, due to the low values of the grid impedance in the default situation, these differences in amplitude represent relative variations that might be significant at specific frequencies. As it can be observed, in general, the spectral form of the impedance, both the modulus and the phase, is maintained in the whole frequency band, except for certain resonances occurring in the frequency range below 150 kHz. Moreover, the charging of an EV does not seem to imply a displacement of the previously existing resonances, as they remain at the same frequencies as in the default situation. The study also addresses the influence of an EVCP at a certain distance from the electrical point where the EVCS is installed, concluding that it seems that an increase in the distance does not reduce the influence on impedance. Therefore, these results lead to conclude that the influence of an EVCP does not only depend on the EVCP itself, but also on the grid impedance in the default situation. With respect to sub-cycle impedance variations, as shown in Fig. 26, this study also leads to conclude that sub-cycle variations, occurring within the 20 ms of the mains, occur when an EV is charging, but also in the default situation. More details about both the mean and sub-cycle variations can be found in the international conference paper (ICP2) shown in Appendix B.1.2. Fig. 26. Modulus and phase of the sub-cycle grid access impedance measured at a certain location when no EV is connected (default situation). Regarding long-term impedance variations, this contribution shows that the spectral characteristics (both amplitude and spectral shape) of the grid impedance vary considerably during the charging process of an EV (see Fig. 27). Specifically, two different impedance states are reported within the 6070 s (impedance state 1 from 0 s to 3710 s and 5880 s to 5900 s, and impedance state 2 from 3710 s to 5880 s and from 5900 s to 6070 s). This demonstrates that significant impedance variations can occur during the charging process of an EV. 66 (a) (b) Fig. 27. Modulus (a) and phase (b) of the grid impedance measured during 6070 s of the charging process of a certain EV with respect to time (vertical axis) and frequency (horizontal axis). Although they are not clearly noticeable in Fig. 27 due to the wide range of values shown in the figure, variations within each impedance state are also given. An example of these variations are depicted in Fig. 28. The variations occurring in the grid impedance of the remaining EVCPs under study can be found in the journal paper (JP4) included in Appendix A.4 (a) (b) Fig. 28. Modulus (a) and phase (b) of the grid impedance measured during the charging process of a certain EV and when no EV is charging (default situation). Fig. 28Fig. 28 clearly shows that, regardless of the impedance state and EVCP under study, not only a difference in the impedance amplitude is reported, but also the spectral shape of the impedance varies. Moreover, Fig. 28 also reveals a displacement of the resonance around 56 kHz occurring in the default situation or even the appearance of new resonances at different frequencies when each EV is charging. This involves that the behavior of the impedance might depend on the measurement scenario, which implies that it is necessary to perform measurement campaigns in different grid conditions (topology, type of cable, number of connected loads…) considering a wide range of EV models and charging configurations. 67 It should also be noted that, at frequencies below 150 kHz, low-amplitude resonances (values lower than 5 Ω) are observed both in the default situation and when each EV is charging. These low impedances can result in high attenuation, which could be critical for the proper performance of PLC [45], [120]. In the 150-500 kHz frequency range, similar spectral features are registered both in the default situation and for all the EVCPs under study (values lower than 15 Ω in the modulus and an inductive behavior in the phase). Thus, the behavior of the grid impedance seems to be more stable at frequencies above 150 kHz (both the amplitude and the spectral shape), which would facilitate the proper design of NB-PLC devices according to the characteristics of the communications channel in this frequency band. In conclusion, this is the first empirical study that analyzes the influence of EVCPs on the sub-cycle, mean, and long-term grid impedance in the 20-500 kHz frequency band. Since the results show that the EVCPs imply both sub-cycle and long-term variations, the characterization of the impedance in presence of EVs is a complex phenomenon that requires further analysis. For this reason, in order to develop a more general model, a large number of measurement campaigns should be carried out all around the world, taking into account different grid topologies, areas with different population densities, different EV models, and charging conditions. This would allow designing communication devices according to the characteristics of the grid, leading to a better performance of PLC systems. The results obtained in this study were presented in an international conference (ICP2), which can be found in Appendix B.1.2, as well as in the journal paper (JP4) included in Appendix A.4. 68 5.2. Evaluation of NB-PLC under different channel conditions 5.2.1. Influence of the spectral characteristics of the conducted emissions on NB-PLC This contribution analyzes the influence of the spectral characteristics of the conducted emissions generated by EVCPs on NB-PLC according to PRIME v1.4 standard. First, a characterization of the emissions in the frequency domain is presented and, then, the performance of PRIME v1.4 in presence of these emissions is evaluated. For this purpose, a measurement campaign is performed (described in 4.2.3), where the emissions generated by a set of EVs are recorded at an EVCS connected to a LISN, and then reproduced in a controlled and previously well-characterized laboratory scenario (described in section 4.2.6). In order to evaluate the performance of Forward Error Correction (FEC) and repetition codes in the presence of sub-cycle variations, the study considers Differential Quadrature Shift Keying (DQPSK) with FEC (DQPSK_C), Differential Binary PSK with FEC (DBPSK_C), Robust DQPSK (R_DQPSK), and R_DBPSK. In all the measurements, 1000 Type B frames are transmitted considering a fixed 256-byte length message and frequency channels 1 and 3-8 of PRIME v1.4. The results related to the characterization of the emissions lead to conclude that the spectral pattern of the emissions is highly dependent on the frequency band under analysis. In channel 1 (42-89 kHz), several high-amplitude tonal or narrowband emissions at different frequencies are observed. By contrast, in channels 3-8 (151-471 kHz), flat background noise with low-amplitude tonal emissions are registered regardless of the EV model and charging current. The tonal emissions in the 42-89 kHz frequency range correspond, in most cases, to harmonics of the switching frequency of the inverter included in the circuitry of the EV charger [126]. The spectral features of these NIEs can be found in the conference papers related to this contribution (ICP3 and NCP1), which are included in Appendix B.1.3 and Appendix B.2.1, respectively. The evaluation results show that, due to the different spectral shape of the emissions depending on the frequency range, the thresholds that set the quality of communications vary considerably from one frequency channel to another. By contrast, no differences in the minimum SNR needed to achieve a FER of 5 % are reported for a given EV model and different charging currents, since similar spectral features are obtained in these cases. Moreover, a better performance of modulations including repetition codes (R_DQPSK and R_DBPSK) than modulations using only FEC (DQPSK_C, DBPSK_C) is shown. In general, the four modulations are capable of correcting tonal or narrowband emissions occurring at certain frequencies, providing a minimum SNR to achieve a FER of 5 % lower than the threshold corresponding to AWGN. However, when background noise with multiple low-amplitude emissions is measured, as presented in channels 3-8 for all EVCPs under study, the thresholds are practically identical to the AWGN case. The only exception to this is EV1 with charging currents of 8 A and 12 A using DQPSK_C and DBPSK_C modulations, where there is a large number of emissions with small frequency gaps 69 between them, which results into additional degradation with respect to the AWGN case. Therefore, the effect of tonal emissions on NB-PLC is determined by the number of emissions, their amplitude with respect to the background noise and the frequency gap between consecutive peaks. The FER-SNR curves presented in this contribution take into consideration the spectral pattern of the emissions and not their actual amplitude. However, since the emissions in channel 1 are significantly higher than the emissions in channels 3-8, the minimum received signal level needed to obtain a FER of 5 % in channel 1 could be higher than the corresponding to channels 3-8. In order to demonstrate this behavior, as an example, the RSSI needed for achieving a FER of 5 % is compared for EV2 with a charging current of 8 A with DBPSK_C modulation in channels 1 and 3. While in channel 1 this parameter takes a value of 83 dBµV, only 60 dBµV are required to obtain the same value of FER in channel 3. However, due to the spectral pattern of the emissions, the SNR required for obtaining a FER of 5 % is higher in channel 3 (5.9 dB in channel 3 compared to 4.9 dB in channel 1). Therefore, with the aim of completely evaluating the influence of the conducted emissions generated by EVCPs in terms of their spectral pattern, actual amplitude of the NIEs and their time-dependent behavior, future work should combine the minimum SNR and RSSI required for obtaining a FER of 5 %. The conference papers related to this contribution (ICP3 and NCP1) can be found in Appendix B.1.3 and Appendix B.2.1. 5.2.2. Influence of the grid impedance variations Influence of frequency-dependent impedance variations This contribution addresses the evaluation of the influence of the frequency-dependent grid impedance on NB-PLC according to PRIME v1.4. The study is carried out in the laboratory setup described in 4.2.7. In this study, frequency channels 1 and 3-8 defined in PRIME v1.4 are evaluated by transmitting 1000 frames with a 256-byte length using a DBPSK modulation. In order to evaluate the influence of frequency-dependent impedance variations on PRIME v1.4, four EMC filters designed according to IEC 60939 are considered, which show important frequency-dependent impedance variations that remain constant over time. These filters, which are used for electromagnetic interference suppression, perform a double function. First, the filter protects an electronic control circuit from voltage spikes in the mains supply, which may be generated, for example, by electromechanical switches and relays. Simultaneously, the same filter also acts in the opposite direction, attenuating the interference generated in the unit towards the power supply line. However, the use of EMC filters might also lead to an undesired effect on NB-PLC signals. Due to the parallel and/or series resonances of EMC filters, notching effects can affect communication signals, which are more likely to be disturbed when the output impedance of the filter is low in comparison with the impedance of the grid. This contribution considers both impedance variations due to the EMC filters in an open-circuit configuration, i.e., when no external load is connected, and in a loaded 70 configuration, when a 50 Ω resistor in series with a 560 nF capacitor is connected to the filter. Fig. 29 and Fig. 30 show the modulus of the impedance measured when the four EMC filters are connected to the setup without and with a load connected in L’-N’ ports, respectively. Fig. 29. Impedance modulus of the four EMC filters connected to the setup without connecting a load in L’-N’ ports. Fig. 30. Impedance modulus of the four EMC filters connected to the setup when a load is connected in L’-N’ ports. Fig. 29 and Fig. 30 reveal that these filters present a wide range of impedance values along the frequency range from 10 kHz to 500 kHz, mainly due to the resonance effects they cause at specific frequencies. The variations with respect to the impedance of the LISN 71 shown in Fig. 12 are the result of the combination of the impedance of each EMC filter and the components of the LISN. The analysis of the schematics of the filters lead to conclude that the impedance variations introduced by the EMC filters are mainly due to the capacitive interface of the filters with the network. Regarding the evaluation of the performance of PRIME v1.4 in the presence of frequency-dependent impedance variations, as previously described in section 4.2.7, the study is based on the calculation of two parameters: the attenuation due to the EMC filters under test and the potential degradation of the FER-SNR curves that set the quality of the communications. Concerning the signal attenuation, the results presented in this study show that the impedance variations due to the connection of the EMC filters result in high attenuation in the frequency band assigned to NB-PLC (see Fig. 31). According to the performed trials, the highest attenuation is reported in those frequency channels in which the modulus of the grid impedance is low and maintained in the frequency channel (low standard deviation) and in which the phase presents abrupt variations. These attenuations could cause communications to fail if the noise level is high or if the received signal power is close to the sensitivity limit of the receiving equipment. Fig. 31. Attenuation due to the EMC filters under study in open-circuit configuration. Blue lines indicate that the filters are connected at point A, red lines indicate connection point B, and green lines indicate connection point C. However, the results also show that the variations in the channel frequency response due to the impedance variations because of the EMC filters are not selective enough to degrade NB-PLC in terms of the SNR thresholds that set the quality of the communications (see Fig. 32). This involves that PRIME v1.4 performs properly under these circumstances, not being affected by the channel characteristics caused by the introduction of the EMC filters. 78 6. Impact The results of this Doctoral Thesis have been published in scientific journals indexed in the Journal Citation Reports (JCR), as well as in international and national conferences. In Table V, the journal papers associated to the contributions of this Doctoral Thesis are gathered. Six of these publications have already been published, while one of them is under review at the time of publication of this thesis. Table V. Summary of the journal papers associated to the contributions of this Doctoral Thesis. Publication code Title Publisher Journal Year IF/Quartile JP1 [42] A Review on the Empirical Characterization of the Low Voltage Distribution Grid as a Communication Channel for Power Line Communications Elsevier Sustainable Energy, Grids and Networks 2023 4.8/Q1 JP2 [43] Emissions Generated by Electric Vehicles in the 9-500 kHz Band: Characterization, Propagation, and Interaction Elsevier Electric Power Systems Research 2024 3.3/Q2 JP3 [44] Comparison of Conducted Emissions Due to Electric Vehicle Charging Processes under Isolated and On-Line Conditions in the 9-500 kHz Frequency Range Elsevier Sustainable Energy, Grids and Networks 2024 4.8/Q1 JP4 Characterization of the Long-Term Impedance Variations due to Electric Vehicle Charging from 20 kHz to 500 kHz IEEE IEEE Open Journal of Power and Energy Under Review 3.3/Q2 JP5 [45] Characterization of the Potential Effects of EMC Filters for Power Converters on Narrowband Power Line Communications MDPI Electronics 2021 2.69/Q3 JP6 [46] Characterization of the Potential Effects of Sub-Cycle Impedance Variations on PRIME v1.4 Elsevier Engineering Science and Technology, an International Journal 2024 5.1/Q1 JP7 [47] Upgrading the Power Grid Functionalities with Broadband Power Line Communications: Basis, Applications, Current Trends and Challenges MDPI Sensors 2022 3.9/Q2 In Table VI and Table VII, a summary of the papers presented in international and national conferences during this Doctoral Thesis are detailed. 79 Table VI. Summary of the international conference papers associated to the contributions of this Doctoral Thesis. Publication code Title Conference Year ICP1 [224] Empirical characterization of the conducted disturbances generated by the electric vehicles during the charging process CIRED Porto Workshop 2022 2022 ICP2 [225] Influence of Electric Vehicle Charging on the Grid Access Impedance from 20 kHz to 500 kHz 2023 International Conference on Smart Energy Systems and Technologies (SEST) 2023 ICP3 [226] Influence of the Spectral Pattern of the Conducted Emissions Generated by Electric Vehicle Charging on PRIME v1.4 IEEE International Conference on Communications, Control, and Computing Technologies for Smart Grids (SmartGridComm) 2024 ICP4 Characterization of the Potential Effects of EMC Filters on Power Line Communications CIGRE Centennial Session Paris 2021 ICP5 Evaluation of PRIME v1.4 in a Reconstructed Low Voltage Grid 14th Workshop for Powerline Communications (WSPLC) 2023 ICP6 BB-PLC over LV networks – a step forward towards Smart Grid implementation ICSC-CITIES: V Ibero-American Congress of Smart Cities 2022 ICP7 [220] Characterization of the LV distribution grid for the deployment of a pilot BB-PLC network 2023 International Symposium on Power Line Communications and its Applications (ISPLC) 2023 Table VII. Summary of the national conference papers associated to the contributions of this Doctoral Thesis. Publication code Title Conference Year NCP1 Análisis del impacto de las emisiones conducidas generadas por la carga del vehículo eléctrico en las tecnologías NB-PLC II Congreso de Redes Inteligentes FutuRed 2023 NCP2 Caracterización de la red eléctrica como medio de transmisión de servicios de banda ancha (BB-PLC) II Congreso de Redes Inteligentes FutuRed 2023 It should also be noted that the following contributions of this Doctoral Thesis have been included in the 4th Study Report of the WG CLC/TC57-219/WG11 of CENELEC: - Emissions from EV chargers. - Propagation of the emissions generated during EV charging. - Impact of the EV charging process on the impedance of the distribution network. 80 - Impedance characteristics of EMI filters. - Attenuation of MCE signals caused by EMI filters. - Sub-cycle variations of the impedance. Finally, this Doctoral Thesis has contributed to the chapter “EMC with communications systems” of the technical brochure of the CIGRE WG C4.68 “Electromagnetic Compatibility issues in modern and future power systems”, that will be published during 2025. 81 7. Bibliography [1] M. Seijo Simó, G. López, J. Matanza, and J. Moreno, “Planning and Performance Challenges in Power Line Communications Networks for Smart Grids,” Int J Distrib Sens Netw, vol. 2016, Mar. 2016, doi: 10.1155/2016/8924081. [2] P. Mlýnek, M. Rusz, L. Benešl, J. Sláčik, and P. Musil, “Possibilities of Broadband Power Line Communications for Smart Home and Smart Building Applications,” Sensors, vol. 21, no. 1, 2021, doi: 10.3390/s21010240. [3] I. Colak, “Introduction to smart grid,” in 2016 International Smart Grid Workshop and Certificate Program (ISGWCP), 2016, pp. 1–5. doi: 10.1109/ISGWCP.2016.7548265. [4] H. Gharavi and R. Ghafurian, “Smart Grid: The Electric Energy System of the Future [Scanning the Issue],” Proceedings of the IEEE, vol. 99, no. 6, pp. 917–921, 2011, doi: 10.1109/JPROC.2011.2124210. [5] G. Hallak and G. Bumiller, “Time variant voltage and current modeling corresponding to access impedance measurements,” in 2018 IEEE International Symposium on Power Line Communications and its Applications (ISPLC), 2018, pp. 1–6. doi: 10.1109/ISPLC.2018.8360226. [6] D. Baimel, S. Tapuchi, and N. Baimel, “Smart grid communication technologiesoverview, research challenges and opportunities,” in 2016 International Symposium on Power Electronics, Electrical Drives, Automation and Motion (SPEEDAM), 2016, pp. 116– 120. doi: 10.1109/SPEEDAM.2016.7526014. [7] A. Majumder and J. J. Caffery, “Power line communications,” IEEE Potentials, vol. 23, no. 4, pp. 4–8, 2004, doi: 10.1109/MP.2004.1343222. [8] A. Sendin et al., “Adaptation of Powerline Communications-Based Smart Metering Deployments to the Requirements of Smart Grids,” Energies (Basel), vol. 8, no. 12, pp. 13481–13507, 2015, doi: 10.3390/en81212372. [9] J. Slacik, P. Mlynek, M. Rusz, P. Musil, L. Benesl, and M. Ptacek, “Broadband Power Line Communication for Integration of Energy Sensors within a Smart City Ecosystem,” Sensors, vol. 21, no. 10, 2021, doi: 10.3390/s21103402. [10] P. Mlynek, J. Misurec, Z. Kolka, J. Slacik, and R. Fujdiak, “Narrowband Power Line Communication for Smart Metering and Street Lighting Control,” IFACPapersOnLine, vol. 48, no. 4, pp. 215–219, 2015, doi: https://doi.org/10.1016/j.ifacol.2015.07.035. [11] A. Llano, D. De La Vega, I. Angulo, and L. Marron, “Impact of Channel Disturbances on Current Narrowband Power Line Communications and Lessons to Be Learnt for the Future Technologies,” IEEE Access, vol. 7, pp. 83797–83811, 2019, doi: 10.1109/ACCESS.2019.2924806. [12] I. Elfeki, S. Jacques, I. Aouichak, T. Doligez, Y. Raingeaud, and J.-C. Le Bunetel, “Characterization of Narrowband Noise and Channel Capacity for Powerline 82 Communication in France,” Energies (Basel), vol. 11, no. 11, 2018, doi: 10.3390/en11113022. [13] M. Antoniali and A. M. Tonello, “Measurement and Characterization of Load Impedances in Home Power Line Grids,” IEEE Trans Instrum Meas, vol. 63, no. 3, pp. 548–556, 2014, doi: 10.1109/TIM.2013.2280490. [14] J. Anatory, N. Theethayi, R. Thottappillil, M. M. Kissaka, and N. H. Mvungi, “The Influence of Load Impedance, Line Length, and Branches on Underground Cable Power-Line Communications (PLC) Systems,” IEEE Transactions on Power Delivery, vol. 23, no. 1, pp. 180–187, 2008, doi: 10.1109/TPWRD.2007.911020. [15] D. Thomas, “Conducted emissions in distribution systems (1 kHz–1 MHz),” IEEE Electromagn Compat Mag, vol. 2, no. 2, pp. 101–104, 2013, doi: 10.1109/MEMC.2013.6550941. [16] I. Fernández, D. de la Vega, A. Arrinda, I. Angulo, N. Uribe-Pérez, and A. Llano, “Field Trials for the Characterization of Non-Intentional Emissions at Low-Voltage Grid in the Frequency Range Assigned to NB-PLC Technologies,” Electronics (Basel), vol. 8, no. 9, 2019, doi: 10.3390/electronics8091044. [17] T. Slangen, T. van Wijk, V. Ćuk, and S. Cobben, “The Propagation and Interaction of Supraharmonics from Electric Vehicle Chargers in a Low-Voltage Grid,” Energies (Basel), vol. 13, no. 15, 2020, doi: 10.3390/en13153865. [18] T. Streubel, C. Kattmann, A. Eisenmann, and K. Rudion, “Detection and Monitoring of Supraharmonic Anomalies of an Electric Vehicle Charging Station,” in 2019 IEEE Milan PowerTech, 2019, pp. 1–5. doi: 10.1109/PTC.2019.8810596. [19] “CENELEC SC 205A. CLC/TR 50669. Investigation results on electromagnetic interference in the frequency range below 150 kHz,” 2017. [20] N. Uribe-Pérez, I. Angulo, L. Hernández-Callejo, T. Arzuaga, D. la Vega, and A. Arrinda, “Study of Unwanted Emissions in the CENELEC-A Band Generated by Distributed Energy Resources and Their Influence over Narrow Band Power Line Communications,” Energies (Basel), vol. 9, no. 12, 2016, doi: 10.3390/en9121007. [21] P. T. Jensen and P. Davari, “Power Converter Impedance and Emission Characterization Below 150 kHz,” in 2021 IEEE International Joint EMC/SI/PI and EMC Europe Symposium, 2021, pp. 255–260. doi: 10.1109/EMC/SI/PI/EMCEurope52599.2021.9559177. [22] M. Yeshalem, B. Khan, and O. Mahela, “Conducted electromagnetic emissions of compact fluorescent lamps and electronic ballast modeling,” AIMS Electronics and Electrical Engineering, vol. 6, pp. 178–187, Apr. 2022, doi: 10.3934/electreng.2022011. [23] J. Meyer, S. Haehle, and P. Schegner, “Impact of higher frequency emission above 2kHz on electronic mass-market equipment,” in 22nd International Conference and Exhibition on Electricity Distribution (CIRED 2013), 2013, pp. 1–4. doi: 10.1049/cp.2013.1027. 83 [24] P. Kotsampopoulos et al., “EMC Issues in the Interaction Between Smart Meters and Power-Electronic Interfaces,” IEEE Transactions on Power Delivery, vol. 32, no. 2, pp. 822–831, 2017, doi: 10.1109/TPWRD.2016.2561238. [25] P. De Falco and P. Varilone, “Statistical Characterization of Supraharmonics in LowVoltage Distribution Networks,” Applied Sciences, vol. 11, no. 8, 2021, doi: 10.3390/app11083574. [26] S. T. Y. Alfalahi et al., “Supraharmonics in Power Grid: Identification, Standards, and Measurement Techniques,” IEEE Access, vol. 9, pp. 103677–103690, 2021, doi: 10.1109/ACCESS.2021.3099013. [27] A. A. Alkahtani et al., “Power Quality in Microgrids Including Supraharmonics: Issues, Standards, and Mitigations,” IEEE Access, vol. 8, pp. 127104–127122, 2020, doi: 10.1109/ACCESS.2020.3008042. [28] “ETSI TS 103 909 V1.1.1. Power Line Telecommunications (PLT) Narrow band transceivers in the range 9 kHz to 500 kHz Power Line Performance Test Method Guide. ,” Dec. 2012. [29] G. Hallak and G. Bumiller, “Impedance measurement of electrical equipment loads on the power line network,” in 2017 IEEE International Symposium on Power Line Communications and its Applications (ISPLC), 2017, pp. 1–6. doi: 10.1109/ISPLC.2017.7897099. [30] G. Hallak, G. Bumiller, and C. Nieß, “Accurate access impedance measurements on the power line with optimized calibration procedures,” in 2017 IEEE International Instrumentation and Measurement Technology Conference (I2MTC), 2017, pp. 1–6. doi: 10.1109/I2MTC.2017.7969803. [31] C. Nieß, J.-P. Kitzig, and G. Bumiller, “Measurement System for Time Variable Subcycle Impedance on Power Lines,” in 2021 IEEE International Instrumentation and Measurement Technology Conference (I2MTC), 2021, pp. 1–6. doi: 10.1109/I2MTC50364.2021.9459875. [32] I. Fernández, M. Alberro, J. Montalbán, A. Arrinda, I. Angulo, and D. de la Vega, “A new voltage probe with improved performance at the 10 kHz–500 kHz frequency range for field measurements in LV networks,” Measurement, vol. 145, pp. 519–524, 2019, doi: https://doi.org/10.1016/j.measurement.2019.05.106. [33] G. Hallak, C. Nieß, and G. Bumiller, “Accurate Low Access Impedance Measurements With Separated Load Impedance Measurements on the Power-Line Network,” IEEE Trans Instrum Meas, vol. 67, no. 10, pp. 2282–2293, 2018, doi: 10.1109/TIM.2018.2814138. [34] F. J. C. Corripio, J. A. C. Arrabal, L. D. del Rio, and J. T. E. Munoz, “Analysis of the cyclic short-term variation of indoor power line channels,” IEEE Journal on Selected Areas in Communications, vol. 24, no. 7, pp. 1327–1338, 2006, doi: 10.1109/JSAC.2006.874402. 84 [35] L. G. da S. Costa et al., “Access impedance in Brazilian in-home, broadband and lowvoltage electric power grids,” Electric Power Systems Research, vol. 171, pp. 141–149, 2019, doi: https://doi.org/10.1016/j.epsr.2019.02.015. [36] I. Fernández, A. Arrinda, I. Angulo, D. D. La Vega, N. Uribe-Pérez, and A. Llano, “Field Trials for the Empirical Characterization of the Low Voltage Grid Access Impedance From 35 kHz to 500 kHz,” IEEE Access, vol. 7, pp. 85786–85795, 2019, doi: 10.1109/ACCESS.2019.2924253. [37] P. A. C. Lopes, J. M. M. Pinto, and J. B. Gerald, “Dealing With Unknown Impedance and Impulsive Noise in the Power-Line Communications Channel,” IEEE Transactions on Power Delivery, vol. 28, no. 1, pp. 58–66, 2013, doi: 10.1109/TPWRD.2012.2214065. [38] S. Rönnberg, M. Lundmark, M. Wahlberg, M. Andersson, A. Larsson, and M. Bollen, “Attenuation and Noise Level - Potential Problems with Communication via the Power Grid,” in 19th International Conference on Electricity Distribution, Vienna, May 2007. [39] M. Antoniali, A. Tonello, and F. Versolatto, “A Study on the Optimal Receiver Impedance for SNR Maximization in Broadband PLC,” Journal of Electrical and Computer Engineering, vol. 2013, Mar. 2013, doi: 10.1155/2013/635086. [40] S. Cassano, F. Silvestro, E. D. Jaeger, and C. Leroi, “Modeling of harmonic propagation of fast DC EV charging station in a Low Voltage network,” in 2019 IEEE Milan PowerTech, 2019, pp. 1–6. doi: 10.1109/PTC.2019.8810969. [41] S. K. Ronnberg, M. H. J. Bollen, A. Larsson, and M. Lundmark, “An overview of the origin and propagation of Supraharmonics (2-150 kHz),” 2014. [42] J. González-Ramos, A. Gallarreta, I. Angulo, I. Fernández, A. Arrinda, and D. de la Vega, “A review on the empirical characterization of the low voltage distribution grid as a communication channel for power line communications,” Sustainable Energy, Grids and Networks, vol. 36, p. 101217, 2023, doi: https://doi.org/10.1016/j.segan.2023.101217. [43] J. González-Ramos, A. Gallarreta, I. Fernández, I. Angulo, D. de la Vega, and A. Arrinda, “Emissions generated by electric vehicles in the 9-500 kHz band: Characterization, propagation, and interaction,” Electric Power Systems Research, vol. 231, p. 110289, 2024, doi: https://doi.org/10.1016/j.epsr.2024.110289. [44] J. González-Ramos, A. Gallarreta, I. Fernández, I. Angulo, D. de la Vega, and A. Arrinda, “Comparison of conducted emissions due to electric vehicle charging processes under isolated and on-line conditions in the 9–500 kHz frequency range,” Sustainable Energy, Grids and Networks, vol. 38, p. 101333, 2024, doi: https://doi.org/10.1016/j.segan.2024.101333. [45] J. González-Ramos, I. Angulo, I. Fernández, A. Arrinda, and D. de la Vega, “Characterization of the Potential Effects of EMC Filters for Power Converters on Narrowband Power Line Communications,” Electronics (Basel), vol. 10, no. 2, 2021, doi: 10.3390/electronics10020152. 85 [46] J. González–Ramos, I. Angulo, I. Fernández, A. Gallarreta, D. de la Vega, and A. Arrinda, “Characterization of the potential effects of sub-cycle impedance variations on PRIME v1.4,” Engineering Science and Technology, an International Journal, vol. 56, p. 101775, 2024, doi: https://doi.org/10.1016/j.jestch.2024.101775. [47] J. González-Ramos et al., “Upgrading the Power Grid Functionalities with Broadband Power Line Communications: Basis, Applications, Current Trends and Challenges,” Sensors, vol. 22, no. 12, 2022, doi: 10.3390/s22124348. [48] S. Galli, A. Scaglione, and Z. Wang, “Power Line Communications and the Smart Grid,” in 2010 First IEEE International Conference on Smart Grid Communications, 2010, pp. 303–308. doi: 10.1109/SMARTGRID.2010.5622060. [49] A. R. Ndjiongue and H. Ferreira, “Power Line Communications (PLC) Technology: More Than 20 Years of Intense Research,” Transactions on Emerging Telecommunications Technologies, pp. 1–20, Jul. 2019, doi: 10.1002/ett.3575. [50] G. López et al., “The Role of Power Line Communications in the Smart Grid Revisited: Applications, Challenges, and Research Initiatives,” IEEE Access, vol. 7, pp. 117346–117368, 2019, doi: 10.1109/ACCESS.2019.2928391. [51] L. Berger, A. Schwager, and J. J. Escudero-Garzas, “Power Line Communications for Smart Grid Applications,” Journal of Electrical and Computer Engineering, vol. 2013, Jan. 2013, doi: 10.1155/2013/712376. [52] P. Chadwick, “CENELEC standards for occupatonal EMF exposure.” Accessed: Sep. 18, 2023. [Online]. Available: https://www.icnirp.org/cms/upload/presentations/Joint/ChadwickP.pdf [53] C. Cano, A. Pittolo, D. Malone, L. Lampe, A. M. Tonello, and A. G. Dabak, “State of the Art in Power Line Communications: From the Applications to the Medium,” IEEE Journal on Selected Areas in Communications, vol. 34, no. 7, pp. 1935–1952, 2016, doi: 10.1109/JSAC.2016.2566018. [54] M. Wasowski et al., “Sources of Non-Intentional Supraharmonics in LV Network and Its Impact on OSGP PLC Communication – Experimental Study,” IEEE Transactions on Power Delivery, vol. 37, no. 6, pp. 5244–5254, 2022, doi: 10.1109/TPWRD.2022.3175090. [55] B. Masood and S. Baig, “Standardization and deployment scenario of next generation NB-PLC technologies,” Renewable and Sustainable Energy Reviews, vol. 65, pp. 1033– 1047, 2016, doi: https://doi.org/10.1016/j.rser.2016.07.060. [56] M. Sanz, J. I. Moreno, G. López, J. Matanza, and J. Berrocal, “Web-Based Toolkit for Performance Simulation and Analysis of Power Line Communication Networks,” Energies (Basel), vol. 14, no. 20, 2021, doi: 10.3390/en14206475. [57] “PRIME Alliance, ‘PRIME v1.4 White Paper’, PRIME Alliance,” Brussels, 2014. Accessed: Sep. 18, 2023. [Online]. Available: https://www.primealliance.org/media/2020/04/whitePaperPrimeV1p4_final.pdf 86 [58] “ITU-T Rec. G9904; Narrowband orthogonal frequency division multiplexing power line communication transceivers for PRIME networks,” 2012. [59] “G3 PLC Alliance, ‘Narrowband OFDM PLC specifications for G3 PLC networks’, G3 PLC Alliance,” Paris, 2021. [60] “ITU-T Rec. G9903; Narrowband orthogonal frequency division multiplexing power line communication transceivers for G3 PLC networks,” 2012. [61] “CLC/TS 50568 4:2015; Electricity metering data exchange Part 4: Lower layer PLC profile using SMITP B PSK modulation,” 2015. [62] “CLC/TS 50590:2015; Electricity metering data exchange. Lower layer PLC profile using Adaptative Multi Carrier Spread Spectrum (AMC SS) modulation,” 2015. [63] “ISO/IEC 14908 1:2012; Information technology – Control network protocol – Part 1: Protocol stack,” 2012. [64] “ITU T Rec. G9902; Narrowband orthogonal frequency division multiplexing power line communication transceivers for ITU T G,hnem networks,” 2012. [65] “‘IEEE Standard for Low-Frequency (less than 500 kHz) Narrowband Power Line Communications for Smart Grid Applications,’ in IEEE Std 1901.2-2013, vol., no., pp.1-269, 6 Dec. 2013, doi: 10.1109/IEEESTD.2013.6679210.” [66] “PRIME Alliance. Deployments & pilots.” Accessed: Oct. 03, 2023. [Online]. Available: https://www.prime-alliance.org/alliance/deployments/ [67] M. Carlsson, Smart Meter in North America. Accessed: Oct. 03, 2023. [Online]. Available: https://media.berginsight.com/2022/12/23180016/bi-smna5-ps.pdf [68] L. Lampe, A. Tonello, and T. Swart, Power line communications: Principles, standards and applications from multimedia to smart grid: Second edition. 2016. doi: 10.1002/9781118676684. [69] S. Galli, A. Scaglione, and Z. Wang, “For the Grid and Through the Grid: The Role of Power Line Communications in the Smart Grid,” Proceedings of the IEEE, vol. 99, no. 6, pp. 998–1027, 2011, doi: 10.1109/JPROC.2011.2109670. [70] J. Granado, A. Torralba, and C. Álvarez-Arroyo, “Partial discharge detection using PLC receivers in MV cables: A theoretical framework,” Electric Power Systems Research, vol. 164, pp. 61–69, 2018, doi: https://doi.org/10.1016/j.epsr.2018.07.030. [71] “IEEE Standard for Broadband over Power Line Networks: Medium Access Control and Physical Layer Specifications,” IEEE Std 1901-2020 (Revision of IEEE Std 19012010), pp. 1–1622, 2021, doi: 10.1109/IEEESTD.2021.9329263. [72] “Open PLC European Research Alliance for New Generation PLC Integrated Network Phase 2.” Accessed: Apr. 10, 2023. [Online]. Available: https://cordis. europa.eu/project/id/026920/es [73] “ITU-T Rec. G9960; Unified High-Speed Wire-Line Based Home Networking Transceivers—System Architecture and Physcial Layer Specification. 2018. .” 87 Accessed: Apr. 10, 2023. [Online]. Available: https://www.itu.int/rec/T-RECG.9960-201811-I/en [74] “HD-PLC Specifications. 4th Generation HD-PLC Quatro Core Overview, HDPLC Alliance. .” Accessed: Apr. 10, 2023. [Online]. Available: https://hd-plc.org/ wp-content/uploads/2021/01/HDPD-P0010E_White-Paper-HD-PLC4-6.pdf [75] “HomePlug AV Specification Version 2.0, HomePlug Powerline Alliance. 2013. .” Accessed: Apr. 10, 2023. [Online]. Available: https://web.archive.org/web/20 121103131751/http://www.homeplug.org/tech/whitepapers/HomePlug_AV2_W hite_Paper_v1.0.pdf [76] “Segalotto, J.-F. Towards a Truly Smart Grid—The Case for Broadband over Power Line. IDC White Paper, PRIME Alliance. ,” 2021. Accessed: Sep. 18, 2023. [Online]. Available: https://www.prime-alliance.org/media/2021/09/TOWARDS-ATRULY-SMART-GRID-Summary.pdf [77] International Electrotechnical Vocabulary (IEV) - Part 161: Electromagnetic Compatibility, “IEC 60050-161:1990.” Accessed: Oct. 04, 2023. [Online]. Available: https://webstore.iec.ch/publication/68718 [78] M. Klatt et al., “Emission levels above 2 kHz — Laboratory results and survey measurements in public low voltage grids,” in 22nd International Conference and Exhibition on Electricity Distribution (CIRED 2013), 2013, pp. 1–4. doi: 10.1049/cp.2013.1102. [79] I. Fernandez et al., “Characterization of non-intentional emissions from distributed energy resources up to 500 kHz: A case study in Spain,” International Journal of Electrical Power & Energy Systems, vol. 105, pp. 549–563, 2019, doi: https://doi.org/10.1016/j.ijepes.2018.08.048. [80] T. Yalcin, M. Özdemir, P. Kostyla, and Z. Leonowicz, “Analysis of supra-harmonics in smart grids,” in 2017 IEEE International Conference on Environment and Electrical Engineering and 2017 IEEE Industrial and Commercial Power Systems Europe (EEEIC / I&CPS Europe), 2017, pp. 1–4. doi: 10.1109/EEEIC.2017.7977812. [81] M. Bollen, M. Olofsson, A. Larsson, S. Rönnberg, and M. Lundmark, “Standards for supraharmonics (2 to 150 kHz),” IEEE Electromagn Compat Mag, vol. 3, no. 1, pp. 114–119, 2014, doi: 10.1109/MEMC.2014.6798813. [82] E. O. A. Larsson, M. H. J. Bollen, M. G. Wahlberg, C. M. Lundmark, and S. K. Rönnberg, “Measurements of High-Frequency (2–150 kHz) Distortion in LowVoltage Networks,” IEEE Transactions on Power Delivery, vol. 25, no. 3, pp. 1749–1757, 2010, doi: 10.1109/TPWRD.2010.2041371. [83] A. Sendin, I. Berganza, A. Arzuaga, A. Pulkkinen, and I. H. Kim, “Performance results from 100,000+ PRIME smart meters deployment in Spain,” in 2012 IEEE Third International Conference on Smart Grid Communications (SmartGridComm), 2012, pp. 145–150. doi: 10.1109/SmartGridComm.2012.6485974. [84] G. López, J. I. Moreno, E. Sánchez, C. Martínez, and F. Martín, “Noise Sources, Effects and Countermeasures in Narrowband Power-Line Communications 94 [153] A. A. M. Picorone, T. R. de Oliveira, R. Sampaio-Neto, M. Khosravy, and M. V Ribeiro, “Channel characterization of low voltage electric power distribution networks for PLC applications based on measurement campaign,” International Journal of Electrical Power & Energy Systems, vol. 116, p. 105554, 2020, doi: https://doi.org/10.1016/j.ijepes.2019.105554. [154] M. Zimmermann and K. Dostert, “A multipath model for the powerline channel,” IEEE Transactions on Communications, vol. 50, no. 4, pp. 553–559, 2002, doi: 10.1109/26.996069. [155] A. M. Tonello and F. Versolatto, “Bottom-Up Statistical PLC Channel Modeling— Part I: Random Topology Model and Efficient Transfer Function Computation,” IEEE Transactions on Power Delivery, vol. 26, no. 2, pp. 891–898, 2011, doi: 10.1109/TPWRD.2010.2096518. [156] A. M. Tonello and F. Versolatto, “Bottom-Up Statistical PLC Channel Modeling— Part II: Inferring the Statistics,” IEEE Transactions on Power Delivery, vol. 25, no. 4, pp. 2356–2363, 2010, doi: 10.1109/TPWRD.2010.2053561. [157] T. Esmailian, F. Kschischang, and P. Gulak, “In-building power lines as high-speed communication channels: Channel characterization and a test channel ensemble,” Int. J. Communication Systems, vol. 16, pp. 381–400, Jun. 2003, doi: 10.1002/dac.596. [158] D. Anastasiadou and T. Antonakopoulos, “Multipath characterization of indoor power-line networks,” IEEE Transactions on Power Delivery, vol. 20, no. 1, pp. 90–99, 2005, doi: 10.1109/TPWRD.2004.832373. [159] J. Anatory, N. Theethayi, and R. Thottappillil, “Power-Line Communication Channel Model for Interconnected Networks—Part I: Two-Conductor System,” IEEE Transactions on Power Delivery, vol. 24, no. 1, pp. 118–123, 2009, doi: 10.1109/TPWRD.2008.2005679. [160] T. Bostoen and O. Van de Wiel, “Modelling the low-voltage power distribution network in the frequency band from 0.5 MHz to 30 MHz for broadband powerline communications (PLC),” in 2000 International Zurich Seminar on Broadband Communications. Accessing, Transmission, Networking. Proceedings (Cat. No.00TH8475), 2000, pp. 171–178. doi: 10.1109/IZSBC.2000.829248. [161] F. J. Canete, J. A. Cortés, L. Díez, and J. T. Entrambasaguas, “A channel model proposal for indoor power line communications,” IEEE Communications Magazine, vol. 49, no. 12, pp. 166–174, 2011, doi: 10.1109/MCOM.2011.6094022. [162] S. Galli and T. C. Banwell, “A deterministic frequency-domain model for the indoor power line transfer function,” IEEE Journal on Selected Areas in Communications, vol. 24, no. 7, pp. 1304–1316, 2006, doi: 10.1109/JSAC.2006.874428. [163] F. Versolatto and A. M. Tonello, “An MTL Theory Approach for the Simulation of MIMO Power-Line Communication Channels,” IEEE Transactions on Power Delivery, vol. 26, no. 3, pp. 1710–1717, 2011, doi: 10.1109/TPWRD.2011.2126608. [164] J. Anatory, N. Theethayi, and R. Thottappillil, “Power-Line Communication Channel Model for Interconnected Networks—Part II: Multiconductor System,” IEEE 95 Transactions on Power Delivery, vol. 24, no. 1, pp. 124–128, 2009, doi: 10.1109/TPWRD.2008.2005681. [165] T. Sartenaer and P. Delogne, “Deterministic modeling of the (shielded) outdoor power line channel based on the multiconductor transmission line equations,” IEEE Journal on Selected Areas in Communications, vol. 24, no. 7, pp. 1277–1291, 2006, doi: 10.1109/JSAC.2006.874423. [166] H. Meng et al., “Modeling of transfer Characteristics for the broadband power line communication channel,” IEEE Transactions on Power Delivery, vol. 19, no. 3, pp. 1057–1064, 2004, doi: 10.1109/TPWRD.2004.824430. [167] E. G. Bakhoum, “${\rm S}$-Parameters Model for Data Communications Over 3Phase Transmission Lines,” IEEE Trans Smart Grid, vol. 2, no. 4, pp. 615–623, 2011, doi: 10.1109/TSG.2011.2168613. [168] T. Banwell and S. Galli, “A novel approach to the modeling of the indoor power line channel part I: circuit analysis and companion model,” IEEE Transactions on Power Delivery, vol. 20, no. 2, pp. 655–663, 2005, doi: 10.1109/TPWRD.2005.844326. [169] S. Galli and T. Banwell, “A novel approach to the modeling of the indoor power line channel-Part II: transfer function and its properties,” IEEE Transactions on Power Delivery, vol. 20, no. 3, pp. 1869–1878, 2005, doi: 10.1109/TPWRD.2005.848732. [170] R. Alaya and R. Attia, “Characterization of low voltage access network for narrowband powerline communications,” in 2017 25th International Conference on Software, Telecommunications and Computer Networks (SoftCOM), 2017, pp. 1–6. doi: 10.23919/SOFTCOM.2017.8115514. [171] A.-I. Chiuta and N. Secăreanu, “Theoretical postulation of PLC channel model,” Journal of Electrical and Electronics Engineering, vol. 2, May 2009. [172] A. Lazaropoulos, “Towards Modal Integration of Overhead and Underground LowVoltage and Medium-Voltage Power Line Communication Channels in the Smart Grid Landscape: Model Expansion, Broadband Signal Transmission Characteristics, and Statistical Performance Metrics (Invited Paper),” ISRN Signal Processing, vol. 2012, Oct. 2012, doi: 10.5402/2012/121628. [173] B. Masood, G. Song, S. Baig, M. Rasheed, and J. Hou, “Measurements and Characterization of Low and Medium Voltage Residential, Commercial and Industrial NB-PLC Networks for AMI,” IET Generation, Transmission and Distribution, vol. 14, Dec. 2020, doi: 10.1049/iet-gtd.2020.1233. [174] B. Masood, S. Guobing, R. A. Naqvi, M. B. Rasheed, J. Hou, and A. U. Rehman, “Measurements and channel modeling of low and medium voltage NB-PLC networks for smart metering,” IET Generation, Transmission & Distribution, vol. 15, no. 2, pp. 321–338, Jan. 2021, doi: https://doi.org/10.1049/gtd2.12023. [175] I. Fernández et al., “Characterization of the frequency-dependent transmission losses of the grid up to 500 kHz,” in 25th International Conference on Electricity Distribution, Madrid, Jun. 2019. 96 [176] T. Maenou and M. Katayama, “Study on Signal Attenuation Characteristics in Power Line Communications,” in 2006 IEEE International Symposium on Power Line Communications and Its Applications, 2006, pp. 217–221. doi: 10.1109/ISPLC.2006.247464. [177] D. Shao, Q. Wang, Y. Lu, Y. Shu, C. Lai, and K. Zhang, “Analyses and Modeling of Power Line Channel Attenuation Characteristics for Low Voltage Access Network in China,” in 2014 IEEE 80th Vehicular Technology Conference (VTC2014-Fall), 2014, pp. 1–5. doi: 10.1109/VTCFall.2014.6965873. [178] I. Tsiropoulos, P. Siskos, and P. Capros, “The cost of recharging infrastructure for electric vehicles in the EU in a climate neutrality context: Factors influencing investments in 2030 and 2050,” Appl Energy, vol. 322, p. 119446, 2022, doi: https://doi.org/10.1016/j.apenergy.2022.119446. [179] J. Matanza, S. Alexandres, and C. Rodriguez-Morcillo, “Performance evaluation of two narrowband PLC systems: PRIME and G3,” Comput Stand Interfaces, vol. 36, no. 1, pp. 198–208, 2013, doi: https://doi.org/10.1016/j.csi.2013.05.001. [180] K. Razazian, M. Umari, A. Kamalizad, V. Loginov, and M. Navid, “G3-PLC specification for powerline communication: Overview, system simulation and field trial results,” in ISPLC2010, 2010, pp. 313–318. doi: 10.1109/ISPLC.2010.5479881. [181] A. Sanz, D. Sancho, C. Guemes, and J. A. Cortés, “A physical layer model for G3PLC networks simulation,” in 2017 IEEE International Symposium on Power Line Communications and its Applications (ISPLC), 2017, pp. 1–6. doi: 10.1109/ISPLC.2017.7897116. [182] A. Van Laere, C. Wawrzyniak, S. Bette, and V. Moeyaert, “Development, validation and utilization of an ITU-T G.9903 PHY simulator for communication performance evaluation,” in 2016 International Symposium on Power Line Communications and its Applications (ISPLC), 2016, pp. 167–172. doi: 10.1109/ISPLC.2016.7476276. [183] B. Reynders, E. Van Lil, S. Pollin, and S. van den Bergh, “Comparison of the differential mode against the coherent mode in G3-PLC,” 2014. [Online]. Available: https://api.semanticscholar.org/CorpusID:57866622 [184] K. Razazian, M. Umari, and A. Kamalizad, “Error correction mechanism in the new G3-PLC specification for powerline communication,” in ISPLC2010, 2010, pp. 50– 55. doi: 10.1109/ISPLC.2010.5479944. [185] L. Di Bert, S. D’Alessandro, and A. M. Tonello, “A G3-PLC simulator for access networks,” in 18th IEEE International Symposium on Power Line Communications and Its Applications, 2014, pp. 99–104. doi: 10.1109/ISPLC.2014.6812329. [186] A. Atayero, A. Alatishe, and Y. Ivanov, “Power line communication technologies: Modeling and simulation of PRIME physical layer,” vol. 2, pp. 931–936, Oct. 2012. [187] P. Chebotayev, D. Urazayev, I. Pospelova, A. Karassenko, D. Zykov, and A. Shelupanov, “Analysis of the level of SNR in the PLC channel based on the results of mathematical and physical modelling,” J Phys Conf Ser, vol. 1145, no. 1, p. 12013, Jan. 2019, doi: 10.1088/1742-6596/1145/1/012013. 97 [188] K. Razazian, A. Kamalizad, M. Umari, Q. Qu, V. Loginov, and M. Navid, “G3-PLC field trials in U.S. distribution grid: Initial results and requirements,” in 2011 IEEE International Symposium on Power Line Communications and Its Applications, 2011, pp. 153– 158. doi: 10.1109/ISPLC.2011.5764382. [189] A. E. Dulay, R. I. R. Astillero, P. G. P. La Rosa, R. J. L. L. Tan, A. D. Tapang, and A. Santos, “Performance evaluation of G3 narrowband PLC standard for transmission through single and paired transformer,” in 2016 3rd MEC International Conference on Big Data and Smart City (ICBDSC), 2016, pp. 1–6. doi: 10.1109/ICBDSC.2016.7460333. [190] C. Males, V. Popa, A. Lavric, and I. Finis, “Performance evaluation of Power Line Communications over power transformers,” in 2012 20th Telecommunications Forum (TELFOR), 2012, pp. 627–630. doi: 10.1109/TELFOR.2012.6419288. [191] S. Ustun Ercan, “Power line Communication: Revolutionizing data transfer over electrical distribution networks,” Engineering Science and Technology, an International Journal, vol. 52, p. 101680, 2024, doi: https://doi.org/10.1016/j.jestch.2024.101680. [192] V. Blazek et al., “Error Analysis of Narrowband Power-Line Communication in the Off-Grid Electrical System,” Sensors, vol. 22, no. 6, 2022, doi: 10.3390/s22062265. [193] M. A. Wibisono, N. Moonen, and F. Leferink, “Interference of LED Lamps on Narrowband Power Line Communication,” in 2020 IEEE International Symposium on Electromagnetic Compatibility & Signal/Power Integrity (EMCSI), 2020, pp. 219–221. doi: 10.1109/EMCSI38923.2020.9191485. [194] A. H. Beshir et al., “Effects of the Switching Frequency of Random Modulated Power Converter on the G3 Power Line Communication System,” in 2022 International Symposium on Electromagnetic Compatibility – EMC Europe, 2022, pp. 778–782. doi: 10.1109/EMCEurope51680.2022.9901074. [195] W. El Sayed, P. Crovetti, N. Moonen, P. Lezynski, R. Smolenski, and F. Leferink, “Electromagnetic Interference of Spread-Spectrum Modulated Power Converters in G3-PLC Power Line Communication Systems,” IEEE Letters on Electromagnetic Compatibility Practice and Applications, vol. 3, no. 4, pp. 118–122, 2021, doi: 10.1109/LEMCPA.2021.3121091. [196] W. El Sayed, H. Loschi, M. A. Wibisono, N. Moonerr, P. Lezynski, and R. Smolenski, “The Influence of Spread-Spectrum Modulation on the G3-PLC Performance,” in 2021 Asia-Pacific International Symposium on Electromagnetic Compatibility (APEMC), 2021, pp. 1–4. doi: 10.1109/APEMC49932.2021.9597060. [197] W. Elsayed, P. Lezynski, R. Smolenski, A. Madi, M. Pazera, and A. Kempski, “Deterministic vs. Random Modulated Interference on G3 Power Line Communication,” Energies (Basel), vol. 14, Jun. 2021, doi: 10.3390/en14113257. [198] L. Da Rocha Farias, L. F. Monteiro, M. O. Leme, and S. L. Stevan, “Empirical Analysis of the Communication in Industrial Environment Based on G3-Power Line Communication and Influences from Electrical Grid,” Electronics (Basel), vol. 7, no. 9, 2018, doi: 10.3390/electronics7090194. 98 [199] A. Sendin, I. Berganza, A. Arzuaga, A. Pulkkinen, and I. H. Kim, “Performance results from 100,000+ PRIME smart meters deployment in Spain,” in 2012 IEEE Third International Conference on Smart Grid Communications (SmartGridComm), 2012, pp. 145–150. doi: 10.1109/SmartGridComm.2012.6485974. [200] A. Sendin, I. Berganza, A. Arzuaga, X. Osorio, I. Urrutia, and P. Angueira, “Enhanced Operation of Electricity Distribution Grids Through Smart Metering PLC Network Monitoring, Analysis and Grid Conditioning,” Energies (Basel), vol. 6, no. 1, pp. 539–556, 2013, doi: 10.3390/en6010539. [201] I. Arechalde, M. Castro, I. García-Borreguero, A. Sendín, I. Urrutia, and A. Fernandez, “Performance of PLC communications in frequency bands from 150 kHz to 500 kHz,” in 2017 IEEE International Symposium on Power Line Communications and its Applications (ISPLC), 2017, pp. 1–5. doi: 10.1109/ISPLC.2017.7897123. [202] A. Mengi, M. Waechter, and M. Koch, “500 kHz G3-PLC access technology for the roll-outs in Germany,” in 18th IEEE International Symposium on Power Line Communications and Its Applications, 2014, pp. 179–183. doi: 10.1109/ISPLC.2014.6812358. [203] P. S. Sausen, A. Sausen, M. De Campos, L. F. Sauthier, A. C. Oliveira, and R. R. Emmel Júnior, “Power Line Communication Applied in a Typical Brazilian Urban Power Network,” IEEE Access, vol. 9, pp. 72844–72856, 2021, doi: 10.1109/ACCESS.2021.3078697. [204] A. Aruzuaga, I. Berganza, A. Sendin, M. Sharma, and B. Varadarajan, “PRIME Interoperability Tests and Results from Field,” in 2010 First IEEE International Conference on Smart Grid Communications, 2010, pp. 126–130. doi: 10.1109/SMARTGRID.2010.5622029. [205] P. Mlynek, M. Koutny, J. Misurec, and Z. Kolka, “Measurements and evaluation of PLC modem with G3 and PRIME standards for Street Lighting Control,” in 18th IEEE International Symposium on Power Line Communications and Its Applications, 2014, pp. 238–243. doi: 10.1109/ISPLC.2014.6812318. [206] J. Slacik, P. Mlynek, R. Fujdiak, J. Misurec, and P. Silhavy, “Performance evaluation of multi-carrier power line communication systems in real condition for smart grid neighborhood area networks,” in 2017 9th International Congress on Ultra Modern Telecommunications and Control Systems and Workshops (ICUMT), 2017, pp. 391–397. doi: 10.1109/ICUMT.2017.8255157. [207] N. Uribe-Pérez, I. Angulo, D. de la Vega, T. Arzuaga, A. Arrinda, and I. Fernández, “On-field evaluation of the performance of IP-based data transmission over narrowband PLC for smart grid applications,” International Journal of Electrical Power & Energy Systems, vol. 100, pp. 350–364, 2018, doi: https://doi.org/10.1016/j.ijepes.2018.02.030. [208] I. Berganza, A. Sendin, R. Ayala, and J. S. Gomez, “Low voltage as the final frontier for broadband over power line,” in 27th International Conference on Electricity Distribution (CIRED 2023), 2023, pp. 3804–3810. doi: 10.1049/icp.2023.0717. 99 [209] A. Sendin, P. Losada-Sanisidro, J. García-Rodríguez, P. González-Méndez, and I. Berganza, “Broadband Over Power Line Communication Prototype Development for Next Generation Smart Meters: Validation in Access Electric Power Distribution Networks,” IEEE Access, vol. 12, pp. 30191–30208, 2024, doi: 10.1109/ACCESS.2024.3367987. [210] “ETSI Technical Committee. Power Line Telecommunications (PLT) Narrow Band Transceivers in the Range 9 kHz to500 kHz Power Line Performance Test Method Guide ETSI Technical Committee Powerline Communications TechnicalSpecification; ETSI TS 103 909 V1.1.1 ,” Sophia Antipolis, Dec. 2012. [211] “International Electrotechnical Commission, ‘IEC 61557-3:2007. Electrical safety in low voltage distribution systems up to 1000 Va.c. and 1500 Vd.c. - Equipment for testing, measuring or monitoring of protective measures - Part 3: Loop impedance,’” 2007. [212] “International Electrotechnical Commission, ‘IEC 61000-4-19:2015. Electromagnetic compatibility (EMC). Part 4-19: Testing and measurement techniques: Test for immunity to conducted, differential mode disturbances and signaling in the frequency range 2 kHz to 150 kHz at a.c. power ports,’” 2015. [213] I. Fernández et al., “Measurement System of the Mean and Sub-cycle LV Grid Access Impedance from 20 kHz to 10 MHz,” IEEE Transactions on Power Delivery, pp. 1–9, 2023, doi: 10.1109/TPWRD.2023.3238647. [214] I. Fernández et al., “Comparison of Measurement Methods of LV Grid Access Impedance in the Frequency Range Assigned to Nb-Plc Technologies,” Electronics (Basel), vol. 8, no. 10, 2019, doi: 10.3390/electronics8101155. [215] “International Electrotechnical Commission, ‘IEC 61851-1:2017. Electric vehicle conductive charging system - Part 1: General requirements’, ,” 2017. [216] HAMEG Instruments. A Rohde & Schwarz Company, “HM6050-2 Line Impedance Stabilization Network. Technical Data/Insertion Loss.” Accessed: Jun. 26, 2023. [Online]. Available: https://scdn.rohdeschwarz.com/ur/pws/dl_downloads/dl_common_library/dl_manuals/gb_1/h/h m6050_2/HM6050-2_UserManual_de_en_04.pdf [217] “Polylux PD4000 specifications.” Accessed: Jan. 22, 2024. [Online]. Available: https://polylux.com/images/files/es/pd4000.pdf [218] “Microchip PL360G55CF Evaluation Board.” Accessed: Jun. 26, 2023. [Online]. Available: https://www.microchip.com/en-us/developmenttool/ PL360G55CFEK# [219] Ettus Research, “USRP N200/N210 Networked Series.” Accessed: Jun. 26, 2023. [Online]. Available: https://www.ettus.com/wpcontent/uploads/2019/01/07495_Ettus_N200-210_DS_Flyer_HR_1.pdf [220] J. González-Ramos et al., “Characterization of the LV distribution grid for the deployment of a pilot BB-PLC network,” in 2023 IEEE International Symposium on 100 Power Line Communications and its Applications (ISPLC), 2023, pp. 19–24. doi: 10.1109/ISPLC57122.2023.10104185. [221] D. Chakravorty, J. Meyer, P. Schegner, S. Yanchenko, and M. Schocke, “Impact of Modern Electronic Equipment on the Assessment of Network Harmonic Impedance,” IEEE Trans Smart Grid, vol. 8, no. 1, pp. 382–390, 2017, doi: 10.1109/TSG.2016.2587120. [222] A. Idris, M. Sabry, N. I. Abdul Razak, and A. L. Yusof, “Fast time-varying channels in MIMO-OFDM system using different diversity technique,” in 2013 IEEE Symposium on Wireless Technology & Applications (ISWTA), 2013, pp. 138–141. doi: 10.1109/ISWTA.2013.6688756. [223] J. Montalbán, M. M. Velez, G. Prieto, I. Eizmendi, G. Berjon-Eriz, and J. L. Ordiales, “On approaching to generic channel equalization techniques for OFDM based systems in time variant channels,” in 2011 IEEE International Symposium on Broadband Multimedia Systems and Broadcasting (BMSB), 2011, pp. 1–6. doi: 10.1109/BMSB.2011.5954940. [224] J. González-ramos, I. Fernández, I. Angulo, A. Gallarreta, D. D. La Vega, and A. Arrinda, “Empirical characterization of the conducted disturbances generated by the electric vehicles during the charging process,” in CIRED Porto Workshop 2022: Emobility and power distribution systems, 2022, pp. 319–323. doi: 10.1049/icp.2022.0719. [225] J. González-Ramos, I. Angulo, I. Fernández, A. Gallarreta, A. Arrinda, and D. de la Vega, “Influence of Electric Vehicle Charging on the Grid Access Impedance from 20 kHz to 500 kHz,” in 2023 International Conference on Smart Energy Systems and Technologies (SEST), 2023, pp. 1–6. doi: 10.1109/SEST57387.2023.10257356. [226] J. González-Ramos et al., “Influence of the Spectral Pattern of the Conducted Emissions Generated by Electric Vehicle Charging on PRIME v1.4,” in 2024 IEEE International Conference on Communications, Control, and Computing Technologies for Smart Grids (SmartGridComm), 2024, pp. 167–173. doi: 10.1109/SmartGridComm60555.2024.10738085. 101 Chapter 2 Conclusions and Future Work 102 1. Conclusions This Doctoral Thesis has focused, first, on characterizing the electrical grid as a transmission medium in terms of the grid access impedance, channel response, and NIEs covering the 9-500 kHz frequency band. Second, NB-PLC technologies under different channel conditions have been evaluated by means of laboratory measurements. Characterization of the electrical grid NIEs Characterization of the NIEs generated by EVs during their charging process. A novel methodology for the evaluation of any conducted emission in the LV grid in the frequency and time domains has been proposed. The spectral analysis, which allows evaluating not only the amplitude of the emissions, but also the distribution of the emissions over the whole frequency band, shows that the emissions depend on the EV model, charging current, and SoC. The time analysis, in turn, reveals that, except for the tonal emissions whose central frequency oscillates with time, the NIEs generated by EVCPs follow a sub-cycle periodic pattern (20 ms) in the frequency band of interest. In the specific case of the tonal emissions with a time-oscillating central frequency, a time-dependent behavior within 24.4 ms is reported. Propagation and interaction of the emissions generated by EVCPs. Concerning the propagation of the previously characterized emissions, this Doctoral Thesis has proven that they propagate several meters through the LV grid. Although in most cases the disturbances are attenuated with distance, there might be resonances that lead to higher amplitudes at an electrical point distant from the source of the emissions. Moreover, the influence of the simultaneous charging of several EVs has been studied, showing that, in general, the amplitudes correspond to the superposition of the individual emissions, in addition to intermodulation products due to the switching frequencies of the inverters. Comparison of the emissions generated by EVPCs using a LISN (isolated conditions) and when measuring directly in the LV grid (on-line conditions). The results show that the emissions measured under on-line conditions generally exceed the ones measured using a LISN, even though the highest-amplitude tonal emissions are several dB lower for on-line conditions. The study also reveals that the high-amplitude emissions are concentrated in a few frequency bins at frequencies up to 150 kHz when considering isolated conditions. Under on-line conditions, in turn, high-amplitude emissions exceeding the PLC out-of-band emission limits are reported in the whole 9-500 kHz frequency band for certain EVCPs. Therefore, since the frequency and time characteristics of the emissions depend on the measurement conditions, this characterization cannot be only based on trials conducted using a LISN. Grid impedance Characterization of the mean grid impedance in the presence of EVCPs. This Doctoral Thesis has revealed that the measured impedances, regardless of whether an EV is charging in the network or not, are considerably lower than the reference impedances defined in the CISPR 16-1-2 standard. Besides, the influence on the grid impedance of the distance between the measurement point and the POC where the EVCSs are installed has been 103 evaluated, concluding, first, that the highest impedance variations do not necessarily occur at the electrical point where the EV is connected; and, second, that the influence of a certain EVCP depends not only on the EVCP itself, but also on the impedance in the default situation (when no EV is connected). Characterization of the sub-cycle impedance variations in the presence of EVCPs. This Doctoral Thesis has also reported a time-dependent behavior of the grid impedance within the fundamental period of the mains (20 ms) at specific frequencies both in the absence of EVs and when each EV is charging. Characterization of the long-term impedance variations in the presence of EVCPs. This Doctoral Thesis has also dealt with the long-term variations (some hours) of the grid impedance, concluding that both the amplitude and the spectral characteristics are modified during the charging process of an EV. Specifically, different impedance states (different spectral patterns and amplitudes) have been reported for all the EVCPs under analysis. Additionally, variations within an impedance state have been also registered, which show significant differences in amplitude and phase at the high-amplitude resonances occurring at specific frequencies. Evaluation of NB-PLC under different channel conditions NIEs Influence of the spectral pattern of the emissions generated by EVCPs on PRIME v1.4. The results show that the four modulations used in this study (DQPSK_C, DBPSK_C, R_DQPSK, and R_DBPSK) are generally capable of correcting tonal or narrowband emissions occurring at certain frequencies (in all cases if modulations with repetition codes are used; for some EVCPs, also when modulations including FEC are used), since they only affect a limited number of subcarriers. AWGN, in turn, affects all subcarriers equally and is critical for OFDM communications. Considering that the emissions show high-amplitude tonal and narrowband emissions in channel 1 and a spectral pattern similar to AWGN in channels 3-8, this Doctoral Thesis leads to conclude that the spectral pattern of the emissions in channels 3-8 is more critical for communications than in channel 1 in terms of FER-SNR curves. Thus, the results presented in this document provide evidence that it is necessary for NB-PLC devices to use modulations with FEC (with repetition codes if possible), in order to avoid the negative effects of the conducted emissions generated by EVs during their charging process. Grid impedance Influence of the frequency-dependent impedance variations on PRIME v1.4. The results reveal that the highest signal attenuation is reported in those frequency channels in which the modulus of the impedance is low and maintained within the frequency channel (low standard deviation) and in which the phase presents abrupt variations. Regarding the evaluation of the quality of the communications, notching effects in frequency-dependent impedances have been demonstrated to degrade FER-SNR curves. Influence of the sub-cycle impedance variations on PRIME v1.4. This Doctoral Thesis has also dealt with sub-cycle impedance variations, concluding that, due to the difference between the impedance frequency responses in the ON and OFF states, a higher SNR is required to obtain a FER of 5 %. Finally, if, in addition to the sub-cycle impedance variations, the Sustainable Energy, Grids and Networks 36 (2023) 101217 3 version). Although different BB-PLC technologies have been developed for inhome and MV channels, there is still no BB-PLC system specifically designed to be deployed over the LV distribution grid. Considering the enhanced performance with respect to NB-PLC in terms of bandwidth, latency, and security requirements, BB-PLC is an alternative that is being considered by some distribution system operators (DSOs) for data transmission through the LV distribution grid, in order to fulfill the demanding requirements of new SG applications [41]. As the behavior of the outdoor channel is expected to be considerably different from the characteristics of the indoor environment, DSOs are taking into consideration two different options. First, the adaptation of the existing indoor technologies to the characteristics of the LV grid; and, second, the development of a new standard based on the outdoor transmission medium. In any case, the development of BB-PLC technologies would definitely be a solution for covering the needs of future SG applications, including the integration of distributed energy resources or EV charging management, among others [7]. With the aim of providing a visual overview of the evolution of PLC, Fig. 1 shows the past, present, and future of the technology development. 3. Main parameters to be measured In this paper, the most relevant aspects to be considered for the empirical characterization of the electrical grid as a transmission medium are addressed: NIEs, grid access impedance and attenuation/ channel response. Throughout the text, the term noise corresponds to an electromagnetic phenomenon not conveying information and which is combined with a wanted signal [42]. Since the emissions referred to in this paper are always non-intentional, emissions and NIEs are used as synonyms and are defined as unwanted signals generated by the power electronics included in the circuitry of connected electronic devices. The terms interference and disturbance, in turn, imply a degradation of a system. In the case of an interference, the performance of a communications system is jeopardized, while a disturbance is related to the degradation of Power Quality (PQ), or the lifetime of a device, among others. Finally, the term distortion refers to an undesired change in the waveform of a signal that might lead to the appearance of new frequency components. 3.1. Non-intentional emissions 3.1.1. Background In the last years, there has been a considerable increase in the number of electronic devices connected to the electrical grid via power converters [43]. These devices, as they are based on inverters with switching frequencies above 10 kHz, have a significant impact on the levels of the NIEs generated in the frequency range of kHz [44–46], also known as supraharmonics [47]. These high-amplitude emissions, which are inherent to the operation of power electronics devices [47], can cause PQ issues, such as equipment malfunction, falsification of energy/smart meters, or overvoltage [48–53], in addition to degrading the quality of PLC technologies [54,55]. As the term supraharmonics only refers to the disturbances in the 2–150 kHz frequency band, it is not sufficient to address EMC issues that may affect NB-PLC up to 500 kHz, neither BB-PLC in the MHz bands. The relevance of the effect of the NIEs depends directly on the amplitude, spectral form, and time-variant behavior of the emission [7]. With the aim of analyzing their influence on communications under laboratory conditions, the ETSI TS 103 909 [56] defined a set of reference noises representative of the LV grid. However, due to the large increase of Distributed Energy Resources (DERs) in recent years, the time and spectral patterns of the disturbances present in the grid are expected to have varied considerably and, thus, their effect on communications is still unknown. For this reason, the characterization of the disturbances present in the LV distribution grid, in both the time and frequency domains, requires further study and becomes essential for the further development of PLC technologies. To do so, it is necessary to carry out extensive field measurement campaigns, in which different configurations and grid topologies are considered, so that the emissions of the grid are correctly characterized. 3.1.2. Classification of noise and main sources of NIEs The noise of the electrical grid can be classified as background noise, which varies over long periods of time, and impulsive noise, which shows fast time-varying behavior. The background noise, in turn, can be divided into colored noise, which is mainly caused by residential electronic equipment, such as computers, dimmers, or hair dryers [57], and has a low power spectral density, and narrowband emissions caused by broadcasting emissions. The connection or disconnection of electronic devices can imply impulsive noise that is aperiodic with the mains frequency. In other cases, due to the rectifiers included in the power supplies, periodic impulsive noise synchronous with the fundamental frequency can be found in the grid, with repetition rates multiple to 50/60 Hz. Asynchronous periodic impulsive noise is also present in the grid, showing repetition rates of 50–200 kHz [7,57–61]. In Fig. 2, a summary of the classification of the noise present in the LV grid is shown. Some authors have statistically characterized the impulsive [62–65] and background [66,67] noise in indoor channels in the frequency band assigned to PLC, which can serve as a basis for the Fig. 1. Evolution of PLC technology: past, present, and future. J. Gonz´ alez-Ramos et al. Sustainable Energy, Grids and Networks 36 (2023) 101217 4 characterization of the noise in outdoor environments. In compliance with [68–71], the noise of the electrical grid can be patterned as the sum of the previously mentioned types of noises. In [72], it is stated that the cyclostationary noise, caused by devices showing an impedance changing in the short-term [73], is predominant in the electrical grid and has a significant influence on NB-PLC, as it reduces considerably the data rate of the OFDM-based technologies [74]. Some recently published articles [75–79] show that photovoltaic inverters (PV), battery chargers, energy-efficient lighting, hydropower systems, wind turbines, or EV chargers, among others, are the main sources of the high-amplitude conducted emissions. In several field and laboratory trials, a frequency and time characterization of these emissions has been carried out, concluding that, as they occur at frequencies assigned to NB-PLC, the quality of communications can be substantially affected and degraded [30,54,80,81]. In some instances, as for the EV chargers and PV panels, the sources of disturbance are located near the smart meter, which might imply an additional challenge for the correct performance of NB-PLC [50]. Other non-intentional emitting devices, such as motors [76], lighting devices [82–84], or electronic amplifiers [47], introduce high-amplitude emissions with a wide range of spectral patterns. A significant time-dependent behavior has also been observed for the previously mentioned sources [76]. It should also be noted that high-amplitude impulsive disturbances are generated by this equipment when commuting between different states or working regimes [76], which may have an additional negative effect on communications. In the BB-PLC frequency band, in turn, conducted emissions are not expected to be that high. Fig. 3 shows the main sources of the emissions present in the LV distribution grid, as well as the publications addressing their characterization. 3.1.3. Normative framework 3.1.3.1. Emission limits. There is great interest from the standardization and regulation organizations in the characterization of NIEs in the frequency and time domains for frequencies above 2 kHz. For example, the former SC 205 Working Group 11 of CENELEC, currently TC219, is in charge of studying the disturbances in the electrical grid and determining the immunity levels for communications [47]. It should be mentioned that this working group has recently requested data for inclusion in a report about the propagation of conducted emissions up to 500 kHz. Moreover, through the working groups TC77A and CISPR SC/H, the International Electrotechnical Commission (IEC) defines the requirements for regulating emissions so that the compatibility of electrical products in the frequency range up to 500 kHz is ensured. The maximum amplitude of the emissions generated by certain equipment connected to the electrical grid has already been specified by the International Special Committee on Radio Interference CISPR. CISPR15 (EN 55015) [82] defines the limits for the lighting equipment, whereas CISPR11 (EN 55011) [85] addresses the maximum amplitudes Fig. 2. Classification of the noise of the LV distribution grid. Fig. 3. Main sources of emissions in the LV grid and summary of the publications addressing their characterization. J. Gonz´ alez-Ramos et al. Sustainable Energy, Grids and Networks 36 (2023) 101217 5 for induction cooking devices. By contrast, no specific limits have been specified for many other sources of emissions, such as PVs, EVs, or hydropower systems. For these devices, the out-of-band limits defined for communications equipment in EN 50065–1 [86] might be considered as a conservative criterion [57]. This technical specification defines limits adapted to the frequency band of transmission (2–9 kHz, 9–150 kHz, or 150 kHz-30 MHz), and the type of detector that should be used in the measurements, Root Mean Square (RMS), Quasi Peak (QP) or Average [87–89]. These emission limits are defined for laboratory conditions and must be evaluated by means of a Line Impedance Stabilization Network (LISN). A LISN is a standard load impedance allowing the repeatability and comparability of EMI measurements, which also prevents the trials from being affected by external conducted emissions [90]. In the LV distribution grid, the Compatibility Levels (CLs) establish the maximum amplitudes of the emissions that cannot be exceeded at a specific electrical point, as the combination of the emissions generated by all the equipment connected to the network. The grid operator should ensure that at least in 95% of the locations these limits are not exceeded. The Annex B of the IEC 61000–4–7 [91] specifies the limits for the frequency band from 2 kHz to 9 kHz in RMS values, whereas the IEC 61000–2–2 [92] defines the limits for the 9–150 kHz frequency range in QP values. 3.1.3.2. Measurement setups. Existing standards specify measurement setups for the characterization of the emissions from equipment under test. CISPR 16–1–2 [93] defines a standardized measurement setup for the evaluation of the emissions based on a LISN from 9 kHz to 109 MHz. Similarly, IEC 61000–4–7 [91] specifies the use of an Artificial Mains Network (AMN) below 9 kHz. In this way, a controlled and isolated scenario is set and the measurements are not affected by external NIEs [94]. However, the measurement campaigns carried out so far lead to conclude that the reference impedances defined in [91,93] are not a good representation of the actual grid access impedance values of the LV distribution grid [95]. For this reason, the characterization of the emissions should not only be based on measurements using AMN or LISN, where isolated effects can be evaluated, but also on field trials, so that representative disturbance values are obtained. In this context, some authors [96,97] have opted for conducting investigations related with NIEs at reconstructed facilities. This type of scenarios avoid the ideal conditions of laboratory measurements as well as the uncontrolled grid factors present in on-field trials, such as the variety of loads connected to the grid, the grid topology, the grid impedance, or the electrical cables, among others [75,98]. 3.1.3.3. Measurement methods. In the bibliography, the term measurement method refers to the post-processing applied to the recorded signals to evaluate NIEs in the frequency domain. Up to now, no normative measurement method for the assessment of the disturbances in the LV distribution grid has been defined for frequencies above 9 kHz. As previously mentioned, the compatibility levels in the 9–150 kHz frequency range that are included in IEC 61000–2–2 [92] are defined for QP values [92]. For this reason, CISPR16 1–1 [87], a method based on a QP detector, is the method that might be used for the evaluation of the disturbances in grid measurements [99]. Nevertheless, this method presents several drawbacks: first, it is not intended for grid measurements, but rather for laboratory conditions; and, second, it has high complexity, computational burden, and memory requirements. Accordingly, the standardization institution IEC SC77A/WG9 is currently working on the development of a normative method for the evaluation of NIEs in the LV distribution grid. In this context, several authors have proposed new methods and compared their performance based on simple synthetic signals or test signals recorded in the LV grid [100–104]. In Table 2, the main characteristics of the measurement methods (informative or normative) proposed in the standards or recently considered for its inclusion in a standard are gathered. In order to show the different results provided by each method, Fig. 4 presents the spectrum of a particular emission according to IEC 61000–4–7 (max, RMS), IEC 61000–4–30 (max, RMS), CISPR 16–1–1 (QP), and LightQP (QP) methods in the 2–150 kHz frequency band. Once the spectra are obtained, there is no standardized procedure for the quantitative characterization of the emissions in the frequency domain. The literature only considers the Total Supraharmonic Voltage (TSHV) [110–112], a parameter that gives an insight into the total amplitude of the emissions in the frequency band under analysis. However, this parameter is highly dependent on the measurement method used for the post-processing of the corresponding emission [102] and does not take into account the frequency distribution of the amplitude of the emissions, which plays an essential role in the proper design of PLC technologies. Moreover, the concepts of narrowband and broadband emission have not yet been defined and, therefore, there is no normalized procedure for their proper characterization. Table 2 Comparison of the measurement methods for the assessment of the conducted emissions in the LV grid. Method Frequency band Time window Window length Overlapping Frequency step size Gaps Output metrics CISPR 16–1–1[87] 9–150 kHz Gaussian/Kaiser/Lanczos, etc. 20 ms ≥75% ≤100 Hz No QP 150 kHz-30 MHz 0.5 ms ≤4.5 kHz IEC 61000–4–7[91] 2–9 kHz Rectangular 200 ms 0% 200 Hz No RMS, max IEC 61000–4–30[105] 9–150 kHz Rectangular 0.5 ms - 2 kHz Yes RMS, max Subsampling approach[106] ≤150 kHz Rectangular 5 ms 0% - No RMS, max OMP compressive sensing[107] 2–150 kHz Rectangular 0.5 ms 0% 200 Hz No RMS, max Bayesian compressive sensing [108] 2–150 kHz Rectangular 0.5 ms 0% 200 Hz No RMS, max Wavelet approach[109] 2–150 kHz Rectangular 200 ms 0% 200 Hz No RMS, max Light QP[99] 9–150 kHz Rectangular 20 ms 0% 100 Hz No QP, RMS, max Fig. 4. Spectrum of a particular emission generated by the charging process of an EV according to IEC 61000–4–7 (max, RMS), IEC 61000–4–30 (max, RMS), CISPR-16–1–1 (QP), and LightQP (QP) methods in the 2–150 kHz frequency band. J. Gonz´ alez-Ramos et al. Sustainable Energy, Grids and Networks 36 (2023) 101217 6 Finally, it should be mentioned that the measurement methods presented in this section are only intended to characterize the emissions in the frequency domain, without considering their time-dependent behavior. Since these time variations can negatively affect PLC, the development of methods that address the time characterization in conjunction with the frequency characterization is necessary, for which the definition of standardized methodologies in the time domain is of upmost importance. A joint time-frequency domain characterization of the emissions in the frequency range 9–150 kHz can be found in [113]. 3.1.4. Propagation/interaction of the emissions According to the literature, the disturbances generated by the devices connected to the grid propagate through the LV network, and may even be transferred to the MV grid over distances of several kilometers [97,114–116]. As a result, they may affect the operation of energy meters and PLC equipment [117,118]. As individual devices have a greater impact on higher frequencies than on harmonics [119], the prediction of the disturbances in this frequency range should consider the whole installation and not only individual devices [120]. In accordance to [95,119,121–124], the disturbances generated by a certain device can be classified as primary and secondary emissions. The primary emission corresponds to the emission originated inside the device, whereas the secondary emission is generated by other electronic equipment or the grid itself, and propagated into the device. This propagation highly depends on the impedance of neighboring devices in relation to the impedance of the electrical grid [121,125,126]. Resonances are a key aspect in the propagation of the disturbances, as they imply increases in the emission at the switching frequency [95,121, 127]. In [119], it is stated that a resonance results in an increase in the secondary emission, whereas the primary emission is attenuated. Regarding the interaction, some works have pointed out the existence of frequency beating and intermodulation effects. For instance, in [45], several simulations carried out in MATLAB led to conclude that the charging of EVs of the same type, due to slightly different switching frequencies (f 1 , f 1 ’), implies an emission at |f 1 -f 1 ’|. In general, the beating frequency is in the order of some Hz and does not affect PLC. Intermodulation distortion in the frequency band up to 100 kHz, in turn, occurs due to the interaction of considerably different switching frequencies (f 1 , f 2 ) and is in the order of tens of kHz [45]. This effect is also reported in [97], where the interaction between three Electric Vehicle Charging Processes (EVCPs) in a controlled and isolated LV grid is analyzed in the frequency range up to 500 kHz. Multiple tonal emissions not registered when measuring the disturbances generated by each EVCP individually are shown when connecting several EVCPs simultaneously. The article concludes that these emissions are linear combinations of the fundamental frequencies of the inverters (20 •f2−f1, 20 •f2+f1, 22 •f2−f1…), i.e., intermodulation products. Another similar analysis concerning intermodulation products can be found in [128]. In this article, the interaction between a PV panel and an EV is reported, evidencing the existence of intermodulation products of up to 4th order. The interaction between end-user and PLC equipment has also been studied in the literature. 5 types of interactions between these devices are described in [129], concluding that the performance of communications is degraded and the lifetime of the end-user equipment is considerably reduced. 3.1.5. Electric vehicles (EVs) The high penetration of EVs in the following years will significantly contribute to reach the main objectives of the SGs, allowing, among others, the reduction of costs and environmental impacts [81,130]. However, the rising number of EVs also involves considerable challenges. For example, EV Charging Stations (EVCSs) generate higher amplitude emissions than other electronic devices connected to the grid [131], as they are based on recently developed technologies, aimed at minimizing the charging time and increasing the charging efficiency. For this reason, there is great interest in the scientific community in analyzing and characterizing the disturbances generated by EVs during the charging process. Since the manufacturers of the EVCSs do not publicly share the details of the implemented converters, the theoretical analysis of the disturbances is not generally possible. For this reason, as stated in [132], the characterization of these emissions can only be carried out empirically, by means of extensive measurement campaigns. In recent years, some laboratory and field trials have been developed for the analysis of the disturbances generated by several EVCPs. For instance, in [96], the emissions introduced by a Bi-Directional Vehicle to Grid (V2G) EVCS are studied in a controlled grid scenario up to 150 kHz. This work concludes that narrowband emissions are present at the switching frequency and multiples of it, caused by the Pulse Width Modulation (PWM), whereas broadband emissions occur at higher frequencies due to the DC-DC converter when the Positive Temperature Coefficient heater is activated. Field trials presented in [133] show that the high amplitude emissions generated by an electric bus do not decrease with frequency in the band 2–150 kHz and that they could be as high as the CL defined in [92]. An isolated grid scenario, connected to the public electricity grid, is also used in [45], where the propagation and interaction of the NIEs of four EVs are analyzed up to 100 kHz. As the grid allows switching to microgrid mode, measurements in both configurations have been considered, obtaining different results in some cases. In [97], a procedure for the frequency and time characterization of the emissions in a controlled LV grid is presented. The study, based on the calculation of several parameters in the frequency domain and a FFT analysis, reveals high-amplitude emissions varying within the fundamental period of the mains in the frequency range 9–500 kHz. Reference [134] analyzes the long-term supraharmonic emissions of three EVs in the time and frequency domains up to 100 kHz at three parking garages. The presented results lead to conclude that the power grid disturbance levels increase when the number of connected EVs rise. In [135], the disturbances generated by different types of chargers at five sites in China and Germany are characterized in the frequency domain. The article shows the dominant emission frequencies as well as their voltage amplitude up to 50 kHz. A similar analysis is presented in [136–138], where the switching frequencies of twelve, ten, and eight EVs, respectively, are identified. A huge variability in the switching frequency and its amplitude can be observed in all three cases. However, in on-site measurements emissions are affected by several grid factors [139], including the time and frequency-dependent grid access impedance. Thus, in some works, such as [94], the EVCS is directly connected to a LISN in order to isolate the setup from the LV grid, avoiding external noise and ensuring that only the emissions generated by the EVCPs are measured. In [140], a time and frequency analysis of the disturbances in such a controlled environment is presented. However, it only considers two EV models and does not provide an in-depth study of the time-variant behavior of the disturbances. 3.2. Impedance and channel response 3.2.1. Background The grid access impedance is one of the main factors affecting PLC systems. The time [141] and frequency variations of the grid access impedance [73,142] results, in turn, into considerable changes in the channel frequency response (CFR). As the modulus of the CFR corresponds to the losses of the transmitted signal, PLC can be greatly attenuated [143–146]. This high attenuation can significantly reduce the maximum range of the communications [147] and is, generally, due to sudden changes in loads [6]. However, these impedance variations do not only affect the PLC signal but also the NIEs present in the grid [95, 119]. Hence, the CFR may have a double effect on the quality of the communications. Additionally, as there is no impedance matching between the impedance of the communications equipment and the grid access J. Gonz´ alez-Ramos et al. Sustainable Energy, Grids and Networks 36 (2023) 101217 7 impedance, the maximum power transfer is not achieved and, thus, the transmitted signal might be substantially attenuated [148]. 3.2.2. Main factors affecting impedance and channel response Load variations on the power lines, day-time energy or night-time energy demands, among others, are the main causes for the timedependent behavior of the impedance [149]. Traditionally, only the long-term impedance variations, which are caused by the sudden connection or disconnection of electric equipment connected to the grid, have been considered. According to the literature, some measurement campaigns for the characterization of the grid access impedance, such as [150] or [151], have been carried out all around the world, concluding that no significant long-term variations were observed. Nevertheless, in the last years, some works highlight the existence of short-term variations, periodical and synchronous with the mains signal, caused by certain electronic equipment [5,73,146]. Normally, these sub-cycle variations are due to the switched components (power supplies, AC/DC converters, and inverters) included in the circuitry of certain electronic equipment, which result in two differentiated impedance states [5,73,152,153]. An example of the modulus and phase of a time-varying impedance within the 20 ms period in 50 Hz grids is shown in Fig. 5. An analysis of the influence of these sub-cycle variations on NB-PLC PRIME can be found in [152]. The study concludes that these variations, not only due to the difference between the two impedance states within the cycle, but also due to the time instants in which the variations occur, have a significant negative effect on the channel response estimation and the channel equalization processes defined by NB-PLC technologies. This occurs since the receiver of the communication system uses the symbols of the header to estimate the channel frequency response and, when applying that estimation to the payload, the CFR has rapidly changed due to the sub-cycle impedance variations. Therefore, the equalization is performed by using an incorrect estimation of the channel. It should be mentioned that, as stated in [152], the length of the transmitted frame, as well as the modulation used, have an impact on the system performance under these circumstances. According to the literature, the attenuation measured at low frequencies, 50 Hz or 60 Hz, cannot be extrapolated to those presented in the power grid at frequencies where PLC communications are established [33]. As it is detailed and analyzed in [154], at these higher frequencies, high attenuations can be due to different causes: grid topology, multipath, distance, frequency or type of electrical cables. According to [155] the reflection caused by the impedance mismatching has a significant impact on the signal loss. In addition, the influence of distance and number of branches in the attenuation suffered by the PLC signal is studied. The article concludes that the greater the distance between the communications equipment and the number of branches, the higher the attenuation measured. Similar results are presented in [156], where the additional high losses generated by the branches in the tree-like topologies are highlighted. Reference [57] points out the impact of distance in attenuation, especially when transmitting at high frequencies. It is worth mentioning that [155] also considers the study of two different lines with the same number of branches and the same distance between transmitter and receiver, obtaining different results of attenuation. It concludes that even though, as mentioned before, the transmission losses are directly proportional to distance and number of branches, other factors, such as the access impedance modified by the different equipment connected to the power grid, can also reduce the level of the received signal. This is also reported in [6], where it is indicated that the highest attenuations are due to the switching ON/OFF equipment (sudden changes in loads) and the addition of new lines or costumers. As an example, in Fig. 6, the attenuation in the 1.7–10 MHz frequency range from a Secondary Substation (SS) to a SM room in the LV grid in Spain is shown. 3.2.3. Normative framework As the behavior of the grid access impedance in the frequency band assigned to PLC technologies is still unknown [157], the regulatory framework in this field is still limited. IEC 61000–4–7 defines a reference grid impedance in the frequency range 2–9 kHz [91], but some authors have extended its definition up to 150 kHz [96,158,159] or even up to 500 kHz [98]. However, the measurement campaign carried out in the LV grid in Austria, Switzerland, Czech Republic, and Germany demonstrates that the reference grid impedance in IEC 61000–4–7 Fig. 5. Example of the modulus and phase of an impedance varying within the fundamental period of the mains (20 ms) corresponding to the measurement of the charging of a mobile phone carried out over a LISN. Fig. 6. Example of the attenuation (dB) from a SS to a SM room in the LV grid in Spain. J. Gonz´ alez-Ramos et al. Sustainable Energy, Grids and Networks 36 (2023) 101217 8 overestimates even the highest values of the impedances measured in the actual distribution grid [158]. CISPR 16–1–2 describes three different reference impedances for defining specific LISNs depending on the frequency range (9–150 kHz, 150 kHz-30 MHz, or 150 kHz-108 MHz) [93]. According to [159], the CISPR 16–1–2 reference impedance defined in the frequency range 9–150 kHz also exceeds the values of the impedances measured in indoor environments in Germany, United Kingdom, and Spain. In Fig. 7, the modulus and phase of the reference impedances defined in CISPR 16–1–2 and IEC 61000–4–7 are presented. As the IEC 61000–4–7 only defines the modulus of the impedance, Fig. 7(b) only includes the phase specified in CISPR 16–1–2. The empirical characterization of the LV grid requires measurement systems adapted to the characteristics of the network. The lack of available commercial devices has led some authors to design their own systems with different characteristics and based on different hardware/ software. These systems should comply with the maximum tolerable measurement uncertainty defined in IEC 61557–3 [160] and IEC 61000–4–19:2014 [161]. However, the absence of a standardized frequencyand time-dependent reference impedance has prevented the definition of a normalized method for the accuracy assessment of a specific system, as well as the comparison between different implementation techniques. The first approach for defining a reference impedance is reported in the European Project Z-NET [162]. In this project, only the definition of a static impedance is considered, which is not suitable for the evaluation of sub-cycle variations. 4. Characteristics and limitations of the existing measurement systems 4.1. Non-intentional emissions The correct characterization of the emissions requires properly designed and calibrated measurement systems. In principle, the emissions can be recorded by means of commercial equipment, such as spectrum analyzers or vector network analyzers (VNAs), connected to the grid. Some authors have proposed measurement systems for the assessment of the NIEs in the LV grid based on these devices [72,76, 163]. However, the fundamental component of the grid (220 V, 50 Hz), in addition to the overvoltages generated due to the connection or disconnection of other electronic equipment or the measuring equipment itself, can damage the measurement system. For this reason, many authors have opted for using oscilloscopes, also referred to as data acquisition (DAQ) devices in the literature, connected to the grid through coupling circuits (CCs) and voltage/current probes [44,45,54, 55,59,62–65,70,71,73,79,83,129,132–134,137]. An alternative based on equipment designed for the transmission of PLC signals is presented in [164]. In Fig. 8, a summary of the measurement systems for the assessment of the NIEs in the PLC frequency bands is presented. Thus, the protection of the measuring devices is a key aspect to be taken into consideration, without overlooking that the circuits used to protect the system can considerably reduce the accuracy of the measurement [165,166]. It should be noted that, regardless of whether a current or voltage probe is used, a prior characterization is essential, so that the measurement is not affected. For this reason, it is necessary to verify that the frequency response of the probe, regardless of the impedance of the grid, is flat in the whole frequency band of analysis [167]. Since the emissions show a time-dependent behavior [76], the characterization of the propagation and interaction of the emissions generated by the electronic equipment can only be performed by synchronized trials at different electrical points of the LV distribution grid [168]. A solution implemented for achieving synchronization with a precision of ns by means of a GPS module responsible for the generation of a Pulse per Second (PPS) signal is described in [97]. 4.2. Impedance and channel response In the same way as for the NIEs, one of the main obstacles for assessing the grid access impedance and the CFR is the limited number of systems developed for measurements in the LV distribution grid, most of them for frequencies below 500 kHz. The protection requirements to Fig. 7. Modulus and phase of the reference impedances defined in CISPR 16–1–2 and IEC 61000–4–7. Fig. 8. Summary of the existing measurement systems for the assessment of the NIEs in the PLC frequency band. J. Gonz´ alez-Ramos et al. Sustainable Energy, Grids and Networks 36 (2023) 101217 9 avoid damage to the measurement devices, the requisite of portability and independent power supply are some of the challenges to be addressed in developing a measurement system that can be used for onfield trials. Besides, it is essential to obtain information on the modulus and phase of both parameters, with the aim of completely understanding the behavior of the propagation channel. Concerning the grid access impedance systems, some commercial impedance analyzers and VNAs provide accurate impedance measurements of electronic isolated devices in controlled laboratory conditions [163,169–172]. However, these measurement devices are not designed to be used in the field and are not usually valid for measurements of grounded devices. Moreover, they present mismatch issues, due to the difference between the internal impedance (50 Ω) and the grid values. Additionally, to avoid damage of these devices due to high power emissions [167], they would require high-protection coupling circuits that modify the injection and reception conditions and, consequently, reduce the measurement accuracy. For this reason, some authors have developed ad-hoc impedance measurement systems adapted to the characteristics of the LV grid. According to the literature, these systems can be classified depending on the origin of the harmonic current used for the system excitation [173]. Under this classification, there are non-invasive methods, based on generating a current by means of existing network components and which are only able to measure at those frequencies where considerable emissions are recorded [174,175], and invasive methods, which use an external system for generating and injecting this current into the network. In order to provide information over the whole frequency band, a test signal should be injected into the grid. The invasive methods, in turn, can be classified as single or multiple frequency sweep methods. The latter allows a considerable reduction of the measuring time, obtaining similar accuracy to the single frequency sweep method. Some examples of invasive methods can be found in [55,157,176–180]. Regarding the measurement systems for the assessment of the subcycle impedance variations in the LV grid, different methods have been published so far. For example, in [173], a setup based on the injection of a single sweep by means of a linear amplifier is presented for the frequency band from 130 Hz to 150 kHz. Reference [181], in contrast, calculates the impedance with respect to time and frequency by transmitting multiple sweeps in the band 30–500 kHz. To this end, a developed card for the synchronization with the mains frequency is required. Another measurement system is reported in [145], which also allows to measure current and voltage after injecting a signal into the power line. This system includes signal processing for error correction and a set of calibrations setups. It should be noted that similar systems have been presented by the same authors in [166,182,183]. References [166,182] only cover the frequency band assigned to NB-PLC (30–500 kHz), whereas [183] provides impedance values up to 1 MHz. Finally, in [184], a measurement system for the characterization of the mean and sub-cycle grid access impedance in the frequency range up to 10 MHz is presented. This system is based on the injection of a test signal composed of short single-frequency bursts in the frequency band of interest obtaining a maximum deviation within ±8%. In Table 3, a comparison of the published measurement systems for the assessment of the grid access impedance in the PLC frequency bands is included. In the literature, CFR measurement systems based on different hardware and considering different methods can be found. In some cases, VNAs are used, including coupling circuits for the protection of the system from the mains signal [163,187–191]. It is important to mention that the use of VNAs implies a great limitation in distance, since it must be connected at both sides between which the CFR is to be measured. In general, these systems include calibration processes so that the effect of the couplers does not affect the measurements. In other cases, such as [169,192–195], a signal generator is used at the Table 3 Comparison of the published measurement systems for the assessment of the grid access impedance in the PLC frequency bands. Reference Mean/Subcycle Frequency range Connection to the grid Measurement method Stiegler, 2015[173] Mean & subcycle 130 Hz-150 kHz Coupling device +power amplifier Invasive method (single sweep) Chruszczyk, 2015[178] Mean 10 kHz-1 MHz Transformer +RC filter +power amplifier Invasive method (sinusoidal signal sweep) Hallak, 2017[145] Mean & subcycle 30–500 kHz Transformer +capacitor +shunt resistor +amplifier Invasive method (single-frequency bursts) Hallak, 2018[166] Mean & subcycle 30–500 kHz Transformer +capacitor +shunt resistor +amplifier Invasive method (single-frequency bursts) Elfeki, 2018[181] Mean & subcycle 30–500 kHz Capacitive coupler +LCL filter Invasive method (single-frequency bursts) Network analyzer Fern´ andez (IGOR-Meter), 2019[177] Mean 1–500 kHz Direct connection (external probes) Invasive method (single-frequency steps) Fern´ andez (UPV/EHU), 2019 [177] Mean 20–500 kHz Via capacitive coupler (external probes) Invasive method (single tonecontinuous sweep) Fern´ andez (TUD), 2019[177] Mean 0–200 kHz Direct connection (internal probes) Invasive method (single-frequency steps) Nieβ, 2020[185] Mean & subcycle 30–500 kHz Coupling circuit +resistive shunts +amplifier Invasive method (sweep sine wave) Saathoff, 2020[186] Mean DC-200 kHz TRIAC-based phase-controlled switch Invasive method (Inrush transient) Nieβ, 2021[182] Sub-cycle 40–500 kHz Coupling circuit Invasive method (cosine current signal) Szymczyk, 2021[183] Mean & subcycle 10 kHz-1 MHz High-pass filter (HPF) +shunt resistor Invasive method (single-frequency bursts) Jensen, 2021[55] Mean 50 Hz-150 kHz Power amplifier +transformer +current probe +voltage/current probe Invasive method (multi-tone signal) Artale, 2022[180] Mean 35.9375–90.625 kHz 98.4375–121.875 kHz 154.6875–487.5 kHz Shunt +50 Hz filter +PLC signal conditioning circuit Invasive method (G3-PLC signal) Erhan, 2022[157] Mean 12–150 kHz Amplifier +injection transformer +current probe +voltage probe Invasive method (single-frequency sweep) Aderibole, 2023[179] Mean 100–200 kHz Shunt +two capacitors +3 inductances +transformer Invasive method (sweep sinusoid) Fern´ andez, 2023[184] Mean & subcycle 20 kHz-10 MHz Capacitive coupler (external probes) Invasive method (single-frequency bursts) J. Gonz´ alez-Ramos et al. Sustainable Energy, Grids and Networks 36 (2023) 101217 10 transmission side for the injection of a signal into the grid, while a data digitizer board is responsible for capturing the signal at the receiver side. Each equipment is connected to the grid through a coupling circuit, in charge of protecting the measuring devices. In Table 4, a comparison of the existing systems for the assessment of the CFR in the PLC frequency bands is gathered. The impedance and CFR measurement systems show similar difficulties to be faced. First, the probes used for measuring the currents should be correctly characterized, as well as the measuring circuits that are part of the system. Moreover, especially at higher frequencies, the parasitic effects of the cables can have a significant impact on the measurement [198] and, thus, a prior and accurate characterization of the connection cables is also needed. Finally, the measurement system should be capable not only of measuring the mean grid access impedance or CFR, but also the sub-cycle variations given within the 20 ms. 5. Measurement campaigns in the LV distribution grid In this section, a summary of the measurement campaigns performed in the LV distribution grid for the characterization of the NIEs, attenuation/CFR, and impedance are presented. Fig. 9 shows a map with the countries in which the characterization of the LV electrical grid has been addressed. 5.1. Non-intentional emissions Several authors have performed on-field trials in different LV distribution grids all around the world considering certain sources of disturbances and grid topologies. As an example, in [163], the NIEs at industrial, residential, and rural areas in the Turkish LV grid are characterized up to 100 kHz, obtaining emission amplitudes around 90 dBµV, 100 dBµV, and 75 dBµV, respectively. A statistical characterization of the emissions in the LV grid based on probabilistic functions is presented in [51], covering the 2–150 kHz frequency band. Another statistical analysis can be found in [164], where a long-term characterization of the emissions in the LV grid in Qatar is presented considering more than 1.8 billion samples at three different locations over 10 days. The analysis is based on the stationarity, autocorrelation, and independence of the NIEs in the frequency range from 10 kHz to 490 kHz. A measurement campaign carried out in Spain [44], in turn, concludes that the propagation channel in rural scenarios is remarkably noisy in low frequencies, as the highest NIEs occur at frequencies lower than 150 kHz, mainly lower than 40 kHz. However, high-amplitude emissions can be found in the whole frequency range up to 500 kHz in urban environments. In [72], a similar characterization of the NIEs in the LV grid in France in the NB-PLC frequency band is gathered. This article points out that the noise is predominantly cyclostationary and concludes that the background noise is stable over periods of hours. 5.2. Channel response and attenuation The huge variety of causes that involve attenuations in PLC signals and the dependency of the behavior of the electrical grid on the geographical area [190,192,199], mainly because of the heterogeneity of grid topologies and connected loads, makes the prediction of the PLC channel response a complex study that should be addressed. In [163], for example, the attenuation between the SS and the SMs in the LV grid in Turkey up to 1 MHz is analyzed in residential, industrial, and rural areas, revealing attenuations around 20 dB (distances from 50 m to 150 m), 30 dB (distances from 80 m to 270 m), and 40 dB (distances from 150 m to 500 m) at each scenario, respectively. Reference [200] considers a set of representative grid topologies, concluding, as in [201], that an increase in distance implies higher transmission losses. However, it is also stated that there is not a linear trend between the attenuation and the distance between communications equipment, due to the great impact of the number of branches in tree like topologies. In [202], apart Table 4 Comparison of the existing measurement systems for the assessment of the CFR in the PLC frequency bands. Reference Indoor/ Outdoor Frequency range Connection to the grid/Protection Measurement basis Cort´ es, 2005[194] Indoor 1–20 MHz Coupling circuit Signal generation board (Tx)-Data acquisition board (Rx) Tlich, 2008[189], [190] Indoor 30 kHz-100 MHz Coupler box Vector Network Analyzer Colen, 2013[169] Indoor/ Outdoor 1.7–50 MHz PLC coupler +Amplifier (Tx) Signal generation board (Tx)-Data acquisition board (Rx) Gassara, 2014 [188] Indoor 10–500 kHz Coupling circuits Vector Network Analyzer Tonello, 2014 [187] Indoor 1.8–30 MHz Couplers +extension cables (characteristic impedance of 50 Ω) Vector Network Analyzer Cort´ es, 2015[191] Outdoor 3–95 kHz Coupling circuit +High power amplifier (Tx)- Coupling circuit +Band-pass filter (BPF) +Analog to digital converter (Rx) Signal generation board (Tx)-Data acquisition board (Rx) Oliveira, 2017 [192] Indoor 1.7–30 MHz (band A) 1.7–50 MHz (band B) 1.7–100 MHz (band C) Coupler Signal generator (Tx)-Data digitizer (Rx) Picorone, 2020 [193] Outdoor 1.7–100 MHz Coupler Signal generator +Digital analog converter (Tx)-Data digitizer +analog digital converter Masood, 2020/ 2021[196], [197] Outdoor 50–150 kHz - PLC transceivers Aderibole, 2023 [179] Indoor/ Outdoor 100–200 kHz Coupling circuit (2 transformers, 3 capacitors, 3 solid-state relays, and 2 shunt resistors) +analog front-end (Low-pass filter +Power amplifier in Tx, BPF +Programmable Gain Amplifiers in Rx) PLC modems Gonz´ alez-Ramos, 2023[195] Indoor/ Outdoor 1.7–15MHz Coupling circuit Signal generator (Tx)-Oscilloscope (Rx) J. Gonz´ alez-Ramos et al. Sustainable Energy, Grids and Networks 36 (2023) 101217 11 from measuring the attenuation in the frequency band associated to NB-PLC, a characterization of the transmission losses measured from 500 kHz to 10 MHz in the LV grid in China is presented. A comparison between the mean transmission losses measured in urban and rural residential areas is carried out, concluding that the attenuation with respect to frequency in the urban scenarios is flatter than what is measured in rural areas. As previously mentioned, the characterization of the PLC channel is carried out according to two approaches: the top-down approach, where a model is obtained from a large number of field trials, and the bottomup approach, based on modeling the channel by applying TL theory [10]. Regarding the top-down approach, the indoor PLC channel has been extensively analyzed in terms of average channel attenuation or channel gain, delay spread, coherence bandwidth, and channel capacity in [187–190,192,203–210]. In [211], the characterization of the indoor PLC channel based on multipath phenomenon can be found. Concerning the characterization of the outdoor scenario, for instance, in [191], the grid access impedance, the channel response (in terms of delay spread, channel bandwidth, and attenuation), and the NIEs in different urban, semiurban, and rural scenarios in the LV distribution grid are analyzed, as well as the achievable data rates. The results presented in this paper only cover the CENELEC A NB-PLC frequency band (3–95 kHz). A similar analysis in the frequency range assigned to BB-PLC (1.7–100 MHz) is gathered in [193], where the Brazilian outdoor scenario is characterized considering the average channel attenuation, root mean squared delay spread, coherence bandwidth, coherence time, and the achievable data rate. In reference [212], a characterization of the outdoor channel considering the multipath signal propagation theory is presented. This model is verified by a set of channel responses obtained by means of field trials. Since in the bottom-up approach the network elements are mathematically modeled, a thorough knowledge of the electrical grid (topology, cable characteristics, load impedances, etc.) is needed. This approach considers the two-conductor TL (2TL) theory [213–219], used for power networks connected with two-conductor transmission lines, and the multi-conductor TL (MTL) theory [220–227], which is a generalization of the 2TL approach. The TL theory has also been applied by some authors for the characterization of the outdoor PLC channel [217,222,223,228,229]. In reference [230], both the LV aerial and underground cable distribution lines are model by means of the MTL theory. A combination of the top-down and bottom-up approaches can be found in [196,197], where a statistical analysis of the NB-PLC channel in the LV grid in Pakistan is presented up to 150 kHz and 500 kHz, respectively. Therefore, taking into account that the CFR is one of the most critical factors that could cause communications to fail and that there is a lack of knowledge about its behavior on the electrical grid, a more exhaustive characterization of the channel in the frequencies assigned to PLC technologies is still needed. 5.3. Impedance Only a few measurement campaigns to characterize the grid impedance have been performed all around the world. In [163], the grid access impedance in the LV grid in Turkey is characterized up to 100 kHz, considering residential, rural, and industrial scenarios. The paper concludes that, regardless of the area, impedance values below 10 Ω are measured in the whole frequency band. Another measurement campaign covering the same frequency range in Austria, Switzerland, Czech Republic, and Germany is presented in [158], showing that the impedance increases with frequency and that remains below the IEC 61000–4–7 reference impedance in the whole frequency band. In [157], phase to neutral impedance measurements are carried out at an electrical vehicle charging plaza at the university and four residential household installations in The Netherlands in the 9–150 kHz frequency range. This article concludes, first, that, due to the capacitor included in their circuitry, both EVs and PVs imply a low-impedance path for communications signals, and, second, that household appliances have a considerable effect on the grid impedance. Moreover, the importance of the topology on the impedance is also highlighted. Although most of the published studies focus, exclusively, on the 9–150 kHz frequency band, some publications provide information up to 500 kHz. For example, in [8], the access impedance from 35 kHz to 500 kHz at three Transformer Stations (TCs) and at a set of access points of the LV distribution grid in the Basque Country (Spain) has been characterized. For this purpose, two urban and a rural scenario have been considered. The study provides an insight into the great differences between the impedances measured at each electrical point and concludes that the urban distribution can be modeled as a particular scenario of short cable sections and numerous homes. In [150], a similar characterization in the frequency band 30–500 kHz is performed, measuring the grid access impedance in an urban/suburban area in China with low-rise apartment Fig. 9. World map with the countries in which measurement campaigns have been carried out for the characterization of the NIEs, attenuation/CFR, and impedance in the outdoor LV grid. J. Gonz´ alez-Ramos et al. Sustainable Energy, Grids and Networks 36 (2023) 101217 12 buildings. The behavior of the grid impedance in the frequency range above 1 MHz is still unknown. In [231], a measurement campaign for its statistical analysis in the Brazilian LV network is presented, concluding that the impedance is a non-stationary random process. However, these results cannot be extrapolated to any distribution network, since the performance of the electrical grid at the frequencies associated to PLC changes drastically from one country to another, mainly due to the diversity of loads and grid topologies [190,192,199]. 6. Open issues The state of the art presented above shows some important limitations. First, there are no commercial measurement systems adapted to the challenging conditions of the LV grid. Impedance and network analyzers, commonly used under laboratory conditions, are not intended for network measurements, mainly because they can be damaged by the fundamental frequency of the grid (50 Hz, 220 V) or by overvoltages due to the connection/disconnection of nearby equipment. For this reason, coupling devices are used for protection purposes, although they reduce considerably the accuracy of the measurement. In this context, the development of measurement systems adapted to the electrical grid, covering both the NB-PLC and BB-PLC frequency bands, in addition to being capable of measuring sub-cycle variations and providing information on the modulus and phase, is of particular importance. For this purpose, all the components that are part of the system (current/voltage probes, protection circuits, etc.), as well as the connection cables, which, due to parasitic effects, could have a considerable impact on the measurement accuracy at higher frequencies, should be correctly characterized. In consequence, the characterization of the behavior of connection cables, especially in the BB-PLC frequency band, is also a matter to be analyzed [198], which should be based on theoretical and physical models considering different cable lengths, cable diameters, and cable types. The measurement systems for the characterization of the LV grid should be evaluated following a standardized method based on reference measurements. However, up to now, no reference impedance has been defined for the proper calibration of a specific measurement system, as well as the comparison between systems based on different techniques or hardware/software. The first attempt of this process can be found in [162], concluding that the measuring conditions (connection of the measuring equipment to the earth and neutral conductors or the supply of the reference impedance) could considerably affect the obtained results. Besides, only static impedances, i.e., impedances not showing sub-cycle variations, have been taken into consideration. Regarding the conducted emissions, there is no normative measurement method for the post-processing applied to the recorded signals to evaluate NIEs in the frequency domain. CISPR 16–1–1 might be used for obtaining the QP values of the recorded signals. However, this method shows considerable drawbacks, such as high complexity, computational burden or memory requirements [99]. Thus, the design of new and higher-performance methods is essential for the correct characterization of the NIEs in real grid conditions. It is important to mention that this evaluation should be based on reference signals with different time and frequency features, which have not been defined yet. For this reason, several authors have opted for assessing the performance of measurement methods based on grid measurements, which are highly dependent on the characteristics and accuracy of the acquisition measurement systems, and ad-hoc designed synthetic signals, which are not commonly of public use. Moreover, these methods only consider the spectral analysis of the emissions, without addressing their variation over time and, therefore, the definition of standardized methods for the time characterization of the NIEs in the LV grid is also a research line to be explored. Another aspect that should be taken into account is the definition of emissions limits for the wide range of electronic devices connected to the grid. Although the CISPR has already specified the maximum amplitude of conducted emissions that can be generated by lighting (CISPR15 [82]) and cooking appliances (CISPR11 [85]), there are scarce specific limits for other devices that will have a high penetration in the following years, such as EVs, PVs, or hydropower bombs. These limits should be defined for both the NB-PLC and BB-PLC frequency bands in RMS, average, and QP values. Similarly, the CL, the limits that should not be overpassed by conducted emissions in the LV grid, have only been established for the 2–150 kHz frequency band and, therefore, there is a need from standardization bodies to extend them to the upper NB-PLC (150–500 kHz) and BB-PLC frequency ranges. The definition of laboratory scenarios that are more representative of the actual characteristics of the LV distribution grid is another aspect currently demanded by the scientific community. These scenarios would allow the characterization of the emissions generated by specific equipment, in addition to analyzing their capacity of propagation and interaction with other electronic devices, avoiding uncontrollable factors such as grid impedance, unknown emission sources, etc. Finally, it should be mentioned that, due to the high penetration of EVs and DERs in recent years, the characteristics of the LV grid as a transmission medium are expected to have changed considerably, and many more changes will follow in the near future. Since the majority of the studies addressing the characterization of the outdoor propagation channel were published in the early 2000 s, further research in this area is needed, in order to define and implement high-performance PLC technologies allowing the development of future SG applications. Although the results obtained for indoor channels cannot be directly extrapolated to the distribution grid scenario, since at each environment different activities are occurring [1], the experience and methodologies developed in this scenario can serve as a first step towards the definition of new methods for the characterization of the outdoor transmission channel. 7. Conclusions The electrical grid is a hostile medium for data transmission, as the power line cable is not designed for communications and its characteristics change over time and frequency. Moreover, the overall behavior of the LV distribution grid is drastically changing due to the introduction of the Smart Grid paradigm. These changes also impose stronger communication requirements for future services, which highlights the need for a detailed characterization of the performance of the propagation medium in both NB-PLC and BB-PLC frequency bands. This characterization should become the basis for the design of robust algorithms for data coding, modulation, and multiplexing. In this context, this article gives an insight into the aspects to be considered for the characterization of the LV distribution grid as a transmission medium for PLC signals based on field trials. The main characteristics of NB-PLC and BB-PLC technologies have been analyzed, as well as the historical evolution of BB-PLC, from their origin in 1990 s as an alternative to DSL for Internet Access to their current purpose of facilitating the development of new SG services. Then, the main parameters to be measured, namely, NIEs, attenuation/channel response, and impedance, have been outlined, describing their fundamentals, along with the still insufficient normative framework. As a consequence of the number of works related to the propagation and interaction of the emissions through the grid and the special interest of the scientific community in the emissions generated by EVs, these topics have been covered more deeply in separate sections. Then, the need to develop properly designed and calibrated measurement systems adapted to the grid has been highlighted, giving an overview of their characteristics and limitations to be faced. Moreover, an overview of the measurement systems found in the literature, based on different techniques, has also been presented. Finally, a comprehensive review of the measurement campaigns carried out in different LV grids all around the world for the characterization of the outdoor electrical network as a propagation J. Gonz´ alez-Ramos et al. Electric Power Systems Research 231 (2024) 110289 Available online 16 March 2024 0378-7796/© 2024 The Author(s). Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/bync-nd/4.0/). Emissions generated by electric vehicles in the 9-500 kHz band: Characterization, propagation, and interaction Jon Gonz´ alez-Ramos * , Alexander Gallarreta , Igor Fern´ andez , Itziar Angulo , David de la Vega , Amaia Arrinda University of the Basque Country (UPV/EHU), Bilbao, Spain ARTICLE INFO Keywords: Conducted non-intentional emissions Electric vehicles Interaction Low voltage grid Propagation ABSTRACT Electric Vehicle Charging Processes (EVCPs), due to the involved power electronics with high-switching frequencies, generate high-amplitude emissions that could have a negative impact on Power Quality (PQ) and Narrowband Power Line Communications (NB-PLC) systems. In this context, this paper deals, first, with the frequency and time characterization of the disturbances generated by four EVCPs in the 9-500 kHz frequency range. The study is based on measurements carried out in a controlled Low Voltage (LV) grid. For this purpose, a novel procedure for the frequency and time characterization is proposed. The frequency characterization is based on the calculation of a set of parameters, which allows evaluating not only the amplitude of the disturbances in the frequency band under analysis, but also their spectral distribution. The time variability is characterized by means of a Fast Fourier Transform (FFT) analysis that leads to a simplified model to evaluate the time-dependent behavior of the disturbances, which shows a sub-cycle periodic pattern of the emissions in the frequency band of interest. Second, the propagation of the previously characterized emissions is analyzed, concluding that they propagate several meters through the LV grid. Although in most cases the disturbances are attenuated with distance, there might be resonances that lead to higher amplitudes at an electrical point distant from the source of the emissions. Finally, the influence of the simultaneous charging of several EVs is studied. The results show that, in general, the amplitudes correspond to the superposition of the individual disturbances, in addition to intermodulation products due to the switching frequencies of the inverters. As a conclusion, the high amplitude time-varying emissions, together with their capacity for propagation and interaction, are a matter to be analyzed due to their influence on PQ and their potential degradation of the performance of PLC in the frequency band from 9 kHz to 500 kHz. 1. Introduction Electric Vehicles (EVs) play a fundamental role in the concept of Smart Grid, and any Smart City initiative should include EV charging in their plans [1]. However, EVs face significant battery-related challenges, which lead to continuous innovation in order to improve aspects such as driving range, charging time, battery cost, etc. [2]. One unintended side effect of EV Charging Stations (EVCSs) is electromagnetic compatibility (EMC) issues due to conducted NonIntentional Emissions (NIEs) [3], which may potentially impact Power Line Communications (PLC) [4–6]. Interfering emissions due to rectification caused by power electronics with high-switching frequencies up to several hundred kHz could propagate and have an effect not only close to the EVCS but also on the surrounding Low Voltage (LV) grid [7, 8]. In order to cover all the frequency range where Narrowband PLC (NB-PLC) technologies can be developed, it is important to extend the Abbreviations: CDF, Cumulative Distribution Function; EDF, ´ Electricite de France; EMC, Electromagnetic Compatibility; EV, Electric Vehicle; EVCS, Electric Vehicle Charging Station; EVCP, Electric Vehicle Charging Process; FFT, Fast Fourier Transform; IFFT, Inverse Fast Fourier Transform; LISN, Line Impedance Stabilization Network; LV, Low Voltage; MV, Medium Voltage; NB-PLC, Narrowband Power Line Communications; NIEs, Non-Intentional Emissions; PFBL, Percentage of frequency bins exceeding the PLC out-of-band emission limits; PLC, Power Line Communications; POC, Point of Connection; PPS, Pulse per Second; PQ, Power Quality; QP, Quasi-Peak; SoC, State of charge; STFT, Short-term Fourier Transform; TSHV, Total Supraharmonic Voltage. * Corresponding author. E-mail addresses: [email protected] (J. Gonz´ alez-Ramos), [email protected] (A. Gallarreta), [email protected] (I. Fern´ andez), itziar. [email protected] (I. Angulo), [email protected] (D. de la Vega), [email protected] (A. Arrinda). Contents lists available at ScienceDirect Electric Power Systems Research journal homepage: www.elsevier.com/locate/epsr https://doi.org/10.1016/j.epsr.2024.110289 Received 8 June 2023; Received in revised form 11 October 2023; Accepted 27 February 2024 Electric Power Systems Research 231 (2024) 110289 2 study of these emissions from the classical supraharmonic range (2–150 kHz) [9] to the frequency band up to 500 kHz. In this context, this paper is focused on the characterization of the NIEs generated by Electric Vehicle Charging Processes (EVCPs) in a controlled LV grid in the frequency range 9-500 kHz, in addition to analyzing their capacity of propagation and interaction. The study is based on measurements carried out in a controlled LV grid scenario. 2. State of the art 2.1. Non-Intentional emissions generated by EVCPs The effect of the penetration of EVs on the Power Quality (PQ) parameters (up to 2 kHz) has been deeply studied in the existing literature [10–12]. Although according to [13], the rectifiers included in the circuitry for the battery charging process are a source of high-amplitude emissions, only a few research works have been carried out for their proper characterization in the supraharmonic range (2 kHz to 150 kHz). In [14], the main causes of the spectral components of the emissions generated by an electric bus in the 2-150 kHz frequency range are identified. In [7], the supraharmonic currents of a bidirectional EVCS are measured and analyzed in the frequency domain, showing both wide-band and narrow-band emissions. Similarly, reference [15] analyzes the supraharmonic emissions generated by nine popular EV models in the Netherlands for frequencies up to 100 kHz. As stated in [16] and [17], the spectral pattern of the disturbances is highly dependent on the EV model and the charging regime, and thus, extensive measurement campaigns are necessary to model this type of NIEs. The authors of the current work also presented a timeand frequency-characterization of the conducted disturbances generated by two commercial EV models from 9 kHz up to 500 kHz in [18]. The measurements analyzed in [18] were performed on an artificial Line Impedance Stabilization Network (LISN). However, conducted emissions in real scenarios will strongly depend on local grid conditions and the influence and interaction with other electrical equipment connected close-by [7]. Therefore, the current state-of-the-art presents some research gaps that still need to be covered in this area. First, there are scarce studies available for the frequency range above 150 kHz, so that their potential negative effect on NB-PLC technology covering the frequency band up to 500 kHz is unknown. Second, as already shown in the literature, the conducted emissions depend on the specific charging process (EVCS, EV model and state of charge). This highlights the need for extensive measurement campaigns considering different charging situations, in order to provide a big picture of the current and future situations of EV development. 2.2. Propagation and interaction of NIEs The conducted disturbances in this frequency range have been proven to propagate over several kilometers in the LV and Medium Voltage (MV) grids [19–21]. Moreover, as in a real situation an EV is surrounded by other sources of disturbances, the resultant emissions at a certain electrical point are expected to be the combination of the propagated emissions generated by each source. For this reason, in order to fully determine the effects of the EVCPs on the amplitude of the emissions in the LV grid, it is essential not only to properly characterize these emissions in the time and frequency domains at the Point of Connection (POC) where the EVCS is installed, but also to analyze their propagation and interaction with other connected equipment. Up to now, several laboratory and on-site measurements have been carried out for the characterization of the interaction of the disturbances generated by different electronic devices. According to the existing studies, the disturbances measured at the terminal of a certain device consist of a primary and a secondary emission. The primary emission, dependent on the grid impedance and, therefore, on the location and time of the day, is defined as the part of the current generated by the sources inside the device. In contrast, the secondary emission corresponds to the part of the current originated by sources outside the device [19,22–27]. In [24], where a model for estimating the emissions of a laboratory installation composed of multiple devices is presented, it is stated that this secondary emission cannot be disregarded and should be taken into account for the correct characterization of the resultant emission generated by an individual appliance. In [23], by means of a simulation model, the influence of the length of the line on the disturbance levels is studied. For that purpose, the interaction between some specific devices connected at the same and different electrical points has been considered. The study in [23] is based on the analysis of the emissions at specific frequencies depending on the number of connected devices and the length of the electrical cable. Some works have also analyzed the influence of the interaction between electronic equipment in a laboratory environment. This interaction depends on the devices and the characteristics of the source impedance [19], and the time of the day [28]. According to the authors of those contributions, the emissions generated from a specific device increase if more appliances are connected at the same electrical installation, whereas the total emission of the grid is reduced [24]. Regarding the propagation and interaction of the emissions generated by EVCPs, in [29], the long-term variations of the NIEs generated from three electric charging infrastructures are analyzed up to 100 kHz, both in the frequency and time domains. The conclusion of the work is that the higher the number of EVs, the higher the disturbances measured in the power grid. A similar study is performed in [30], where the time and frequency analysis of the behavior of the supraharmonics from EVCSs is addressed for frequencies up to 100 kHz in laboratory conditions. However, the existing literature only considers the propagation of emissions from individual devices, without analyzing the cumulative effect of NIEs from several sources on their propagation through the grid. Moreover, as stated in [31], the disturbances generated by electronic appliances show a considerable time-dependent behavior and, consequently, the propagation analysis can only be carried out by means of synchronized measurements at different electrical points [32]. Consequently, the state-of-the-art shows that the aggregation of simultaneous emissions from various EV charging processes should be investigated. Finally, conducted NIEs from EVs should be analyzed in more realistic grid conditions, avoiding uncontrolled grid factors, in order to evaluate propagation and interaction with other electrical equipment. 3. Paper contributions and structure Considering the research gaps presented above, the main objective of this paper relies on the frequency and time characterization of the emissions generated by a set of EVs during their charging process in the 9-500 kHz frequency range. The study, based on measurements carried out in a controlled LV grid, proposes a novel procedure that allows evaluating a certain disturbance by means of the calculation of a set of parameters and a Fast Fourier Transform (FFT) analysis. Moreover, the propagation of the NIEs along the grid is analyzed by means of synchronized measurements at different connection points. Finally, in order to study the interaction of the emissions from various EVCPs, the aggregation of simultaneous emissions is investigated. With the aim of highlighting the novelty of this paper, Table I shows a summary of the previously presented state of the art, in addition to the main contributions of the current paper. The rest of the paper is organized as follows. In section IV, the measurement scenario is presented, including the available technical specifications of the EVCSs under study, as well as the measurement system used for the recording of the NIEs and the signal processing. In section V, the NIEs generated by four different EVCPs are characterized J. Gonz´ alez-Ramos et al. Electric Power Systems Research 231 (2024) 110289 3 in the frequency and time domains up to 500 kHz. In section V.A, the Quasi-Peak (QP) values of the amplitude of the disturbances, according to the CISPR16-1-1 [33], are presented. The study of the time-variant behavior of the disturbances over the measurement time is discussed in section III.B. In sections VI and VII, the propagation and interaction of the emissions previously characterized are addressed. Finally, in section VIII, the main conclusions of the work are gathered. 4. Methodology 4.1. Description of the measurement scenario The measurements that support this study are carried out in the “Concept Grid” laboratory of ´ Electricite de France (EDF), a unique testing facility that goes beyond ideal conditions of laboratory trials, but at the same time, avoids uncontrolled background distortion that may substantially affect the results [7]. This testing scenario simulates a LV distribution grid composed of a Secondary Substation (SS) and five houses, with a three-phase installation, to which different electronic devices can be connected. In this study, three different EVCSs are analyzed. EVCS1 is installed at H2, EVCS2 at H3, and EVCS3 at H5. There are 3 EV models available, and each EV model can only be charged at its corresponding EVCS. A representation of the measurement scenario is shown in Fig. 1, where the distance between the different houses and the location of the EVCSs are indicated. In order to have a predominantly resistive load at the POC, domestic heaters were connected to each house. The background noise due to these loads was characterized at H4, considering this as the default situation. All the measurement results presented in the following sections were conducted in the same electrical phase (monophasic measurements). The four situations to be analyzed correspond to the charging process of EV1 at 81% state of charge (SoC), EV2 at 100% and 75% SoC, and EV3 at 68% SoC. The disturbances were measured at the POC of each house to which the EVCS under study is installed. As the emissions generated by EVCP1 and EVCP3 were only measured at a given SoC, hereafter the SoC of these two EVCPs will not be specified each time these two EVCPs are cited in the text. In order to analyze the propagation of the emissions generated by each EVCP individually, synchronized measurements were carried out at H2, H3, and H5. The analysis of the interaction of the NIEs is based on measurements performed at H2, H3, and H5 when the three EVs are charging simultaneously. Limited technical information of the EVCSs under study is available, with no information about EVCS2. In Table II, the technical specifications of EVCS1 and EVCS3 are gathered. In all the trials, EVCS1 corresponds to the charging post of EV1, EVCS2 to the charging circuitry of EV2, which is installed inside EV2, and EVCS3 to the charging post of EV3. 4.2. Measurement system for NIEs The measurement system for NIEs, shown in Fig. 2, is based on the voltage probe presented in [36], which is connected to the electrical point where the disturbances are measured. This voltage probe shows a flat response for a wide range of impedance values that may be found in the grid. A digital oscilloscope, controlled by a laptop, is responsible for recording the signal with high accuracy (15 bits of resolution in magnitude) and a high sampling frequency (8.92 MHz). The acquisition of the NIEs is performed by means of purpose-specific software developed by the authors. In order to carry out synchronized measurements at three different Table I Summary of the previously presented state of the art and the main contributions of the current paper. Reference Frequency range Measurement scenario Scope of the study Grasel, [7] 9-150 kHz Reconstructed LV grid Characterization of the conducted emissions generated by a V2G EVCS Meyer, [13] 50 Hz-150 kHz Laboratory Characterization of the conducted emissions generated by 19 EVCPs Lodetti, [14] 2-150 kHz Point of Common Connection (LV grid) Characterization of the conducted emissions generated by an electric bus inductive charging Slangen, [15] 0-100 kHz Testlab of ElaadNL Characterization of the conducted emissions generated by 9 EVCPs Sch¨ ottke, [16] 2-150 kHz Laboratory/LV grid Characterization of the conducted emissions generated by 6 EVCPs Darmawardana, [17] 2-150 kHz LV grid/Waveform generator (supply) Characterization of the conducted emissions generated by two EVCPs Gonz´ alez-Ramos, [18] 9-500 kHz LISN Characterization of the conducted emissions generated by two EVCPs Espín-Delgado, [20] 2-150 kHz LV grid (TU Dresden laboratory) Propagation and interaction study of the conducted emissions generated by PV inverters and LEDs based on a correlation and impedance analysis Cassano, [23] 5.1 kHz and 15.1 kHz Simulations (LV grid) Propagation and interaction analysis of the conducted emissions generated by a fast DC EVCS based on a primary/secondary emission approach Sutaria, [26] 2-150 kHz Laboratory at the University of Luleå Characterization of the conducted emissions generated by different Power Factor Corrected circuits. The propagation of these emissions is analyzed based on a primary/secondary emission approach Streubel, [29] 2-150 kHz Three different parking garages with charging infrastructures Characterization of the conducted emissions generated by three EV charging infrastructures in the long-term Slangen, [30] 0-100 kHz Smart Grid Interoperability Lab Characterization of the conducted emissions generated by four EVs. The propagation and interaction of these emissions are analyzed based on a primary/secondary emission approach Current paper 9-500 kHz Controlled LV distribution grid Definition of a novel procedure for the characterization of the conducted emissions generated by four EVCPs (time and frequency domains). The propagation and interaction of these emissions is analyzed based on synchronized measurements at different electrical points in the LV distribution grid Fig. 1. Measurement scenario in the “Concept Grid” laboratory of EDF in ´ Ecuelles. J. Gonz´ alez-Ramos et al. Electric Power Systems Research 231 (2024) 110289 4 locations, three measurement systems as the one presented in Fig. 2 are needed, as well as an identical time setting in each computer. For this purpose, a GPS module is required for generating a Pulse per Second (PPS) signal with a precision of 20 ns. In this way, a deviation between computers of less than a millisecond is achieved. In each measurement, the GPS signal is recorded at the same time as the disturbances. 4.3. Signal processing Once the emissions are recorded, the results are represented in a color scale with respect to frequency (horizontal axis) and time (vertical axis) in the form of spectrograms by means of a Matlab script, providing a time and frequency resolution of 2 ms and 50 Hz, respectively. For this purpose, a sliding Lanczos windowing with a duration of 20 ms and an overlapping of 90% between consecutive windows is applied. The Lanczos window is defined in (1), where N is the number of samples per 20 ms window. Then, in order to obtain the measured spectrum every 2 ms, a ShortTerm Fourier Transform (STFT) is applied to each time window according to (2), Z[fc,k] = ∑ N−1 n=0 x[n−k]w[n]e−j2 π fn (2) where f c are the frequency components, and k the time steps between consecutive STFT outputs. As the recordings of the emissions last 600 s (except for EV2 at 75% SoC, where the emissions were recorded for 250 s), due to computational effort, only 50 s of the recordings are represented. As an example, Fig. 3 shows the spectrogram during the first 50 s of the emissions generated by the charging process of EV1. Subsequently, in order to characterize the disturbances in the frequency domain, the QP values of the amplitude of the emissions are obtained according to CISPR 16–1-1 standard [33], obtaining a single figure of QP values of the amplitude of the NIEs as a function of frequency for each time interval of 50 s. By contrast, the time characterization of the emissions is performed by evaluating the variability of each frequency bin over the measurement time. 5. Characterization of individual emissions 5.1. Frequency analysis Fig. 4 shows the QP values of the amplitude of the emissions generated by the EVCP1, EVCP2 at 75% and 100% SoC, and EVCP3, respectively, together with the emissions corresponding to the default situation. As no emission limits have been specified for the conducted disturbances generated by EVs during the charging process, the out-of-band limits defined for communications equipment in EN 50065-1 [37] might be considered as a conservative criterion [38] for comparison purposes. In general, the amplitude of the emissions is very high if compared to the background noise corresponding to the default situation in which no EV is connected to the grid. This does not go in line with the observations in [16], where no significant emission is visible above 50 kHz. This reinforces the idea that the emissions are very dependent on the specific characteristics of the charging process under analysis. For EVCP1 and EVCP2, the measured disturbances are in the form of harmonics of the switching frequency of an inverter of the EV charger [7],[13],[16]. For instance, Fig. 4(a) shows this spectral pattern, where harmonics of 10 kHz (amplitudes between 62 dBµV and 98 dBµV) are measured in the whole frequency band. In the case of EVCP2, harmonics of 16 kHz with amplitudes varying between 82 dBµV and 115 dBµV at 75% SoC and between 82 dBµV and 106 dBµV at 100% SoC are registered. By contrast, as shown in Fig. 4(c), a high-amplitude emission decreasing with frequency in the form of colored noise is observed for EVCP3 between 70 kHz and 500 kHz (amplitudes varying from 79 dBµV Table II Technical specifications of EVCS1 and EVCS3 [34,35]. EVCS Model Mode Output Voltage Output Current Maximum singlephase power EVCS1 NSQC442G 3 50–500 V DC 0 – 120 A DC – EVCS3 WittyXEV100 3 230-400 V (Adjust.) – 32 KVA Fig. 2. Measurement system for Non-Intentional Emissions. Fig. 3. Spectrogram of the disturbances generated by the charging process of EV1 during 50 s. w[n] = ⎧ ⎪ ⎪ ⎪ ⎨ ⎪ ⎪ ⎪ ⎩ sinc(2(2n N−1))sinc(2n N−1−1),n∕= N−1 2 1,n=N−1 2 ,n=0,1,…,N−1(1) J. Gonz´ alez-Ramos et al. Electric Power Systems Research 231 (2024) 110289 5 to 48 dBµV). This might be due to an efficient grid-side filter circuit, PWM-controlled devices using active power factor correction and/or a control regime that continuously changes the switching frequency [7,13, 30]. In order to verify this, a time domain analysis is required (see section V.B). Moreover, Fig. 4(b) demonstrates that the amplitudes of the spectral components do not necessarily decrease with frequency, as observed in [14]. Fig. 4(b) also shows that the spectral pattern varies with the state of charge [13,16,39]. Although similar amplitudes are obtained at the narrowband emissions if both states of charge are compared, there is a minor shift in the fundamental frequency of the spectral components and, hence, the high-amplitude narrowband emissions are not centered at exactly the same frequencies. This effect is also found in the results presented in [13,17]. As an example, Fig. 5 shows the emission generated by the charging process of EV2 at 100% and 75% SoC around 350 kHz. The peaks of the emission are located at 351.9 kHz and 352.4 kHz respectively. Therefore, the previous figures clearly show that the QP values of the amplitude of the emissions highly depend on the particularities of the EVCPs (EVCS, EV model and state of charge), as different spectral components and amplitudes of the emissions have been registered. This effect might be caused by the filter circuits of the EVs [13]. Considering that the details of the electronic components are not usually provided by the manufacturers, this highlights the need for basing the analysis of emissions from EVCPs on experimental evidence [17]. In order to numerically characterize the QP values of the amplitude of the emissions for the whole frequency band of interest, the Total Supraharmonic Voltage (TSHV) [40–42] is calculated for the four situations under study according to (3). TSHV = ∑ i=9821 i=1 Vi2 √ √ √ √(3) This parameter, calculated in linear scale and then converted to logarithmic, gives a general overview of the amplitude of the disturbances generated by each EVCP in the whole frequency band. The summation considers the QP values of the amplitude of the emissions of the 9821 frequency bins covering the band under analysis (9-500 kHz) with a frequency step size of 50 Hz. Moreover, in order to determine if the emissions above the limit are concentrated at specific frequencies or spread over wider frequency bands, the percentage of frequency bins exceeding the PLC out-of-band emission limits (PFBL) is calculated. PFBL =number of frequency bins above the limit total number of frequency bins ⋅100 (4) Fig. 4. QP values of the amplitude of the emissions (dBµV) generated by EVCP1 (a), EVCP2 at 75% and 100% SoC (b), and EVCP3 (c). J. Gonz´ alez-Ramos et al. Electric Power Systems Research 231 (2024) 110289 6 In Table III, the TSHV and PFBL are gathered for each situation under study. Table III clearly shows that the emissions generated by the charging process of EV1 in the 9-500 kHz frequency band are considerably lower in comparison with EVCP2 and EVCP3 (differences between 9 dB and 19 dB), in addition to being concentrated in fewer frequency bins (18%). If the TSHV values for EVCP2 at 75% and 100% SoC are compared, it can be concluded that the emissions introduced by this EVCP are higher when EV2 is not fully charged. This difference in the emission amplitudes is also reflected in the noise floor obtained at 75% and 100% SoC, which explains the change in the PFBL from 97% at 75% SoC to 62% at 100% SoC. Finally, it should be mentioned that, despite distributing the emissions in only 48% of the frequency bins, the charging process of EV3 generates the highest disturbance amplitudes according to the TSHV. The PFBL shows if the emission amplitudes above the PLC out-ofband emission limits are concentrated at particular frequencies or spread over wider frequency bands, but it does not reflect the spectral distribution of the emissions over the 9-500 kHz frequency band. In order to show the spectral distribution of the emissions above the PLC out-of-band emissions limits, the Cumulative Distribution Function (CDF) of the frequency bins that correspond to emission amplitudes above the emission limit is calculated for each situation under test. In Fig. 6, the obtained CDFs for EVCP1, EVCP2 at 75% and 100% SoC, and EVCP3, are depicted. As shown in Fig. 6, the slope of the CDF of EVCP1 is much steeper at frequencies above 150 kHz, which implies that a higher number of frequency bins exceed the PLC out-of-band emission limits in the frequency range 150-500 kHz. A similar behavior is observed above 70 kHz if the CDF of EVCP2 at 100% SoC is analyzed. Regarding the CDF obtained for EVCP2 at 75% SoC, a linear trend is shown in almost the whole frequency band of analysis, which demonstrates a continuous uniform distribution from 23 kHz to 500 kHz (i.e., all the emissions are above the limit in this frequency range). Again, the difference in the results obtained for both SoCs is affected by the difference in the noise floor measured in the two situations. Finally, the CDF of EVCP3 allows to conclude that, as shown in Fig. 6, all the frequency bins exceed the limits between 9 kHz and 250 kHz (uniform distribution in this frequency band). These results are of practical interest for planning PLC systems in this frequency band because, as shown in Fig. 4, the QP values of the amplitude of the emissions generated in the four situations under study exceed the limits defined for the PLC out-of-band emissions at least for some frequency bands. If the eight channels defined for PRIME v1.4 are considered [43], and evaluating the CDFs obtained in Fig. 6, it is possible to opt for the communication channels that present lower disturbance amplitudes. For example, in presence of emissions similar to the situation corresponding to the charging process of EV1, channel 1 (42-89 kHz) and channel 2 (97–144 kHz) would present better channel conditions in terms of disturbances, as very few spectral components of the emissions are above the PLC out-of-band emission limits. In the case of EV3, channels 5 (261-308 kHz), 6 (315–362 kHz), 7 (370–417 kHz) and 8 (424-471 kHz) would imply less disturbances caused by the EVCP. 5.2. Time analysis The measurements give rise to spectrograms that are composed of many samples corresponding to each frequency bin and time instant, which cover a wide range of amplitude values. Therefore, due to computational issues, the analysis in the time domain is performed for shorter time periods of 50 s length. The analysis of the time variability of the disturbances is organized as follows. First, in section V.B.1), the differences in the QP values of the amplitudes of the emissions are compared for the resultant time periods of 50 s. Second, in V.B.2), the variations over time given within the first 50 s period of each recording are modelled by means of a FFT analysis. 5.2.1. Analysis of the time variability within the recording time The analysis of the time variability within the recording time is based, first, on the comparison of the QP values of the amplitude of the emissions calculated over each period of 50 s within that recording time (600 s, i.e. 12 periods, except for EVCP2 at 75% SoC, where 250 s are available, i.e. 5 periods). In Fig. 7, the QP values of the amplitude of the emissions generated by the charging process of each EV over each period of 50 s are superimposed, together with the PLC out-of-band emissions limits. Although the spectral form of the emissions is maintained in the whole frequency band for the consecutive periods of 50 s for the four Fig. 5. Frequency shift of the emission generated by EVCP2 around 350 kHz for different states of charge. Table III TSHV and PFBL for each EVCP. EVCP TSHV PFBL EVCP1 112 dBµV 18% EVCP2 75% 128 dBµV 97% EVCP2 100% 121 dBµV 62% EVCP3 131 dBµV 48% Fig. 6. CDFs of the frequency bins that correspond to QP emission amplitudes above the PLC out-of-band emission limits. J. Gonz´ alez-Ramos et al. Electric Power Systems Research 231 (2024) 110289 7 situations under study, slight variations in the amplitude of the disturbances can be observed at certain frequencies. For this reason, in order to quantify these differences, the TSHV and PFBL are calculated for the 12 periods of 50 s (5 periods in the case of EVCP2 at 75% SoC). Then, the following figures are obtained in order to characterize the maximum difference of the TSHV (ΔTSHV) and PFBL (ΔPFBL) for each analyzed situation: ΔTSHV =max i∈{1,2,…n}TSHVi−min i∈{1,2,…n}TSHVi(5) ΔPFBL =max i∈{1,2,…n}PFBLi−min i∈{1,2,…n}PFBLi(6) where n =12 except for EVCP2 at 75% SoC, where n =5. In Table IV, ΔTSHV and ΔPFBL values for each EVCP are gathered. In the case of EVCP1, a maximum difference of the TSHV of 3 dB is observed between different periods of the emissions generated by its charging process. In the remaining analyzed cases, these differences do not exceed 1 dB. Regarding the PFBL, in the cases of EV1 and EV2 at 75% SoC, maximum differences of 1% and less than 1% are obtained, respectively. The differences in the PFBL for EV2 at 100% SoC and EV3, in turn, increase up to 11% and 6%, respectively. However, these larger differences do not result in great variations in the TSHV in that cases, which implies that the emissions exceeding occasionally the limits do not present high amplitude. Therefore, it can be concluded that, in general, the disturbances generated by the EVCPs under study present a quasi-stationary behavior in periods ranging from 50 s to several minutes. Apart from that, although not explicitly shown in Table III, it is important to highlight that the TSHV of the last period of 50 s for EV2 at 75% SoC (129 dBµV) is significantly higher if compared to the TSHV recorded for the first 50-second period at 100% SoC (122 dBµV). This means that the emissions can substantially differ for different states of charge. Therefore, despite the fact that it is not possible to determine when these time variations occur, it can be assumed that variations Fig. 7. QP values of the amplitude of the emissions (dBµV) generated by EVCP1 (a), EVCP2 at 75% (b) and 100% (c) SoC, and EVCP3 (d) in each period of 50 s within the recording time. Table IV ΔTSHV and ΔPFBL. ΔTSHV ΔPFBL EVCP1 3 dB 1% EVCP2 at 75% SoC 1 dB 0% EVCP2 at 100% SoC 0 dB 11% EVCP3 1 dB 6% J. Gonz´ alez-Ramos et al. Electric Power Systems Research 231 (2024) 110289 8 given in periods greater than several minutes due to changes in the state of charge are to be expected. As EVs use constant current/constant voltage modes for charging their batteries [44], this effect can be due to the charging profile of these two modes. Abrupt differences in the emission amplitudes 30 minutes after start of the charging cycle have been also reported in [16]. 5.2.2. Analysis of the time variability within 50 s In the previous section, it was concluded that no time variations in periods longer than 50 s are occurring, since similar TSHV and PFBL were obtained in each 50-second period within the recording time. For this reason, time variations can only occur in periods of less than 50 s. In order to quantify this potential time variability, a FFT is applied to the time samples corresponding to the first 50-second period of each frequency bin of the spectrogram, obtaining the FFT frequency components of the time variability of each frequency bin. With the aim of comparing the amplitudes of the FFT frequency components corresponding to the different frequency bins, the resultant FFTs are normalized with respect to the corresponding amplitude of the FFT component at 0 Hz. As an example, in Fig. 8, the modulus of the normalized FFT obtained for the first 50-second period of the emissions generated by the charging process of EV2 at 100% SoC at the frequency bin 383.901 kHz is presented. Fig. 8 only shows the positive part of the FFT spectrum, since the emissions generated by the charging process of an EV are real signals, whose FFT is even. In the 39,284 analyzed frequency bins (9821 frequency bins/signal ⋅ 4 signals), the modulus of the normalized FFTs were similar to the one shown in Fig. 8, where narrowband FFT components at 50 Hz, 100 Hz, 150 Hz, 200 Hz and 250 Hz are observed. This FFT pattern corresponds to a periodic function with period T =1/50 =20 ms [45]. Therefore, the normalized FFT, regardless of the frequency bin and EVCP under study, can be characterized by a simplified model composed of the main components at 50 Hz, 100 Hz, 150 Hz, 200 Hz, and 250 Hz. For each frequency bin, the main amplitude of those components is detected using the normalized FFT in dB, as represented in Fig. 8. Then, a simplified model is proposed in linear scale according to (7): FFT (fFFT) = δ(fFFT) + ∑ 5 n=1 cn⋅δ(fFFT −50n)(7) where δ(fFFT)is the unit sample function. In (7), c n are 2ℼ times the complex Fourier coefficients of the signal, given in linear scale and normalized with respect to the 0 Hz FFT component. With the aim of determining how the original signal resembles the signal obtained by approximating the FFT with the proposed simplified model (synthesized signal), the absolute value of the Inverse FFT (IFFT) of the impulse train in (7) is calculated and expressed in logarithmic scale. For this purpose, the non-normalized coefficients in (7) are considered. In Fig. 9, the original and synthesized emissions generated by the charging processes of EVCP2 at 100% SoC at 255.951 kHz (a) and EVCP3 at 10 kHz (b) are shown during a time period of 0.11 s. As it can be observed, the synthesized signal approximates accurately the amplitude of the original signal, as well as the peak-to-peak and the shape of the variation. This means that the complex coefficients of the simplified model cnprovide a quantitative characterization of the signal variation in the time domain. Therefore, in order to quantify the total time variations of the emissions and relate their time-dependent behavior with their spectral characteristics, the total variability of the disturbances is calculated, as shown in (8), as the sum in linear scale of the modulus of the amplitude of the FFT components at 50 Hz, 100 Hz, 150 Hz, and 200 Hz, which is then converted to logarithmic. Total variability (dB) = 20⋅log10(∑ 4 n=1 |cn|)(8) For the two examples shown in Fig. 9, the total variability is -11.8 dB for (a) and -7.9 dB for (b), so that the higher the sum of the FFT components, the greater the variation of the signal over time. Fig. 10 shows the total variability of the disturbances obtained in terms of the FFT components, along with the QP values of the amplitude of the emissions. Similarly as obtained for the frequency analysis, the time variations of the amplitudes of the NIEs also depend on the specific charging process (EVCS, EV model, and state of charge). Finally, from Fig. 10 it can be assumed that, while the background noise remains practically static over time in the four charging situations under study, higher variations occur at the frequency bins corresponding to disturbances, both for narrowband emissions and colored noise. 6. Propagation of the measured emissions 6.1. Frequency analysis of the propagation of the emissions In order to analyze the propagation of the emissions in the frequency domain, the QP values of the amplitude of the emissions measured at H2, H3, and H5 are superimposed when each EV model is charged individually. Fig. 11 shows, for the three measurement locations, the emissions generated by the charging process of EV1 (a), EV2 at 75% (b) and 100% (c) SoC, and EV3 (d). As it can be seen, the emissions do not remain in the proximity of the source and they propagate several meters through the electrical grid. In most of the cases, the spectral patterns of the disturbances are maintained, but they are attenuated with the distance. In order to clearly observe this behavior, Fig. 12 shows, as an example, a zoom of the QP values of the amplitude of the emissions (dBµV) generated by EVCP1 measured at H2, H3, and H5 in the frequency range 230-250 kHz. If the emission at 240 kHz is taken into account, a decrease in the amplitudes generated by the charging process of this EV model can be observed when measuring at distances of 46 m (POC of H3) and 120 m (POC of H5). The generated emission suffers an attenuation of 5 dB (from 79.5 dBµV to 74.5 dBµV) if the NIEs are measured at H3, while the attenuation increases up to 10.7 dB (from 79.5 dBµV to 68.8 dBµV) at H5. By contrast, at certain frequencies of the spectrum when EV1 is charging, higher emissions are measured at locations different to the POC of the source. As an example, Fig. 13 presents the QP values of the amplitude of the emissions generated by EVCP1 at H2, H3, and H5 in the frequency range 70-90 kHz. Despite EV1 being charged at H2, the component around 80 kHz is greater when measured at H5 than when measured at the same POC where the EV is charging. Fig. 8. Modulus of the normalized FFT (dB) of the time samples corresponding to the emissions generated by EVCP2 at 100% SoC at 383.901 kHz. J. Gonz´ alez-Ramos et al. Electric Power Systems Research 231 (2024) 110289 9 Fig. 9. Synthesized and original emissions generated by the charging processes of EVCP2 at 100% SoC at 255.951 kHz (a) and EVCP3 at 10 kHz (b). Fig. 10. Normalized amplitude (dB) of the variability of the emissions generated by EVCP1 (a), EVCP2 at 75% (b) and 100% (c) SoC, and EVCP3 (d) in the 9-500 kHz frequency band during the first 50 s of measurement time together with the QP values of the amplitude of the emissions. J. Gonz´ alez-Ramos et al. Electric Power Systems Research 231 (2024) 110289 10 As it is stated in [46],[47], the introduction of modern energy-efficient appliances implies resonances in the grid access impedance, which may involve significant increases in the emissions at the switching frequency and its harmonics. This might be the main cause why, in spite of the attenuation suffered by the generated emissions due to the distance, the amplitudes of the disturbances are higher at an electrical point distant from the source of emissions. Hence, as it is pointed out in [23],[25],[46–48], resonances in the grid impedance are a key aspect in the amplitude of NIEs generated by electronic devices. An analysis of the influence of the grid impedance on the propagation of the disturbances is presented in [49]. Reference [7] points out that a high-impedance parallel resonance causes high-amplitude emissions at the resonance frequency, whereas a decrease in the disturbance amplitudes is observed if there is a series resonance with low impedance values. As carried out in section V.A, in order to numerically characterize the propagation of the emissions, TSHV and PFBL values are calculated for the disturbances recorded at each location for each charging situation under test (see Table V). The results presented in Table V lead to conclude that higher TSHVs are obtained at the POC of each house to which the EVCS under study is installed. This goes in line with the conclusions presented from Fig. 12, where, in general, a decrease in the emissions with distance was observed. The greatest difference occurs when comparing the TSHV calculated for EVCP3 at H5, 131 dBµV, and H3, 124 dBµV (difference of 7 dB). However, as concluded from Fig. 13, the distance is not the only factor affecting the amplitude of the emissions, since a resonance in the grid access impedance could imply an increase in the disturbances at that frequency. For this reason, similar or even higher TSHVs are calculated at electrical points more distant from the source of the emissions than at nearby POCs. This behavior can be observed when comparing, for example, the TSHV for EVCP1 at H3 and H5. For this EV, an identical TSHV is obtained at both locations, despite the fact that H5 is 74 meters farther away from H2 than H3. This effect arises in the four charging situations under study. The PFBL calculated for EVCP2 at 75% SoC and EVCP3 is similar at the three POCs where the emissions are registered. In the case of EVCP2 at 75% SoC, values close to 100% (90%-100%) are obtained at the three measurement locations. The slight differences are due to the noise floor, which does not exceed the PLC out-of-band emissions limits at some frequencies when measuring the disturbances at H2 and H5. Similarly, this is the case for EVCP3, where 50% of the frequency bins (approximately up to 250 kHz) correspond to emission amplitudes above the emission limits at H2, H3, and H5. By contrast, for EVCP2 at 100% SoC, Fig. 11. Synchronized QP values of the amplitude of the emissions (dBµV) measured at H2, H3, and H5 generated by EVCP1 when EV1 is charging at H2 (a), EVCP2 at 75% and 100% SoC when EV2 is charging at H3 (b), and EVCP3 when EV3 is charging at H5 (c). J. Gonz´ alez-Ramos et al. 142 A.3. Journal Paper JP3 This subsection presents the following journal paper:  J. González-Ramos, A. Gallarreta, I. Fernández, I. Angulo, A. Arrinda, D. de la Vega, “Comparison of Conducted Emissions Due to Electric Vehicle Charging Processes under Isolated and On-Line Conditions in the 9-500 kHz Frequency Range”, Sustainable Energy, Grids and Networks, Volume 38, 2024, 101333, ISSN 2352-4677, https://doi.org/10.1016/j.segan.2024.101333. The most representative quality indicators concerning this paper are listed below:  Publisher: Elsevier  Journal: Sustainable Energy, Grids and Networks  Year of publication: 2024  Type of publication: Journal paper indexed in JCR  Area: Electrical & Electronic Engineering | Energy & Fuels  Ranking (JCR - 2023): 70/353 (Q1) | 75/171 (Q2)  Impact factor (JCR - 2023): 4.8 Sustainable Energy, Grids and Networks 38 (2024) 101333 Available online 28 February 2024 2352-4677/© 2024 The Author(s). Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/bync-nd/4.0/). Comparison of conducted emissions due to electric vehicle charging processes under isolated and on-line conditions in the 9–500 kHz frequency range Jon Gonz´ alez-Ramos a , * , Alexander Gallarreta a , Igor Fern´ andez a , Itziar Angulo b , David de la Vega a , Amaia Arrinda a a Department of Communications Engineering, University of the Basque Country (UPV/EHU), Bilbao, Spain b Department of Applied Mathematics, University of the Basque Country (UPV/EHU), Bilbao, Spain ARTICLE INFO Keywords: Conducted non-intentional emissions Electric vehicle charging process Low voltage grid Line impedance stabilization network ABSTRACT This paper aims at comparing the conducted emissions generated by Electric Vehicle Charging Processes (EVCPs) under isolated and on-line conditions in the frequency and time domains, covering the 9–500 kHz frequency range. The isolated conditions correspond to the use of a Line Impedance Stabilization Network (LISN), whereas measurements in the Low Voltage (LV) grid where the EV is connected are referred to as on-line conditions. Regarding the frequency analysis, the results lead to conclude that different amplitude and spectral features of the Non-Intentional Emissions (NIEs) are measured depending on the measurement conditions, registering higher-amplitude NIEs when the measurements are conducted in the LV grid. Moreover, the paper also shows that the spectral characteristics of the NIEs correspond to tonal and narrowband emissions at specific frequencies in the 9–150 kHz frequency range, while background noise with low-amplitude emissions (in the case of isolated conditions) and colored noise (in the case of on-line conditions) are reported in the 150–500 kHz band. In the time domain, a periodic time-dependent behavior within the fundamental cycle of the mains (20 ms) is reported in all the analyzed cases, except for the tonal emissions with oscillating central frequency generated by a specific EV model. The time variability between both measurement configurations only differs by more than 3 dB for three out of the twelve analyzed EVCPs, where a higher variability is observed if the measurements are carried out using a LISN. As a conclusion, considering the differences in the frequency and time characteristics of the NIEs under isolated and on-line conditions, evaluating the emissions generated by EVCPs might lead to an underestimation of the NIEs in real LV grids. 1. Introduction In recent years, there has been a considerable increase in the number of inverter-based devices connected to the power grid, such as Electric Vehicles (EVs) charging infrastructure or Photovoltaic (PV) panels. This kind of power electronic devices generate Non-Intentional Emissions (NIEs) at the switching frequency of the inverter and its multiples [1,2]. Although the improvement of the efficiency of the power conversion has led to a reduction in the amplitude of the distortion in the harmonic frequency range (<2 kHz), the emission levels in the 2–500 kHz frequency band have significantly increased [3]. These high-amplitude and time-variant distortions cover the whole Narrowband Power Line Communications (NB-PLC) frequency range and can be a critical aspect for communications using the Low Voltage (LV) grid infrastructure [1,4, 5]. Moreover, they can lead to Power Quality (PQ) issues, such as equipment malfunction, reduction of the lifetime of the electronic Abbreviations: BPF, Band-Pass Filter; CDF, Cumulative Distribution Function; EUT, Equipment Under Test; EV, Electric Vehicle; EVCS, Electric Vehicle Charging Station; EVCP, Electric Vehicle Charging Process; FFT, Fast Fourier Transform; LISN, Line Impedance Stabilization Network; LV, Low Voltage; NB-PLC, Narrowband Power Line Communications; NIE, Non-Intentional Emission; PFBHOn-line, Percentage of frequency bins in which the emissions are higher when the measurement is carried out under on-line conditions; PFBL, Percentage of frequency bins exceeding the PLC out-of-band emission limits; POC, Point of Connection; PQ, Power Quality; PV, Photovoltaic; RF, Radiofrequency; TSHV, Total Supraharmonic Voltage; QP, Quasi-Peak. * Corresponding author. E-mail address: [email protected] (J. Gonz´ alez-Ramos). Contents lists available at ScienceDirect Sustainable Energy, Grids and Networks journal homepage: www.elsevier.com/locate/segan https://doi.org/10.1016/j.segan.2024.101333 Received 30 November 2023; Received in revised form 2 February 2024; Accepted 24 February 2024 Sustainable Energy, Grids and Networks 38 (2024) 101333 2 devices, audible noise or light flicker, among others [6,7]. Up to now, several contributions have been published addressing the characterization of the emissions generated by EV charging processes (EVCPs) in different measurement conditions covering the frequency range from 2 kHz to 150 kHz (also referred to as supraharmonic range). Some of these works have been carried out in the LV distribution grid [8–13], where a more realistic overview of the behavior of the disturbances is offered. In this scenario, the influence of other connected devices and the frequencyand time-dependent grid impedance cannot be disregarded [14]. Other studies, such as [15–17], have addressed this characterization in reconstructed LV grids with the aim of simulating real grid conditions in a controlled scenario. Finally, in other publications [1,18], the EV is connected to a Line Impedance Stabilization Network (LISN). This way, the recorded emissions are only due to the EVCP under study and are evaluated for a well-characterized impedance. In this sense, they can be considered as isolated conditions [15]. There is a lack of studies in the literature in which a comparison of the emissions generated by the same EVCPs under both isolated and LV grid conditions is addressed. In this context, this paper deals with the characterization of the emissions generated by twelve EVCPs, comparing the results obtained for the measurements in the LV grid (online conditions) and when using a commercially available LISN (isolated conditions). The study evaluates the disturbances in the frequency and time domains and covers the 9–500 kHz frequency band. This paper is organized as follows. In section II, the methodology of the study is presented. This includes the description of the measurement scenarios for the recording of the emissions with and without LISN, the signal processing applied to the recorded emissions, and a summary of the empirical data obtained. Section III discusses the frequency and time characterization of the emissions. In section III. A, a comparison of the spectral features of the disturbances generated by these EVCPs are presented, while section III. B gives an overview of their time-dependent behavior. Finally, in section IV, the main conclusions of this study are gathered. 2. Methodology 2.1. Description of the measurement scenario 2.1.1. Measurement scenario for the recording of the emissions under isolated conditions The on-site measurements were carried out on a commercial Electric Vehicle Charging Station (EVCS) operating in mode 3 according to IEC 61851–1 [19]. All the measurements presented in this article were performed in monophasic mode. With the aim of implementing a controlled measurement scenario, the setup was isolated from the LV grid by means of a LISN [20], a small transformer (Polylux PD400 [21] with a rated power of 4000 VA), and a band-pass filter (BPF) adjusted to the PLC frequency band. Both the LISN and the transformer operated in monophasic mode. The EVCS was directly connected to the Equipment Under Test (EUT) port of the LISN. The resulting emissions were measured from the Radiofrequency (RF) port with a digital oscilloscope, which was controlled by a laptop charging at a power station. The oscilloscope (Picoscope Series 5000 [22]) registered the NIEs with a vertical resolution of 16 bits, using a sampling rate of 4.1667 MS/s for the frequency range 9–500 kHz. As the measurements are obtained from the RF port, a correction factor should be applied in the post-processing stage to compensate the insertion loss of the LISN between the EUT and RF ports. Typical values of the insertion loss of the LISN as a function of frequency can be found in [20]. In Fig. 1, a representation of the measurement setup used for the recording of the NIEs using a LISN is shown. 2.1.2. Measurement scenario for the recording of the emissions under online conditions As shown in Fig. 2, the measurements were performed at a parking plaza, where the EVCS was directly connected to the LV distribution grid. As in section II. A. 1), the measurement system for the assessment of the NIEs is composed of an oscilloscope responsible for registering the emissions with a vertical resolution of 16 bits and a sampling rate of 4.1667 MS/s. In order to avoid the measurement system from being damaged by the fundamental component of the grid (220 V, 50 Hz), the voltage probe presented in [23] is connected as a coupling circuit at the Point of Connection (POC). This voltage probe was designed for measuring the NIEs under unknown grid impedance conditions in the frequency range under analysis. 2.2. Signal post-processing Once the disturbances are measured, a Lanczos window with a duration of 20 ms and an overlapping of 90% is applied to the recorded signals [24]. Subsequently, in order to characterize the disturbances in the frequency domain, the Quasi-Peak (QP) values of the amplitude of the emissions are obtained according to the CISPR 16–1–1 standard [25]. More detailed information about the signal post-processing stage can be found in [16]. 2.3. Summary of the recorded emissions The analysis presented in this paper is based on the disturbances generated by five commercially available EV models. With the aim of Fig. 1. Measurement scenario for the recording of the NIEs generated by EVs during their charging process using a LISN (isolated conditions). Fig. 2. Measurement scenario for the recording of the NIEs generated by EVs during their charging process in the LV grid (on-line conditions). J. Gonz´ alez-Ramos et al. Sustainable Energy, Grids and Networks 38 (2024) 101333 3 analyzing the potential influence of the charging current on the recorded emissions, the different charging currents supported by the EVCS are sequentially configured for each EV model under study. In the case of EV1, only one charging current was possible (16 A), whereas two charging currents were possible for EV4 (12 A, 16 A). For the rest of the EV models under test, three charging currents were evaluated (8 A, 12 A, and 16 A). Therefore, the study addresses the characterization of the emissions generated by twelve different EVCPs (combinations of EV model and charging current) with and without using a LISN. This characterization considers recordings of 600 s of duration, except for EV3 with a charging current of 8 A under isolated conditions, for which only 150 s were recorded. 3. Results 3.1. Frequency characterization The spectral characterization of the emissions in the 9–500 kHz frequency band, addressed in section III. A. 1), is based on the procedure proposed by the current authors in [16]. First, the QP values of the amplitude of the emissions corresponding to the first 50 s of measurement are calculated and represented according to CISPR 16–1–1 standard [19]. Since no specific emission limits are defined for the charging of EVs in this frequency range, the PLC out-of-band emission limits defined in EN 50065–1 [20] are included for comparison purposes. Then, the Total Supraharmonic Voltage (TSHV), the percentage of frequency bins exceeding the PLC out-of-band emission limits (PFBL), and the Cumulative Distribution Function (CDF) of the frequency bins corresponding to disturbances above these limits are calculated. Detailed information about the definition and calculation of these parameters can be found in [16]. Moreover, in order to thoroughly analyze the behavior of the emissions in the ranges 9–150 kHz and 150–500 kHz, separate analyses of each frequency band are presented in sections III. A. 2) i) and III. A. 2) ii). 3.1.1. Comparison of the emissions generated by EVCPs in the 9–500 kHz frequency band As an example, in Fig. 3, the QP values of the amplitude of the emissions generated by EV2 and EV3 using a charging current of 8 A for the first period of 50 s of recording are presented under isolated and online conditions. Fig. 3 clearly shows that both the spectral shape and the amplitude of the emissions depend on the measurement conditions. The disturbances measured using a LISN are lower in practically the whole frequency band, except for emissions occurring at certain frequencies (for example, around 45 kHz for EVCP3 with a charging current of 8 A). It should be noted that, unlike when the emissions are recorded directly in the LV grid, the disturbances measured using a LISN do not generally exceed the PLC out-of-band limits. According to Fig. 3, there are differences in the emissions recorded for different models of EV, even though the same EVCS was used. However, similar behaviors can be observed. In the 9–150 kHz frequency range, EVCPs generate tonal emissions (for example, around 45 kHz and 90 kHz for EVCP3 and a charging current of 8 A) or narrowband emissions (for instance, between 45 kHz and 50 kHz and between 90 kHz and 98 kHz for EVCP2 and a charging current of 8 A). These emissions show similar amplitude in both measurement conditions. Since the narrowband emissions might be caused by a periodically changing switching frequency [17], a time analysis is required (see section III. B). By contrast, both the amplitude and the spectral shape of the background emissions vary considerably in this frequency band if isolated and on-line conditions are compared. In the 150–500 kHz frequency range, flat and low-amplitude distortions (40 dBµV, approximately) are measured when a LISN is used for both EVCPs. Under on-line conditions, in turn, colored noise decreasing with frequency (from 80 dBµV to 55 dBµV for EV2 and from 70 dBµV to 50 dBµV for EV3) is recorded. In order to show that the spectral pattern of the emissions is similar for the 12 EVCPs under study, in Fig. 4, the QP values of the amplitude of the emissions generated by the twelve analyzed cases are presented under isolated (a) and on-line (b) conditions. Fig. 4 shows that the behavior observed for EV2 and EV3 with a charging current of 8 A in Fig. 3 applies to the twelve EVCPs under study. Therefore, these results lead to conclude that the understanding of the amplitude and spectral features of the emissions generated by EVCPs from 9 kHz to 150 kHz, the frequency range that is normally addressed in the literature, cannot be extrapolated to the 150–500 kHz frequency band. This higher frequency range is currently being used in North America for PLC transmissions and it is expected to be also adopted in Europe in the coming years. For this reason, further studies considering these higher frequencies are still needed. In order to give an insight into the amplitude of the emissions in the 9–500 kHz frequency band in all the analyzed cases and perform a quantitative comparison between both measurement configurations (with and without LISN), the TSHV and PFBL are calculated for all the (a) (b) Fig. 3. QP values of the amplitude of the emissions (dBµV) generated by EV2 (a) and EV3 (b) with a charging current of 8 A under isolated and on-line conditions. J. Gonz´ alez-Ramos et al. Sustainable Energy, Grids and Networks 38 (2024) 101333 4 EVCPs under study as defined in [16] (see Table 1). As shown in Table 1, a higher TSHV can be measured with or without LISN depending on the EV model and the charging current under analysis. For instance, EV2 with a charging current of 16 A shows a lower TSHV when measuring with a LISN (113 dBµV and 135 dBµV when the measurements are performed with and without LISN, respectively). For EV3 with a charging current of 8 A, in turn, a higher THSV is obtained when a LISN is used for the recording of the emissions (132 dBµV when a LISN is used, and 125 dBµV when the measurement is performed directly in the LV grid). Due to the definition of the TSHV (in linear scale), high-amplitude emissions occurring at certain frequencies can contribute significantly to the value of this parameter, in such a manner that they can cause the TSHV to be higher even if the NIEs are considerably lower in the remaining frequency bins. This phenomenon can be clearly observed if the PFBL is analyzed. For all the EVCPs under study, PFBLs lower than 4% are obtained when a LISN is used. By contrast, when the measurements are performed in the LV grid, considerably higher PFBLs are measured, being up to 100% for some EVCPs. This means that, when a LISN is used, there are high-amplitude emissions at specific frequency bins that give rise to TSHV values higher than the ones obtained when the measurement is performed in the LV grid, although in this measurement configuration most of the emissions exceed the amplitude of the NIEs recorded with LISN. In order to show this effect, the percentage of frequency bins in which the emissions are higher when the measurement is carried out under on-line conditions with respect to the isolated conditions is calculated (PFBHOn-line). Moreover, in order to quantify the difference in amplitude between both measurement conditions, the difference of the QP values of the amplitude of the emissions is calculated for each frequency bin f i according to (1). ΔQP(fi) = QPOn−line(fi) − QPIsolated(fi)(1) In Table 2, the minimum, median, and maximum values of the difference for each frequency bin ΔQP(fi)are gathered for each EVCP under study, together with the PFBHOn-line. Table 2 shows that, in all the cases under study, at least in a 99% of the frequency bins, the emissions generated by an EVCP when measuring in the LV grid exceed the distortions measured using a LISN. Moreover, Table 2 also points out that the median difference is higher than 11 dB in all the analyzed cases, which implies that in 50% of the frequency bins, the NIEs measured in the LV grid are at least 11 dB higher than the ones recorded using a LISN. If the maximum differences between both configurations are considered, values greater than 30 dB are reported, with a maximum difference of 49 dB for EV2 with a charging current of 16 A. Regarding the minimum values, it should be mentioned that negative differences are measured for ten out of the twelve analyzed EVCPs. These negative values indicate that the emissions in those frequency bins are higher if a LISN is used. A maximum negative difference of −19 dB is registered for EV5 with a charging current of 16 A. The previous analysis only considers the amplitude of the emissions, without taking into account their spectral distribution in the 9–500 kHz frequency band. For this reason, following the procedure presented in [16], Fig. 5 shows the CDFs of the frequency bins in which the QP values of the amplitude of the emissions exceed the PLC out-of-band emission for the twelve EVCPs when measuring under isolated (a) and on-line (b) conditions. Fig. 5(a) shows that the frequency bins in which the QP values of the amplitude of the emissions exceed the PLC out-of-band emission limits when measuring with a LISN are concentrated in the frequency range up to, approximately, 150 kHz. This means that, above this frequency, the emissions do not exceed the PLC out-of-band limits for this measurement condition. When the measurement is performed directly in the LV grid, in turn, as presented in Fig. 5(b), the spectral distribution of the higheramplitude emissions is spread over a higher frequency range. In four of the twelve analyzed cases, emissions exceeding the limits are reported in the whole frequency range. Therefore, the 150–500 kHz frequency range seems to be more appropriate for PLC transmissions if the amplitude of the emissions measured with a LISN are taken into consideration. However, due to the different behavior observed under on-line conditions, it can be concluded that the planning of PLC networks should not only be based on isolated observations using a LISN. 3.1.2. Comparison of the emissions in the 9–150 kHz and 150–500 kHz frequency bands In Fig. 4, it was shown that the spectral features of the NIEs depend considerably on the measurement conditions and the frequency band. In the 9–150 kHz frequency band, tonal or narrowband emissions at specific frequencies were reported, while background noise with lowamplitude emissions (with LISN) or colored noise (without LISN) were registered in the 150–500 kHz frequency range. This section aims to give a more in-depth evaluation of the differences occurring between isolated and on-line conditions depending on the frequency range under analysis (9–150 kHz and 150–500 kHz). Fig. 4. QP values of the amplitude of the emissions (dBµV) generated by the twelve EVCPs under study considering the recordings under isolated (a) and on-line (b) conditions. J. Gonz´ alez-Ramos et al. Sustainable Energy, Grids and Networks 38 (2024) 101333 5 3.1.2.1. 9–150 kHz frequency band. This section aims at comparing the amplitude of the tonal and narrowband emissions registered between both measurement configurations. Since the field trials conducted with a LISN isolate the NIEs generated by the EVCP itself, without being affected by external sources, this comparison may shed light on how these emissions behave when the EV is charging under real LV grid conditions. For this purpose, this study focuses on the NIEs with an amplitude exceeding 50 dBµV and a prominence of 5 dB. The prominence measures how much a certain narrowband emission stands out due to its intrinsic height and its location relative with respect to other narrowband disturbances. In order to conduct this analysis, first, the emissions fulfilling the previously mentioned conditions in the measurements performed using a LISN are identified. These emissions are represented by means of a red circle in Fig. 6(a). Then, the same procedure is applied to the NIEs recorded directly in the LV grid. The emissions detected in the same frequency bins under both measurement conditions are represented by means of a red circle in Fig. 6(b). If, by contrast, no tonal or narrowband emission is detected when recording the emissions without LISN in a frequency bin where an emission was present under isolated conditions, it is indicated using a black cross. An example of this can be observed in Fig. 6(b) for EV5 with a charging current of 8 A. Fig. 6(a) shows that twenty tonal or narrowband emissions fulfill the established conditions when recording the NIEs generated by EV5 with a charging current of 8 A with a LISN in the frequency band from 9 kHz to 150 kHz. By contrast, in the measurement carried out in the LV grid (see Fig. 6(b)) only three out of these twenty emissions (115 dBµV at 44 kHz, 92 dBµV at 89 kHz, and 71 dBµV at 133 kHz) are identified as tonal or narrowband NIEs. Thus, it seems that most of the emissions generated by EVCPs occurring in isolated conditions cannot be identified when the measurement is conducted directly in the LV grid. In order to prove if this behavior is observed for all the EVCPs, the percentage of the emissions that occur in both measurement situations (common emissions), as well as the frequency (kHz), width at −10 dB (kHz), and amplitude (dBµV) of these NIEs when measuring with and without LISN are gathered in Table 3 for the twelve EVCPs under study. As concluded from Fig. 6, not all the emissions detected with LISN Table 1 TSHV (dBµV) and PFBL (%) for each EVCP under isolated (Is.) and on-line (O-L) conditions. EV1 EV2 EV3 EV4 EV5 16 A 8 A 12 A 16 A 8 A 12 A 16 A 12 A 16 A 8 A 12 A 16 A Is. O-L Is. O-L Is. O-L Is. O-L Is. O-L Is. O-L Is. O-L Is. O-L Is. O-L Is. O-L Is. O-L Is. O-L TSHV (dBµV) 109 123 136 132 137 133 113 135 132 125 109 121 113 125 110 118 111 118 133 124 110 117 113 119 PFBL (%) 1 33 3 100 4 100 3 100 2 35 1 33 1 95 0 12 0 13 2 19 1 12 1 28 Table 2 Percentage of frequency bins in which the emissions generated by the EVCPs are higher under on-line than isolated conditions, together with the minimum, median, and maximum differences between the QP values of the amplitude of the emissions recorded under on-line and isolated conditions. EV model PFBHOn-line (%) ΔQP (f i ) (dB) Min. Median Max. 8 A 12 A 16 A 8 A 12 A 16 A 8 A 12 A 16 A 8 A 12 A 16 A EV1 - - 100.00 - - 4 - - 26 - - 42 EV2 99.01 99.02 99.21 -6 -6 -2 26 26 36 38 39 49 EV3 99.70 99.90 99.99 -15 -4 0 17 24 29 32 37 42 EV4 - 99.92 99.91 - -7 -7 - 18 13 - 34 30 EV5 99.50 99.80 99.47 -14 -4 -19 11 16 21 30 32 36 Fig. 5. CDFs of the frequency bins in which the QP values of the amplitude of the emissions exceed the PLC out-of-band emission limits when measuring under isolated (a) and on-line (b) conditions. J. Gonz´ alez-Ramos et al. Sustainable Energy, Grids and Networks 38 (2024) 101333 6 are also detected under real conditions in the LV grid. In all the analyzed cases, the maximum percentage of common emissions is 50.0%, which is given for EV3 with a charging current of 12 A, and EV3 and EV5 with a charging current of 12 A (see Table 3). Regarding the difference in amplitude between the NIEs in both measurement conditions, in six out of the eighteen analyzed emissions, a higher amplitude is reported when a LISN is used. In these cases, a highest difference of 11 dB is given at 44 kHz for EV5 with a charging current of 8 A. In the remaining cases, in turn, the amplitude of the NIE is higher in the LV grid, with a greatest difference of 20 dB occurring at 45 kHz for EV2 with a charging current of 16 A. Thus, the results do not show a clear trend in terms of the amplitude of the tonal or narrowband emissions when measured in the LV grid compared to measurements performed using a LISN. In four EVCPs under study, tonal or narrowband emissions exceeding 120 dBµV are reported when measuring under isolated conditions, which are several dB lower in amplitude when detected under on-line conditions. These cases showed a higher TSHV when the measurement is carried out using a LISN (see Table 1 and Table 3, EV2 with a charging current of 8 A and 12 A, EV3 with a charging current of 8 A, and EV5 with a charging current of 8 A). Therefore, these high-amplitude NIEs correspond to the emissions with a major contribution to the calculation of this parameter, regardless of the behavior observed for the remaining frequency bins. The differences observed for those emissions that are detected as tonal or narrowband NIE with LISN, but do not fulfill the criteria to be detected when recorded in the LV grid, have been also analyzed. For this purpose, as in section III. A. 1), the minimum, mean, and maximum values of the differences between the emissions measured without and with LISN for each frequency bin are calculated for each EVCP under study. Table 4 gathers the minimum, mean, and maximum differences between both situations, together with the number of emissions detected as tonal or narrowband emission when the recording is performed with LISN, but not identified as so if the measurement is performed in the LV grid. In general, the amplitude of the non-common emissions between both measurement configurations is considerably higher when the measurement is performed without a LISN. In these cases, the (a) (b) Fig. 6. QP values of the amplitude of the emissions (dBµV) generated by EV5 with a charging current of 8 A under isolated (a) and on-line (b) conditions. Table 3 Frequency (kHz), amplitude (dBµV) of the common emissions recorded under isolated and on-line conditions in the 9–150 kHz frequency range, as well as the percentage of common emissions detected with respect to the total emissions identified under isolated conditions. EV model Charging current On-line emissions / Isolated emissions Percentage of common emissions in both configurations (%) Frequency (kHz) Width at −10 dB (kHz) Amplitude under isolated conditions (dBµV) Amplitude under on-line conditions (dBµV) EV1 16 A 1/11 9.1 130 0.4 69 76 EV2 8 A 2/14 14.3 45 4.3 121 117 90 8.5 91 93 12 A 2/16 12.5 45 4.3 123 118 90 8.3 93 94 16 A 2/6 33.3 45 4.3 80 120 90 8.1 97 96 EV3 8 A 2/13 15.4 44 0.5 125 116 89 0.8 91 92 12 A 1/2 50.0 89 0.8 89 92 16 A 1/5 20.0 89 0.9 86 88 EV4 12 A 1/20 5.0 33 0.7 92 93 16 A 1/10 10.0 33 0.7 93 93 EV5 8 A 3/20 15.0 44 0.5 126 115 89 0.4 91 92 134 0.6 59 71 12 A 1/3 33.3 89 0.4 91 93 16 A 1/5 20.0 44 0.4 95 89 J. Gonz´ alez-Ramos et al. Sustainable Energy, Grids and Networks 38 (2024) 101333 7 background emission of the on-line scenario, in which emissions due to EVCPs are combined with potential NIEs from other sources, brings EV emissions to be undetectable. This behavior can be observed if the median difference is taken into consideration, since, in all the analyzed cases, positive differences ranging from 1 dB to 25 dB are measured. It should be mentioned that, in six out of the twelve cases, a maximum difference between both configurations higher than 20 dB is reported. Finally, if the minimum differences are considered, it can be observed that there are also some cases where emissions with a higher amplitude when measuring using a LISN are registered. 3.1.2.2. 150–500 kHz frequency band. As previously mentioned, in the 150–500 kHz frequency range, flat background with low-amplitude narrowband emissions are recorded when a LISN is used, whereas colored noise is registered if the measurement is conducted in the LV grid. In Fig. 4, it was shown that the emissions in this frequency band when measuring in the LV grid considerably exceed the disturbances recorded if a LISN is used. In order to quantify the existing differences between both measurement conditions in this frequency band, the difference between the QP values of the amplitude of the emissions without and with LISN are calculated (ΔQP(f i ) for 150 <=f i (kHz) <=500). In Fig. 7, a boxplot of these differences for each EVCP under study is shown. In this figure, the red line indicates the median difference value, while the bottom and top edges represent the 25th and 75th percentiles of the differences. The whiskers indicate the most extreme data points not considered outliers. Finally, the outliers are depicted by the red “+” symbol. Fig. 7 shows median values higher than 7.7 dB regardless of the EVCP under test. In seven out of the twelve analyzed EVCPs, the median difference exceeds 20 dB, with a maximum median value of 36.6 dB for EV2 with a charging current of 16 A. The maximum difference, 48.3 dB, is also reported for this EVCP at 150.3 kHz. Therefore, as it was shown in Fig. 4, the amplitude of the emissions measured in the LV grid in the 150–500 kHz frequency band is considerably higher than the ones recorded using a LISN, regardless of the EV model and charging current under test. This implies that the measurements performed under isolated conditions might not be representative of the emissions generated by EVCPs in a real situation in this frequency band. 3.2. Time characterization In order to analyze the time-dependent behavior of the disturbances during the 600 s of recording time, in section III. B. 1), the QP values of the amplitude of the emissions in each period of 50 s are calculated and represented. Then, with the aim of studying the time variability within each period of 50 s, in section III. B. 2), a Fast Fourier Transform (FTT) analysis is performed. 3.2.1. Analysis of the time variability within the recording time (600 s) In Fig. 8, as an example, the QP values of the amplitude of the emissions generated by EV2 with a charging current of 8 A are shown for the twelve 50-second periods (600 s in total) when the measurement is performed using a LISN (a) and in the LV grid (b). Fig. 8(a) shows that the differences in the QP values of the amplitude of the emissions generated by EV2 between consecutive 50-second periods are negligible when a LISN is used. By contrast, in Fig. 8(b), two different states are reported when the recording is performed in the LV grid. Although the spectral features of the disturbances are similar in both states except for the emission occurring around 45 kHz, the amplitude of the NIEs varies considerably in the whole frequency band. It should also be noted that a drastic change in the PFBL can be observed between these two clearly differentiated states. While in the lower-amplitude state the distortions are below the PLC out-of-band emission limits in the majority of the frequency bins, these limits are exceeded in the whole frequency band in the higher-amplitude state. In the specific case shown in Fig. 8(b), the emissions correspond to the higherand lower-amplitude states in five and seven 50-second periods, respectively. With the aim of quantifying the variation of the emissions within the 600 s and evaluating if the time variations shown in Fig. 8 are reported for other EVCPs, the maximum difference between the TSHV and PFBL for the twelve periods of 50 s within 600 s is calculated for each EVCP under study according to (2) and (3) [16]: ΔTSHV =max i∈{1,2,…n}TSHVi−min i∈{1,2,…n}TSHVi(2) ΔPFBL =max i∈{1,2,…n}PFBLi−min i∈{1,2,…n}PFBLi(3) where n is the number of 50-second periods available for each measurement (n=12, except for EV3 with a charging current of 8 A under isolated conditions, for which n=3). Table 4 Number of emissions, minimum, mean, and maximum difference of the emissions identified in the 9–150 kHz frequency band as narrowband emissions under isolated conditions and that are not detected under on-line conditions. EV model Charging current Not detected on-line emissions / Isolated emissions Min. ΔQP (f i ) of the emissions not detected under online conditions⋅ (dB) Median ΔQP(f i ) of the emissions not detected under online conditions⋅ (dB) Max. ΔQP (f i ) of the emissions not detected under online conditions⋅ (dB) EV1 16 A 10/11 4 14 29 EV2 8 A 12/14 13 22 29 12 A 14/16 3 17 30 16 A 4/6 18 25 31 EV3 8 A 11/13 -3 11 18 12 A 1/2 2 2 2 16 A 4/5 4 7 8 EV4 12 A 19/20 -7 9 26 16 A 9/10 -7 7 21 EV5 8 A 17/20 -7 6 18 12 A 2/3 -4 1 6 16 A 4/5 -19 2 11 Fig. 7. Boxplot of the differences of the QP values of the amplitude of the emissions under on-line and isolated conditions (dB) in the 150–500 kHz frequency band. J. Gonz´ alez-Ramos et al. Sustainable Energy, Grids and Networks 38 (2024) 101333 8 The results gathered in Table 5 show that the emissions measured in a LISN barely vary with time in periods greater than 50 s. For this measurement condition, a maximum difference in the TSHV and PFBL of 1 dB and 1%, are reported, respectively, which can be considered negligible. However, as observed in Fig. 8(b) for EV2 with a charging current of 8 A, important variations with time can be observed in four out of the twelve analyzed cases if the measurement is performed directly in the LV grid. In these cases, differences in the TSHV and PFBL between two 50-second periods up to 26 dB and 97% are measured, respectively. Table 5 also shows that these significant time variations mainly occur for EV2 under on-line conditions, regardless of the configured charging current. For this reason, in order to thoroughly analyze the timedependent behavior of the emissions generated by this EV model, in Table 6, the TSHV (dBµV) and PFBL (%) calculated for each 50-second period for EV2 with charging currents of 8 A, 12 A, and 16 A are gathered. For the charging currents of 8 A and 12 A, a drastic decrease in the amplitude of the disturbances is registered in the eighth and sixth 50second period, respectively. In these cases, the TSHV drops from 132 to 133 dBµV and 133–134 dBµV to 119 dBµV, involving a considerable reduction of the PFBL from 100% to 12–13%. It should be noted that, for these charging currents, once the TSHV and PFBL decrease in a certain 50-second period, they do not take higher values at subsequent time instants during the 10 minutes. For the charging current of 16 A, in turn, a different behavior is observed. The amplitude of the emissions varies from one state to another in different periods of 50 seconds, and can increase and decrease depending on the 50-second period considered. For example, in the first and fourth periods, a TSHV and PFBL of 135 dBµV and 100% are measured, while, in the second and third periods, these parameters take values of 109 dBµV and 3%, respectively. Thus, there seems to be different states in the EVCP of EV2 that give rise to differences in the amplitude of the emissions, which means that the amplitude of the NIEs generated by this EVCP rises and falls in different time intervals within 10 minutes. In order to give an insight into the variations of the NIEs from one state to another, as an example, in Fig. 9, the spectrogram of the emissions generated by EV2 with a charging current of 16 A under on-line conditions are depicted during the first 50 seconds of recording. In Fig. 9, sharp changes in the amplitude of the emissions generated by EV2 with a charging current of 16 A when measuring under on-line conditions are shown. This can be clearly observed if the highamplitude emission occurring around 40–50 kHz from 22 seconds onwards is considered, which does not appear in the first 22 seconds of the Fig. 8. QP values of the amplitude of the emissions generated by EV2 with a charging current of 8 A in each period of 50 s during the recording time under isolated (a) and on-line (b) conditions. Table 5 ΔTSHV and ΔPFBL for each EVCP under isolated (Is.) and on-line (O-L) conditions. EV1 EV2 EV3 EV4 EV5 16 A 8 A 12 A 16 A 8 A 12 A 16 A 12 A 16 A 8 A 12 A 16 A Is. O-L Is. O-L Is. O-L Is. O-L Is. O-L Is. O-L Is. O-L Is. O-L Is. O-L Is. O-L Is. O-L Is. O-L ΔTSHV (dB) 1 0 0 14 0 15 0 26 0 0 0 0 0 0 0 0 1 0 1 8 0 0 0 1 ΔPFBL (%) 0 1 1 87 1 88 1 97 0 0 0 1 0 3 0 0 1 2 0 7 0 0 0 2 Table 6 TSHV (dBµV) and PFBL (%) for each 50-second period for EV2 with charging currents of 8 A, 12 A, and 16 A under on-line conditions. Period of 50 s EV2 8 A 12 A 16 A TSHV (dBµV) PFBL (%) TSHV (dBµV) PFBL (%) TSHV (dBµV) PFBL (%) Period 1 132 100 133 100 135 100 Period 2 132 100 134 100 109 3 Period 3 132 100 134 100 109 3 Period 4 132 100 134 100 135 100 Period 5 132 100 134 100 119 12 Period 6 133 100 119 13 119 12 Period 7 133 100 119 12 119 13 Period 8 119 13 119 12 119 12 Period 9 119 13 119 12 119 12 Period 10 119 13 119 13 119 12 Period 11 119 13 119 13 119 12 Period 12 119 13 119 12 119 12 J. Gonz´ alez-Ramos et al. Sustainable Energy, Grids and Networks 38 (2024) 101333 9 recording. This implies that significant time variations can occur abruptly within 50 seconds for this EV model. It should be noted that Fig. 9 also shows that the narrowband emissions around 45–50 kHz and 90–98 kHz correspond to a tonal emission whose central frequency oscillates with time. This oscillation is the main cause of the appearance of narrowband NIE in Fig. 3(a). These emissions have been observed for the three charging currents configured during the charging process of EV2. Therefore, the results presented in this section lead to conclude that for some of the EVCPs under analysis substantial time variations are registered within 10 minutes. These time variations may occur at different time instants within the recording time. This type of time variability was not observed in a previous study where EVCPs corresponding to different EV and EVCS models were evaluated [16], in which only sub-cycle variations were reported. This implies that the NIEs due to EVCPs are very dependent on the specific EV model. 3.2.2. Analysis of the time variability within 50 s The study presented in this section is focused on the common emissions presented in Table 3 (emissions detected both for isolated and on-line conditions). The analysis of the time variability within 50 s is based on the FFT analysis of the signal corresponding to each frequency bin, which gives rise to a model composed of the components around 0 Hz, 50 Hz, 100 Hz, 150 Hz, and 200 Hz. This model accurately resembles the original signal and, thus, it is suitable for characterizing the timedependent behavior of the NIEs generated by EVCPs where a periodic pattern with the fundamental period of 50 Hz is observed [16]. As presented in [16], in order to be able to compare the amplitude of the FFT components of different frequency bins and EVCPs, the amplitudes of the FFT components of each frequency bin are normalized with respect to the amplitude of the component at 0 Hz of this frequency bin. As an example, in Fig. 10, the modulus of the normalized FFT of the time samples corresponding to the emission generated by EV5 with a charging current of 12 A at 89 kHz under isolated and on-line conditions are depicted. Fig. 10 shows that the emissions generated by this EVCP follow the model presented in [16]. This FFT pattern corresponds to a periodic signal with a repetition rate of 50 Hz and, therefore, as presented in [16], the emissions generated by EVCPs vary within the 20 ms of the fundamental period of the mains. This behavior has been observed for all the common emissions analyzed, except for those generated by EV2, which are characterized by tonal emission whose central frequency oscillates with time. For this reason, this EV model will be independently analyzed afterwards. In order to give an overview of the total variation over time of each emission for all the EVCPs (except for EV2), the total variability is calculated as the sum in linear scale of the amplitude of the FFT components at 50 Hz, 100 Hz, 150 Hz, and 200 Hz, which is then converted to logarithmic scale [16]. In Table 7, the total variability of the common emissions detected under isolated and on-line conditions are gathered Fig. 9. Spectrogram of the NIEs generated by EV2 with a charging current of 16 A under on-line conditions considering the first 50 seconds of recording. Fig. 10. Modulus of the normalized FFT (dB) of the time samples corresponding to the emission generated by EV5 with a charging current of 12 A at 89 kHz when measuring under isolated (red) and on-line (black) conditions. J. Gonz´ alez-Ramos et al. 2 of the third LISN, where a PLC device is connected to act as a receiver), considering an opencircuit configuration, i.e., when no external load is connected to the filter. Results and discussion The introduction of the EMC filters in the setup shows variations in the measured impedance that imply variations in signal attenuation. The obtained results lead to the conclusion that the attenuation introduced by the filters greatly depends on the point of connection. Attenuation measured when the filter is connected at points A or B is very low, being practically 0 dB when connecting them in the EUT port of the second LISN. By contrast, the attenuation introduced by the filters when connected at the same electrical port as the receiver is very high on some frequency channels. Therefore, the impact is considerably more noticeable when the filter is connected near the communications equipment, especially if it works as a receiver. It should be taken into account that the communication equipment deployed in the field performs both transmitting and receiving functions. Therefore, in practice, an EMC filter near a communication equipment might imply an additional problem for NB-PLC, especially when receiving frames from other equipment. As the introduction of the filter is only very influential when it is connected close to the receiver, the FER-SNR curves were only represented when the filter was connected at point C. The thresholds obtained using a DBPSK modulation are very similar in the absence of a filter and when each of the four EMC filters are connected. This implies that in a better situation, that is, when using more robust modulations, the FER-SNR curves will neither be affected. As a consequence, it can be concluded that the channel frequency responses due to the filters are not sufficiently selective to degrade NB-PLC in terms of the SNR thresholds. This implies that the estimation and equalization processes defined for PRIME v1.4 perform properly under these circumstances, not being affected by the channel characteristics caused by the introduction of the EMC filters. In order to analyze the considerable attenuation introduced by the EMC filters, the schematics of the four EMC filters were evaluated. The four of them show capacitors in parallel to the LN ports (line side). However, if the filter presented a high-impedance inductive interface, it would decouple its capacitive load from the network, which could reduce the transmission losses. For that reason, we repeated the tests connecting two of the four filters reversed, so that the first elements are inductive. These trials lead to conclude that the impedance measured for the frequency band of interest is not influenced when the filter is connected reversed, obtaining equal results without connecting the filter and when connecting it through the load side to the EUT port of the LISNs. Following the same procedure explained before, the attenuation of the filters in this new configuration was calculated. Regardless of the location of the filter, the attenuation introduced by them when connected to the setup through the L’-N’ ports can be considered negligible. Thus, it is demonstrated that the attenuation introduced by the EMC filters is mainly due to the capacitive burden they present in the network interface. Conclusion As a conclusion, it is shown that the EMC filters used in the study do not introduce abrupt variations in the channel frequency response, so that thresholds that set the quality of communications are not affected. However, they do present considerable signal attenuations when they are connected near the receiver equipment. These attenuations could cause communications to fail if the noise level is high or if the received signal power is close to the sensitivity limit of the receiving equipment. This implies that the EMC filters used for power converters have an unintended side effect of attenuating communication signals that had not been analyzed before. These high attenuations are due to the capacitive interface of the filters with the grid. 255 B.1.5. International Conference Paper ICP5 This subsection presents the following conference paper:  J. González-Ramos, B. Grasel, I. Angulo, I. Fernández, A. Arrinda, "Evaluation of PRIME v1.4 in a Reconstructed Low Voltage Grid," 14th Workshop for Powerline Communications (WSPLC), Mannheim, Germany, 2023. Then, the most representative quality indicators concerning this paper are listed below:  Publisher: -  Conference: 14th Workshop for Powerline Communications (WSPLC)  Year of presentation: 2023  Type of publication: Conference Proceedings  Area: Electrical & Electronic Engineering XXX-X-XXXX-XXXX-X/XX/$XX.00 ©20XX IEEE Evaluation of PRIME v1.4 in a Reconstructed Low Voltage Grid Jon González-Ramos*, Bernhard Grasel†, Itziar Angulo*, Igor Fernández*, Amaia Arrinda* University of the Basque Country (UPV/EHU), Bilbao, Spain* University of Applied Sciences Technikum Vienna, Vienna, Austria† {jon.gonzalezr, itziar.angulo, igor.fernandez, amaia.arrinda }@ehu.eus* [email protected]† Keywords— Narrowband Power Line Communications, PRIME v1.4, Reconstructed Low Voltage Grid. I. INTRODUCTION Power Line Communications (PLC) are the most extensively deployed technologies by the Distribution System Operators (DSOs) to transmit data over the electrical grid [1]. In the recent development of Advanced Metering Infrastructures (AMI), Narrowband PLC (NB-PLC) have been the preferred option to transmit data between Smart Meters (SMs) and data concentrators. However, the efficient performance of PLC is challenging, as the electrical grid was not designed for data communication. Until now, only frequencies below 500 kHz have been considered, as they are expected to allow longer transmission distances. Nevertheless, the high level of interfering noise and disturbances that are present in this range, together with the transmission losses due to the changing impedance of the medium, hamper the communications and prevent from obtaining full-time availability of the transmission devices. In this context, this paper aims at evaluating the performance of NB-PLC according to PRIME v1.4 standard in a reconstructed Low Voltage (LV) grid considering the connection of multiple electronic devices that may have a negative impact on communications. Moreover, a characterization of the emissions at the measurement locations is also presented, in order to relate their amplitude and spectral shape to the potential degradation of NB-PLC. II. METHODOLOGY The measurements were performed in the reconstructed LV grid at University of Applied Sciences Technikum Vienna shown in Fig. 1. This measurement scenario is composed of a Secondary Substation (SS) and four houses (H1, H2, H3, and H4), to which different electronic devices can be connected [2]. The distance between the SS and each house is 17 m. All the measurements were carried out at phase 2 (L2) considering a star grid topology. The connected devices are as follows: 2 SMs (L123) at H1, 3 SMs (L123) and a Photovoltaic (PV) system (L2) at H2, 2 SMs (L123), a PV system (L2) and a 1200 W load (L123) at H3, and 3 SMs (L123), a PV system (L123), and a storage system (L123) at H4. Fig. 1. Reconstructed LV grid at University of Applied Sciences Technikum Vienna. NB-PLC according to PRIME were evaluated by means of two PL360G55CF-EK boards, one acting as a transmitter and connected to the SS and the other operating as a receiver and connected to the corresponding house. The study considers the frequency channels 1 and 3-8 defined in PRIME v1.4 standard [3], as well as DBPSK_C, DQPSK_C, R_DQPSK, and R_DBPSK modulations. In all the trials, 1000 frames with a 256-byte length were transmitted. The recording of the emissions was conducted using the measurement system developed by the current authors presented in [4]. III. RESULTS AND CONCLUSIONS In order to evaluate the performance of PRIME v1.4, for each configuration, the mean value of the Signal to Noise Ratio (SNR) of all the received frames is calculated, in addition to the Frame Error Rate (FER), which is defined as the ratio between the erroneous received frames and the total number of transmitted frames. In Table I and Table II, the mean SNR and FER obtained for each configuration when the receiver is located at H1, H2, H3, and H are gathered. Table I. Mean SNR and FER obtained for each frequency channel and modulation when the receiver is located at H1 (a) and H2 (b). Modulation Channel Mean SNR FER DBPSK_C CH1 5.8 1.000 CH3 20.3 0.000 CH4 13.0 0.000 CH5 15.0 0.002 CH6 20.5 0.000 CH7 20.8 0.000 CH8 19.8 0.000 DQPSK_C CH1 7.5 1.000 CH3 18.3 0.000 CH4 11.3 0.006 CH5 14.8 0.000 CH6 21.8 0.000 CH7 22.0 0.000 CH8 20.5 0.000 R_DBPSK CH1 5.8 0.000 CH3 17.5 0.000 CH4 9.8 0.000 CH5 14.8 0.000 CH6 21.0 0.000 CH7 21.0 0.000 CH8 19.0 0.000 R_DQPSK CH1 7.3 0.213 CH3 18.8 0.000 CH4 11.5 0.000 CH5 16.0 0.000 CH6 21.5 0.000 CH7 21.3 0.000 CH8 19.3 0.000 Modulation Channel Mean SNR FER DBPSK_C CH1 11.1 0.002 CH3 10.1 0.231 CH4 5.6 0.861 CH5 13.0 0.000 CH6 15.9 0.000 CH7 19.8 0.000 CH8 19.0 0.000 DQPSK_C CH1 11.6 0.026 CH3 11.9 0.825 CH4 7.2 0.932 CH5 13.6 0.000 CH6 16.3 0.000 CH7 20.6 0.000 CH8 19.6 0.000 R_DBPSK CH1 11.3 0.000 CH3 10.0 0.000 CH4 5.7 0.000 CH5 13.5 0.000 CH6 16.1 0.000 CH7 19.9 0.000 CH8 19.1 0.000 R_DQPSK CH1 11.7 0.000 CH3 11.4 0.000 CH4 6.9 0.264 CH5 13.6 0.000 CH6 16.2 0.000 CH7 20.5 0.000 CH8 19.7 0.000 Table II. Mean SNR and FER obtained for each frequency channel and modulation when the receiver is located at H3 (a) and H4 (b). Modulation Channel Mean SNR FER DBPSK_C CH1 8.4 0.029 CH3 12.5 0.010 CH4 10.9 0.000 CH5 18.5 0.000 CH6 21.0 0.000 CH7 21.3 0.000 CH8 21.6 0.000 DQPSK_C CH1 9.1 0.936 CH3 14.4 0.085 CH4 11.1 0.000 CH5 19.1 0.000 CH6 22.5 0.000 CH7 22.6 0.000 CH8 23.1 0.000 R_DBPSK CH1 6.6 0.000 CH3 9.4 0.000 CH4 7.6 0.000 CH5 10.7 0.000 CH6 18.1 0.000 CH7 19.8 0.000 CH8 19.5 0.000 R_DQPSK CH1 7.6 0.000 CH3 10.3 0.000 CH4 8.3 0.000 CH5 10.4 0.000 CH6 18.4 0.000 CH7 19.8 0.000 CH8 19.2 0.000 Modulation Channel Mean SNR FER DBPSK_C CH1 4.0 1.000 CH3 12.0 0.001 CH4 4.2 1.000 CH5 5.5 0.999 CH6 15.2 0.000 CH7 18.5 0.000 CH8 16.2 0.000 DQPSK_C CH1 5.9 1.000 CH3 12.8 0.010 CH4 6.1 1.000 CH5 7.6 1.000 CH6 15.8 0.000 CH7 19.3 0.000 CH8 17.1 0.000 R_DBPSK CH1 4.1 0.846 CH3 12.2 0.000 CH4 4.3 0.707 CH5 6.1 0.000 CH6 14.6 0.000 CH7 19.2 0.000 CH8 17.2 0.000 R_DQPSK CH1 6.0 1.000 CH3 14.4 0.000 CH4 6.8 0.896 CH5 7.8 0.018 CH6 16.4 0.000 CH7 19.7 0.000 CH8 19.9 0.000 The results shown in Table I lead to conclude that channel 1 (42-89 kHz) is significantly more hostile to communications. This behavior can be clearly observed, for example, when analyzing the results obtained for the DQPSK_C modulation. While in channel 1 all frames are received with errors, in the upper frequency channels (3-8) only a FER higher than 0 is reported in channel 5 (0.02). The degradation of communications in channel 1 is mainly caused by the emissions of higher amplitude reported in this frequency band (see Fig. 2). It should be noted that, in Europe, this is the frequency range assigned by CENELEC for smart metering applications and, thus, an improper performance of the system could be expected. However, in some cases, as shown in Table II at location 4, a same FER (1.000) is obtained in channel 1 and higher frequency channels (for instance, channel 4 for DBPSK_C modulation), although the amplitude of the emissions is considerably higher in the 42-89 kHz frequency band. For this reason, in these instances, in order to determine the cause of the degradation of the operation of NB-PLC according to PRIME v1.4, further study is needed. This analysis should consider the influence of the grid impedance, both at the transmitter and receiver sides, and the channel frequency response between the SS and the corresponding location. Finally, it should be mentioned that, as it could be expected, modulations including repetition codes, i.e., R_DQPSK and R_DBPSK, perform considerably better than those using only Forward Error Correction (FEC) codes (DBPSK_C and DQPSK_C). In the case of R_DBPSK, all the frames are received correctly regardless of the frequency channel in which the PLC signal is transmitted. Fig. 2. QP values of the amplitude of the emissions measured at H1, H2, H3, and H4. ACKNOWLEDGMENT This work was financially supported in part by the Basque Government under the grants IT436-22, PRE_2021_1_0006 and PRE_2021_1_0051, and by the Spanish Government under the grants PID2021 124706OB-I00 (MCIU/AEI/FEDER, UE, funded by MCIN/AEI/10.13039/5011000011033 and by “ERDF A way of making Europe”). REFERENCES [1] L. Lampe, A. Tonello, and T. Swart, Power line communications: Principles, standards and applications from multimedia to smart grid: Second edition. 2016. doi: 10.1002/9781118676684. [2] B. Grasel, J. Baptista, and M. Tragner, “Supraharmonic and Harmonic Emissions of a Bi-Directional V2G Electric Vehicle Charging Station and Their Impact to the Grid Impedance,” Energies (Basel), vol. 15, no. 8, 2022, doi: 10.3390/en15082920. [3] “PRIME Alliance, ‘PRIME v1.4 White Paper’, PRIME Alliance,” Brussels, 2014. [4] I. Fernández, D. de la Vega, A. Arrinda, I. Angulo, N. Uribe-Pérez, and A. Llano, “Field Trials for the Characterization of Non-Intentional Emissions at Low-Voltage Grid in the Frequency Range Assigned to NB-PLC Technologies,” Electronics (Basel), vol. 8, no. 9, 2019, doi: 10.3390/electronics8091044. [Document text truncated for crawler view.]