Experimental Evaluation and Coordinated Deployment of Ramp-Rate Limitation Against Rapid Voltage Changes in Distribution Systems

IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 61, NO. 2, MARCH/APRIL 2025 1839 Experimental Evaluation and Coordinated Deployment of Ramp-Rate Limitation Against Rapid Voltage Changes in Distribution Systems Stelios C. Dimoulias , Student Member, IEEE, Kyriaki-Nefeli D. Malamaki , Member, IEEE, Andrei Mihai Gross , Francisco de Paula García-López , Georgios C. Kryonidis , Senior Member, IEEE, and Manuel Barragán-Villarejo Abstract—Climate change has expedited the integration of converter-interfaced renewable energy sources (CIRESs) in distribution systems. Nevertheless, the active power volatility of CIRESs and the resulting high ramp rates are known to downgrade the voltage quality of distribution systems, causing rapid voltage changes (RVCs). In this context, the operation of CIRESs under ramp-rate limitation (RRL) schemes, employing an energy storage system, could serve as a preventive action against RVCs, thereby restoring voltage quality. To this end, this paper provides insights into the deployment of RRL control against RVCs in two stages. Firstly, the aptness of RRL as a preventive action is experimentally evaluated in a scaled-down laboratory testbed of the CIGRE European benchmark medium-voltage feeder, hosting CIRES prototypes. The RVCs are defined as per the IEEE 1547:2018 and IEC 610004-30:2015 Standards. The experimental results demonstrate that the operation of CIRESs under RRL control can enhance voltage quality by suppressing RVCs. Subsequently, a novel, system-level strategy for the coordinated mitigation of RVCs is developed. For a given distribution system, the proposed strategy allocates the RRL functionalities among CIRES units, to achieve the elimination of RVCs under minimal CIRES engagement and energy storage requirements. The effectiveness of the strategy is tested via dynamic rms simulations on the IEEE European Low Voltage test feeder. Index Terms—Active distribution networks, energy storage systems, ramp-rate limitation, rapid voltage changes, system planning, voltage quality. Received 2 July 2024; revised 17 September 2024; accepted 24 October 2024. Date of publication 27 December 2024; date of current version 4 April 2025. Paper 2024-ESC-1007.R1, presented at the 2023 International Conference on Smart Energy Systems and Technologies (SEST), Mugla, Türkiye, Sep. 04–06, and approved for publication in the IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS by the Energy Systems Committee of the IEEE Industry Applications Society [DOI: 10.1109/SEST57387.2023.10257447]. This work was supported by the European Union under the Horizon European project COCOON under Grant 101120221. (Corresponding author: Kyriaki-Nefeli D. Malamaki.) SteliosC.Dimoulias,Kyriaki-NefeliD.Malamaki,andGeorgiosC.Kryonidis are with the Department of Electrical and Computer Engineering, Aristotle University of Thessaloniki, 54124 Thessaloniki, Greece (e-mail: [email protected]; [email protected]; [email protected]). AndreiMihai Gross,Francisco dePaula García-López,and ManuelBarragánVillarejo are with the Department of Electrical Engineering, Universidad de Sevilla, 41004 Seville, Spain (e-mail: [email protected]; [email protected]; manuel- [email protected]). Color versions of one or more figures in this article are available at https://doi.org/10.1109/TIA.2024.3523456. Digital Object Identifier 10.1109/TIA.2024.3523456 I. INTRODUCTION AMID the ongoing energy transition, new challenges are introduced to electricity grids, particularly due to the proliferation of converter-interfaced renewable energy sources (CIRESs). Regarding distribution systems, the volatile generationof CIRESscauses technicalproblems, suchas reversepower flows and overvoltages [1]. In this context, a phenomenon that has been highlighted by technical standards [2],[3],[4] and reports from regulatory authorities [5], but remains relatively overlooked by the technical literature, is that of rapid voltage changes (RVCs). Defined in the IEEE 1547:2018 [3] and the IEC 61000-4-30:2015 [4] Standards, RVCs can emerge due to high ramp rates of CIRES active power and have been identified to degrade the performance of electrical equipment and produce flicker, thus compromising voltage quality [6],[7]. As a proactiveresponse,CIRESscanoperate,inconjunctionwithanenergy storage system (ESS), under power smoothing schemes, so as to preventhigh ramprates,andhence,RVCs[8],[9].Amongpower smoothing approaches, ramp rate limitation (RRL) techniques are preferred due to their superior performance in terms of the achieved smoothing effect and the more efficient ESS usage, as attested by recent review [10],[11] and comparative [12],[13], [14],[15],[16] studies. More specifically, both the simulationbased [10],[11],[12],[15] and the experimental [13],[14],[16] studies have demonstrated that RRL techniques outperform alternative approaches, such as moving-average or low-pass filter methods, in terms of the required ESS sizing and operational cycles, while also achieving fewer ramp rate limit violations. Despite the extensive literature on RRL control, only a few works examine the potential mitigating impact of RRL schemes on voltage variations, and even fewer focus on RVCs. More specifically, in [17], RRL control is applied to alleviate voltage fluctuations. However, the study neglects the RVC metric, considering the 1-minute flicker effect as a performance index, and it primarily focuses on techno-economic aspects of the employed ESS. The RVC metric is also neglected in [18], where a voltage-smoothing algorithm, combining RRL with reactive power control, is proposed. Moreover, the study focuses on the effect of reactive power control, while the therein analysis does not quantify the achieved voltage-smoothing effect. In [8], the authors investigate the effect of RRL against rapid voltage © 2024 The Authors. This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 License. For more information, see https://creativecommons.org/licenses/by-nc-nd/4.0/ 1840 IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 61, NO. 2, MARCH/APRIL 2025 variations caused by cloud passing in photovoltaic (PV) plants, and their study is extended to wind turbines in [9]. Nonetheless, these works do not consult the definitions of the relevant Standards [3],[4]. Additionally, and most importantly, the above studies are limited to demonstrating and quantifying the effect of RRL against voltage fluctuations, assuming uniform application of theRRL controlbyall CIRESunits of adistributionsystem. This approach, although effective, is characterized by excessive capitalexpenditure,sinceit implies thateveryCIRESunit is required to be equipped with an ESS and be upgraded with RRL control functionalities. Therefore, in the context of RVC mitigation, the technical literature lacks a strategy for the coordinated and efficient allocation of the RRL functionalities among the CIRES units of a distribution system, considering factors such as the distribution system topology and techno-economic indices, e.g., ESS size and cost. From the point of view of the industry, this impedes distribution system operators (DSOs) from addressing the issue of RVCs in practice, even though its importance is highlighted in the technical standards and literature, and some European countries, e.g., Norway, have officially imposed limits on the allowable number of RVCs per day [5]. Regarding the experimental implementation of RRL, the relevant works have been limited to the experimental validation and comparative evaluation between novel and conventional RRL techniques at CIRES level. In this context, the current literature has focused on the power smoothing effectiveness of the tested techniques and the technical requirements of the employed ESSs. More specifically, in [19], a novel RRL scheme based on the predictor-corrector logic is experimentally validated, while in [20], an alternative method, combining supercapacitors and fuel cells, is tested. Similarly, the authors in [21] propose and experimentally evaluate the performance of a new, filter-based technique that employs supercapacitors. Moreover, in [14],several RRL control schemes are experimentally compared in a microgrid environment. Furthermore, although the correlation between CIRES active power volatility and voltage fluctuations is experimentally verified in [9], the authors do not extend their experimental analysis to the positive impact of RRL on the mitigation of those fluctuations. Consequently, a comprehensive experimental evaluation of the potential mitigating effect of RRL against RVCs in distribution systems is missing from the technical literature. Having identified these existing research gaps, the scope of this paper is two-fold: (a) to experimentally demonstrate the aptness of RRL schemes for the mitigation of RVCs in active distribution systems; (b) to propose a method for the efficient allocation of the RRL control among the distributed CIRES, aiming for the elimination of RVCs under minimal requirements in CIRES engagement and energy storage. An initial experimental stage of the work was originally presented in [1]. Building upon [1], this work enriches the experimental evaluation with additional results and introduces the distribution system-level coordination strategy for RVC mitigation. Concerning the experimental stage, the RRL scheme of [16] is incorporated in three CIRES laboratory prototypes within a scaled-down version of the CIGRE European benchmark medium-voltage (MV) feeder, presented in [22]. Each CIRES is equipped with a supercapacitor, which serves as the ESS to perform the RRL control. To achieve more comprehensive results, two distinct power profiles are used to emulate the CIRES volatility. The experimental results are assessed according to the guidelines of [3] and [4]. The second stage of the paper regards the development of a novel, universal strategy for the coordinated deployment of RRL control against RVCs, designed for distribution system operators. Specifically, by mathematizing the problem of RVC prevention, the proposed method determines which CIRES units need to perform RRL in order to suppress RVCs with the minimum number of RRL-performing CIRESs and with the minimumtotal installedESS capacity. Additionally,therequired degree of RRL per CIRES unit is specified. The performance of the method is assessed via rms simulations on the IEEE European Low Voltage (LV) test feeder. In summary, the paper introduces the following contributions with respect to the current state-of-the-art: rExperimentally evaluates the impact of RRL against RVCs in active distribution systems. rIntroduces a new, system-planning method to efficiently preventRVCsinactivedistributionsystems,underminimal CIRES engagement and energy storage requirements. The rest of the manuscript is organized as follows: Section II provides an overview of the IEC and IEEE definitions of RVCs. Section III briefly presents the considered RRL technique. In Section IV, the experimental evaluation of the RRL control is discussed, with a thorough presentation of the laboratory setup and the performed tests. Section Vpresents the new method for the coordinated deployment of RRL against RVCs. The method is validated in Section VI. Finally, Section VII concludes the paper, discussing its main findings. II. REVIEW OF RVC DEFINITIONS In this section, the RVC definitions of the IEC 61000-430:2015 [4] and IEEE 1547:2018 Standards [3] are presented. A. IEC 61000-4-30 Standard In IEC 61000-4-30:2015 [4]-Class A, RVCs are defined over a reference interval, Δtref, with a duration of 1-s. The definition is based on the Urms(1/2) quantity, i.e., the rms voltage value, measuredoveronecycleandrefreshedateachhalf-cycle.Hence, Δtref accounts for Mcalculated values of Urms(1/2), where Mis 100 or 120, for 50 or 60 Hz systems, respectively. In each halfcycle,themonitoredvoltageisinsteady-state,ifall Mpreceding values and the new Urms(1/2) value of this half-cycle lie within a threshold, calculated as a percentage, th(%), of their arithmetic mean, Umean rms(1/2). Thus, an RVC starts if at least one of the M+1 values of Urms(1/2) exceeds the threshold. Note that, during an RVC event, a hysteresis is applied to the th(%). Percentages between 1%–6% of Urms(1/2) for th; and around 50% of th,for the hysteresis, are suggested. Therefore, both parameters are user-defined. Furthermore, the Standard quantifies RVCs through the following characteristic parameters: magnitude (ΔUss), maximum DIMOULIAS et al.: EXPERIMENTAL EVALUATION AND COORDINATED DEPLOYMENT OF RRL AGAINST RVC IN DISTRIBUTION SYSTEMS 1841 Fig. 1. Indicative RVC event according to [4]. deviation (ΔUmax), and duration (ΔT). ΔUss is the absolute difference between the final Umean rms(1/2) value prior to the event and the first Umean rms(1/2) value after the event. ΔUmax denotes the maximum absolute difference between any of the Urms(1/2) values, during the event, and the final Umean rms(1/2) value, prior to it. Essentially, ΔUss and ΔUmax express the intensity of an RVC. ΔTis defined as ΔT=tr−ts−Δtref, where tsand trare the commencing and the steady-state restoration instants of the event, respectively. To better comprehend the detection process, the reader is referred to Fig. 1, where an indicative RVC event is identified. Here, th =1%, while UB,LB stand for the upper and lower boundary of Urms(1/2), respectively. B. IEEE 1547-2018 Standard The IEEE 1547:2018 Standard [3] specifies that an individual CIRES, at MV level, should not cause step or ramp changes in the rms voltage at its point of connection exceeding 3% of nominal and 3%/s, averaged over a period of 1-s. The respective limit is 5% for the LV level. Nonetheless, when referring to the combined effect of all the RVC-inducing sources, the stricter limit of 3% is used for both voltage levels. Concerning the detection process, no guidelines are provided. III. CONSIDERED RRL CONTROL Generally, RRL control approaches aim to contain the output ramp rate of a CIRES unit, RRout, within a desired, pre-defined limit, denoted as RRlim. More specifically, the input active power of the unit, pin[t], is measured, and the corresponding ramp rate, denoted as RRin[t], is compared against RRlim.If |RRin[t]|<RR lim, the ESS does not intervene; otherwise, it injects/absorbs the required power so that |RRout[t]|=RRlim. In this study, the direct RRL control of [16] is considered, bothforthe experimentaland the simulation-basedanalysis.The employed RRL control has been compared against conventional power smoothing techniques, such as the moving average and low pass filter, both through simulations [15] and experimentally [16], demonstrating superior performance in terms of ESS efficiency and ramp rate limit violations. The control logic behind the considered RRL scheme is depicted in Fig. 2and presented in algorithmic form in Algorithm 1. The inputs of the algorithm are pin[t], and the positive (max[t]=RRlim) and negative (min[t]=-RRlim) limits of RRout, for ramp-ups and ramp-downs, respectively. Δtrefers to the measurement resolution of pin[t]. The output of the control block is the smoothed, Fig. 2. Control logic of the considered RRL technique [16]. Algorithm 1: Algorithm for RRL Control. Require: pin[t],pout[t−Δt],min[t],max[t],Δt Ensure: pout[t],pESS[t] 1: RRin[t]←pin[t]−pout[t−Δt] Δt 2: if RRin[t]<min[t]then 3: RRout[t]←min[t] 4: else if RRin[t]>max[t]then 5: RRout[t]←max[t] 6: else 7: RRout[t]←RRin[t] 8: end if 9: pout[t]←RRout[t]·Δt+pout[t−Δt] 10: pESS[t]←pin[t]−pout[t] output power pout[t], by which the reference power, pESS[t],of the ESS is determined. Among pin[t],pout[t], and pESS[t], it holds that:pESS[t]=pin[t]−pout[t]. Moredetailsregardingthecontrol are available in [16]. IV. EXPERIMENTAL EVALUATION OF RRL AGAINST RVCS In this section, RRL is experimentally evaluated as a preventive action against RVCs. Initially, the laboratory setup with the employed CIRES prototypes is presented. Subsequently, the experimental scenarios are thoroughly assessed according to the IEEE and IEC Standards. A. CIRES Prototypes The RRL algorithm of Section III is incorporated in three laboratory voltage source converter prototypes (VSCPs). More specifically, each VSCP setup features a 20-kVA, three-phase, three-wire voltage source converter (dc/ac converter), with rated dc and ac voltages of 750 V and 400 V, respectively. Regarding the employed ESS, supercapacitors with a capacitance of 6 F, a rated voltage of 160 V, and a maximum instantaneous power of 2 kW, are selected. The supercapacitors are coupled at the dc link of the VSCP via a dedicated dc/dc converter (Fig. 3). Note that, although the theoretical energy storage capacity of the supercapacitors is 21.33 Wh, the actual available energy for RRL control is restricted to 14.58 Wh, due to limitations of the dedicated dc/dc converters to operate with a minimum voltage of 90 V. The primary energy source of the VSCP is emulated by a controllable current source. The control of the prototype has been implemented in a Texas Instruments TMS320F28335 1842 IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 61, NO. 2, MARCH/APRIL 2025 Fig. 3. Configuration of the supercapacitor (SC) and dc/dc converter. Fig. 4. Single-line diagram of the scaled-down MV distribution system used for validation purposes. Delfino microcontroller with a sampling period of 50 μs. Details on the modeling and control of the VSCPs can be found in Appendix A,aswellasin[23]. B. Experimental Testbed The experimental testbed is a scaled-down version of the CIGRE European benchmark MV feeder, presented in [22] and set up in the laboratory of Universidad de Sevilla. The original distribution system consists of two radial feeders, with lengths of 15 km (Feeder 1) and 8 km (Feeder 2), respectively. In the present study, only Feeder 1 of the original grid has been emulated, with its respective single-line diagram presented in Fig. 4. This subsystem comprises 11 buses, hosting two constant-power loads and the three VSCPs of Section IV-A. The scaled-down distribution system has been designed by converting the base magnitudes of the original grid: from 20 kV and 10 MVA, to 400 V and 100 kVA, respectively; allowing for the reproduction of the behavior of the original distribution system in terms of voltage drops, power flows, and losses, in an LV laboratory. Furthermore, a Regatron controllable voltage source is connected at Bus 0 to decouple the scaled-down experimental distribution system from the LV supply grid of the laboratory. More specifically, the Regatron controllable voltage source aims to maintain a constant voltage at Bus 0, emulating a stiff, upstream system. Fig. 5(a) shows the branches of the two feeders and the cabinets with the devices used to emulate loads and CIRESs. The interior of such a cabinet is shown in Fig. 5(b). Details on the parameters of the scaled-down MV distribution system and the architecture for the real-time control of the testbed are provided in [22]. The line and transformer parameters of the testbed are found in Section A of Appendix A, at the end of the manuscript. Fig. 5. Laboratory scaled-down MV distribution system: (a) Distribution system overview, (b) Cabinet interior. Fig. 6. Active (a) and reactive (b) power profiles of the loads at Bus 5 and Bus 8 for Scenario E2. C. Experimental Scenarios To experimentally evaluate the performance of RRL in terms of RVC mitigation, two different scenarios are considered, each corresponding to different power profiles for the VSCPs and different consumption patterns for the loads. In Scenario E1, an artificial step profile with a duration of 300-s and a maximum power of 7.5 kW is used to emulate the volatility of the CIRES primary energy source. The profile is assigned as input to the VSCPs. The loads are controlled to absorb 7 kW at unity power factor. In Scenario E2, the VSCPs are assigned a real, 300-s profile with a peak power of 5 kW. This profile corresponds to the 300-s interval that exhibits the most severe power variations in the PV measurements of [16]. Moreover, in this scenario, the loads are assigned the active power consumption profiles of Fig. 6(a). These profiles derive from the daily profiles designated by CIGRE Task Force C06.04.02 [24] for distribution system loads, and have been re-scaled to a 300-s interval. Similar profiles are considered for the reactive power consumption, as shown in Fig. 6(b). DIMOULIAS et al.: EXPERIMENTAL EVALUATION AND COORDINATED DEPLOYMENT OF RRL AGAINST RVC IN DISTRIBUTION SYSTEMS 1843 Fig. 7. Output active power profile of VSCP2 for Cases 1-3: (a) Scenario E1; (b) Scenario E2. For both scenarios, three different cases are examined: no RRL control applied to the VSCPs (Case 1), the RRL scheme of [16] with RRlim=0.01 pu/s to all VSCPs (Case 2), and the RRLschemeof[16] with RRlim=0.005 pu/s to all VSCPs (Case 3). Case 1 is the base case that highlights the RVC-inducing effect of CIRES volatility. The RRlim values of Cases 2-3 are expressedinperunit(pu) withrespect tothe ratingofeachVSCP and are relatively relaxed, corresponding to 60 %/min and 30 %/min, respectively, thus being significantly less strict than the commonly suggested limit of 10 %/min [17],[19],[25]. These lax RRlim values are selected to prevent the excessive stressing of the supercapacitor and its dc/dc converter. The achieved smoothing effect is indicated in Fig. 7(a) and (b), by the output active power of VSCP2, for Scenarios E1 and E2, respectively. The RVC-suppressing capability of the implemented RRL control is assessed according to the IEEE 1547:2018 and IEC 61000-4-30:2015-Class A Standards. The two Standards differ regarding the required measuring sampling rate. More specifically, provided that the IEC Standard is based on the Urms(1/2) concept, measurements with a sampling period of up to 10ms are required. Conversely, the required sampling period is undefined in IEEE 1547:2018. The only specification concerns the examination of the rate of change of voltage (RoCoV) over an 1-s period. Therefore, any sampling period of up to 0.5s would suffice. Due to the different resolutions of the voltage meters connectedateachbusofthetestbed,thefollowingmeasurements are available: (a)rms measurements, with a resolution of 0.5s, for Buses 2-11; (b)instantaneous measurements, sampled every 50 μs, for Bus 1. Consequently, the RRL effect is assessed according to IEEE 1547:2018 for Buses 2-11, and as per the IEC Standard for Bus 1. D. Results According to IEEE 1547:2018 Results are presented for Buses 5, 6, 8, 9, and 11, where the loads and VSCPs are connected, as depicted in Fig. 4. Fig. 8. Scenario E1: Voltage profiles with respect to time: (a) Bus 8 - Case 1; (b) Bus 8 - Case 3; (c) Bus 6 - Case 1; (d) Bus 6 - Case 3. TABLE I SCENARIO E1: RVCSATLOAD BUSES ACCORDING TO THE IEEE STANDARD TABLE II SCENARIO E1: RVCSATVSCP BUSES ACCORDING TO THE IEEE STANDARD 1) Scenario E1: The capability of the RRL algorithm to suppressRVCphenomenaisvisualizedinFig.8,withthevoltage profiles of Buses 6 and 8 for Cases 1 and 3. By comparing Fig. 8(a) and (c) with Fig. 8(b) and (d), respectively, it is attested that the RRL application suppresses the detected RVC events of Case 1, significantly improving the smoothness of the voltage profiles. Moreover, as shown in Fig. 8, the VSCP bus (Bus 6) exhibits more frequent and significant voltage variations compared to the load bus (Bus 8). This is due to the higher total impedance in the path between Bus 0 and Bus 6. Specifically, as indicated by Tables IX and X, the total impedance in the path between Bus 0 and Bus 6 is 281.1+j262.9 mΩ, whereas the total impedance between Bus 0 and Bus 8 is only 211.1+j246.9 mΩ. This higher total impedance, and more importantly, the significantly higher total resistance in the path to Bus 6, renders the voltage at Bus 6 more sensitive to active power variations [26], resulting in more severe voltage fluctuations. The overall effect of the RRL control is aggregated for all buses and cases in Tables Iand II. The impact is quantified through the total number of RVC events, N, and the maximum 1844 IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 61, NO. 2, MARCH/APRIL 2025 Fig. 9. Scenario E2: Voltage at Bus 6. Fig. 10. Scenario E2: Number of RVC events per VSCP bus and case. The maximum RoCoV (%/s) for each bus and case is shown above the bars. RoCoV, dV dt(%/s), for these events. As shown in Table I,for Cases 2-3, the application of RRL, under both RRlim values, eliminates all RVC events at load buses. Regarding VSCP buses, the results of Table II showcase that the application of RRL with RRlim =0.01 pu/s (Case 2) achieves up to a 60% reduction in the total number of events and up to a 40% reduction in the maximum RoCoV, compared to Case 1. Eventually, the use of RRlim =0.005 pu/s (Case 3) completely suppresses RVCs, restoring voltage quality. 2) Scenario E2: For Scenario E2, the RVC-suppressing effect of the employed RRL control is showcased in Fig. 9, where the voltage of Bus 6 for Case 1 and Case 3 is depicted. As shown, the RRL algorithm effectively smooths the voltage, mitigating its fluctuations and abrupt variations. The overall performance of the RRL algorithm is demonstrated in Fig. 10, by the total number of induced RVCs at VSCP buses and the maximum RoCoV (%/s) per case. As observed, the application of RRL with RRlim =0.01 pu/s (Case 2) significantly improves voltage quality.Eventually,forCase3,RVCsarecompletelysuppressed, and the maximum RoCoV values (%/s) are contained below 1 %/s. Regarding load buses, no RVCs were reported. The tests of Scenario E1 and Scenario E2 have experimentally validated the effectiveness of RRL control as an RVC-mitigation measure under diverse conditions of CIRES power volatilityand load consumption. Comparatively, the conditions of Scenario E2 resulted in fewer and less severe RVC events than those of Scenario E1. This is attributed to the more gradual power variations of the VSCPs and the influence of the loads, whose active and reactive power variations partially balance the CIRES volatility. E. Results According to IEC 61000-4-30:2015-Class a The RVC-suppressing capability of the RRL control is further tested under the IEC definition, examining the changes in the Fig. 11. RVC event at Bus 1: (a) Case 1 - th =2%; (b) Case 1 - th =1%; (c) Case 3 - th =1%. In (a) and (b), red dots indicate the starting and ending instants of RVC events. voltage of Bus 1 for Scenario E1. Note that the conditions of Scenario E1 have been selected, as they correspond to more severe power and voltage variations. Prior to discussing the performance of the RRL control, it is important to highlight the strong dependency of the detection process and the RVC duration on the user-defined threshold, th(%). This dependency is observed in Fig. 11, where a certain RVC event is depicted. Specifically, Fig. 11(a) and (b) represent the same detected event, but for th =2% and th =1%, respectively. In the latter case, due to the stricter threshold, the RVC event is triggered earlier, resulting in a longer duration ΔT. For the sake of better comparison, the start time, ts=116.55 s, of the event for th =2%, is also stamped on Fig. 11(b). Regarding the effect of RRL, applying the considered technique to all VSCPs with RRlim =0.005 pu/s suppresses the event, containing voltage between the upper (UB) and lower boundary (LB). The overall results for the 300-s analysis period and the three distinct cases are summarized in Table III, which presents the total number of events and the maximum reported values of the characteristic RVC parameters. Regarding Case 1, the results highlight the effect of th on the number of RVCs, as Nis almost doubled if th is set to 1%. Furthermore, the RRL scheme effectivelymitigatesRVCs, reducingtheir number,duration,and intensity. More specifically, for Case 2, the number of events is reduced by 75%, compared to Case 1, for both th values. For DIMOULIAS et al.: EXPERIMENTAL EVALUATION AND COORDINATED DEPLOYMENT OF RRL AGAINST RVC IN DISTRIBUTION SYSTEMS 1845 TABLE III RVCSATBUS 1ACCORDING TO THE IEC STANDARD Algorithm 2: Breakdown of the Parameter-Acquisition Stage. 1: Select a worst-case profile for the primary energy source. 2: Determine representative buses for the distribution system under study. 3: Calculate the voltage magnitude-active power sensitivity factors between the buses of Step-2 and the CIRES buses. 4: For the profile of Step-1, detect the induced RVCs at the selected buses of Step-2, as per the IEC Standard. 5: For each bus of Step-2, identify the most severe RVC and store the Mvalues of Urms(1/2) before its occurrence. Case 3, the achieved reduction reaches 92%, owing to the stricter RRlim value. V. PROPOSED METHOD FOR EFFICIENT DEPLOYMENT OF RRL AGAINST RVCS Having validated the RVC-suppressing effect of RRL schemes in Section IV, this section proposes a new systemplanning method for the coordinated mitigation of RVCs in active distribution systems. More specifically, for a given distributionsystem,themethoddeterminestheCIRESunitstooperate under RRL control, as well as the necessary RRlim values, to mitigate RVCs under minimum technical requirements, such as thenumberofCIRESunitswithRRLcontrolorthetotalrequired ESS capacity. Note that the method addresses RVCs according to the IEC Standard, due to its more comprehensive guidelines. The method consists of two stages: parameter acquisition and formulation of the optimization problem. A. Parameter-Acquisition Stage The solution of the method derives from an optimization problem. For the formulation of the optimization problem, a parameter-acquisition stage needs to precede. More specifically, this stage provides a framework for addressing a worst-case scenario in terms of RVCs. As the method is solved offline for system-planning purposes, the parameter-acquisition stage is requiredtobeexecutedonly if one ofthefollowingchanges occurs in the distribution system: (a)a new CIRES is connected; (b)an existing CIRES is decommissioned; (c)a change occurs related to the lines or transformers of the grid. The algorithmic steps of the parameter-acquisition stage, as intended to be followed by DSOs, are summarized in Algorithm 2and thoroughly discussed below: Step-1: ToensurethepreventionofRVCsunderallconditions, the RRL allocation needs to be dimensioned according to a worst-case, volatile input profile, corresponding to the primary energy source of CIRESs. If available, DSOs can derive volatile profiles by analyzing historical datasets with measurements of CIRESactivepowerfromtheirdistributionsystemorbyconsulting historical data of weather conditions, e.g., [27],[28],from the specific location of their distribution system. Alternatively, DSOs can select benchmark volatile profiles available in the technical literature, such as those of [9],[16]. It should be noted that the compact size of distribution systems allows for the assumption of uniform weather conditions across them, e.g., solar irradiance, ensuring that all CIRES units are subject to the same input profile. Moreover, this uniformity represents the worst-case scenario in terms of the induced voltage variations, as the changes in power of all CIRES units will be synchronized, and thus, their effect on voltage maximized. Step-2: The strategy aims to ensure the total suppression of RVC events for all buses of a distribution system. Nevertheless, it is impractical to examine all buses while implementing the method. Therefore, the strategy is based on the reasonable selection of some representative buses, according to the following logic: If the emergence of RVCs is prevented for those buses, then it is safe to assume that it has been prevented for all network buses. These representative buses can be selected by consulting the voltage-to-active power sensitivity matrix of the distribution system and employing network partitioning techniques [29], [30]. Step-3: Subsequently, the voltage-to-active power sensitivity factors between the representative buses of Step-2 and the CIRES buses are calculated. The calculation can either be performedviaapowersystemsanalysissoftwareoraccording toappropriate scientific works, e.g. [26],[31],[32]. It should be noted that the calculation of the sensitivity factors only requires the knowledge of the topology, as well as of the line and transformer impedancesofthedistributionsystem,whichis usuallyavailable to the DSOs. Moreover, in case the precise distribution system parameters are not fully known, measurement-based techniques, e.g., [33],[34],[35], can be employed to estimate the required parameters, and through them, the sensitivity matrix. Step-4: Executing rms simulations on the examined distribution system under the profile of Step-1, the RVCs at the buses of Step-2 are identified. The power of the rest distribution system components, e.g., loads, is not altered, hence RVCs only depend on CIRES volatility. Step-5: The most abrupt RVC event for each bus is identified. Abruptness is defined as the difference between the rms voltage immediately before and after the onset of an event. This event 1846 IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 61, NO. 2, MARCH/APRIL 2025 corresponds to the most abrupt active power change and, reasonably, exhibits the strictest requirements for mitigation. Thus, it is rational to assume that the successful suppression of this event implies the suppression of all RVCs. For this event, the M values of Urms(1/2) before its occurrence are stored. In summary, the profile of Step-1, the sensitivity factors of Step-3, and the Mvoltage values per examined bus of Step-5 are the parameters required for the formulation of the optimization problem, which is presented in Section V-B. B. Optimization Problem Formulation Leveraging the parameters obtained in Section V-A, an optimization problem is formulated to efficiently allocate the RRL functionalities among CIRES units, for the mitigation of RVCs. The problem is centrally solved by the DSO at their premises and is formulated as a mixed-integer linear programming problem to ensure low computational complexity and facilitate its execution. More specifically, let NCIRES be the set of buses with CIRESs, Nrbe the set of representative buses, and K= {1,2,...,M}be an integer set, with parameter Mdefined in Section II-A. The voltage-to-active power sensitivity factor between a bus i∈N rand bus j∈N CIRES is denoted as ΔUi ΔPj, while the rating of the CIRES unit at bus jis Sj. Moreover, Uk past,i denotes the k-th preceding voltage value before the most severe RVC event at bus i, with k∈K,i∈N r. The optimization problem treats the Uk past,i values as known parameters and determines the required ramp rates of the CIRES unitsfor thenexthalf-cycle,suchthatthe occurrenceof thisRVC event would have been prevented. Elaborating, the most severe RVC event at each bus i∈N rwould be avoided, if the rms voltage value of the next half-cycle were such that all Uk past,i values, and the new value, lied within the threshold of their arithmetic mean value. Therefore, let U∗ idenote this desired voltage value at bus i∈N rfor the next half-cycle, so that the RVC event is not triggered. This desired U∗ iwill result in a value for the arithmetic mean of the M+1 half-cycle voltage values, denoted as U∗ mean,i. Eventually, in order for the desired U∗ ivalue to be achieved and the RVC event to be prevented, the ramp rate of the CIRESs for the next half cycle should be RR∗ j, j∈N CIRES. Based on the above, the optimization problem is mathematically formulated as follows: minf(RR∗ j),j ∈N CIRES (1) s.t. (1−th)·U∗ mean,i ≤Uk past,i ≤(1+th)·U∗ mean,i,i∈Nr,k∈K (2) (1−th)·U∗ mean,i ≤U∗ i≤(1+th)·U∗ mean,i,i∈N r(3) U∗ mean,i =1 M+1(U∗ i+ M  k=1 Uk past,i),i∈N r(4) U∗ i=UM past,i + j∈NCIRES ΔUi ΔPj ·Sj·RR∗ j·Δthc,i∈N r (5) RR∗ j=g(RR),j ∈N CIRES (6) Equation(1)expressestheobjectivefunctionofthe optimization problem, which is a function of the required ramp rates. Eq. (2) and (3) represent the conditions for the prevention of the RVC event, while (4) is the definition of the U∗ mean,i.In(2) and (3), th is the considered percentage value for the threshold. Finally, (5) represents the linearized equation that correlates the rms voltage of the next half-cycle, U∗ i, with the rms voltage value of the current half-cycle, UM past,i, with respect to the sensitivity factors and the ramp rates of CIRES units. In (5),Δthc is equal to the half-cycle duration, while RR∗ jis expressed in pu/s with respect to Sj.Eq.(6) determines RR∗ jfrom a permissible range of ramp rate values, represented by vector RR. Equations(1) and(6) areabstractly expressed,astheir specific form depends on the particular version of the optimization problem. In the present study, three distinct versions of the optimization problem are addressed, as follows: rProblem 1: Considering a fixed RRlim value, minimize the number of CIRES units operating under RRL control. rProblem 2: Considering a fixed RRlim value, minimize the total required ESS capacity. rProblem 3: Considering a wider range of applicable RRlim values, minimize the total required ESS capacity. Elaborating on the considered problems: Problem 1: Here,CIRES unitscaneitheroperatewithoutRRL control or under RRL control with RRlim. Thus, the permissible range of values for RRjis RR =[RRlim,RR in], with RRin being the known, input ramp rate of the CIRESs for the next half-cycle, as dictated by the selected volatile profile. Therefore, (6) corresponds to: RR∗ j=(1−b∗ j)·RRin +b∗ j·RRlim,j ∈N CIRES (7) A binary variable, b∗ j, is introduced for each CIRES unit. If b∗ j=1, the corresponding unit is decided to operate under the RRL scheme and be involved in the RVC mitigation. Thus, the objective function becomes: min  j∈NCIRES b∗ j(8) Problem 2: This formulation preserves (7) but takes into account the different power ratings among CIRES units, which, in turn, affect the energy ratings of their corresponding ESSs. More specifically, for the same input power profile and under the same RRlim, the required maximum energy capacity of an ESS is proportional to the power rating of its associated CIRES unit. Therefore, the objective function becomes: min  j∈NCIRES b∗ j·rj(9) DIMOULIAS et al.: EXPERIMENTAL EVALUATION AND COORDINATED DEPLOYMENT OF RRL AGAINST RVC IN DISTRIBUTION SYSTEMS 1847 Here, rjis the rated-power ratio between the CIRES unit at bus j and the CIRES unit with the lowest rated power. Thus, (9) minimizes the requirements in aggregate ESS capacity, assigning weights to CIRESs according to their power rating. Problem 3: This is the most generalized formulation, where a discrete range of RRlim values is assumed. The RRlim values and the RRin value comprise set NRRL and are denoted as RRm,m∈N RRL. Each unit could operate under one of these values or without RRL control. This is expressed by (10), through the usage of binary variables, bm∗ j.  m∈NRRL bm∗ j=1,j ∈N CIRES (10) Therefore, (6) becomes (11), as follows: RR∗ j= m∈NRRL bm∗ j·RRm,j ∈N CIRES (11) The optimization criterion is, similarly to Problem 2, the minimization of the total required ESS capacity. In this problem, the energy capacity requirements of an ESS depend on both the rating of the associated CIRES and the RRlim value. To estimate the effect of RRlim on energy rating, the energy profile corresponding to each examined RRlim is theoretically calculated, for the CIRES with the lowest rated power. The resulting energy usage,alongwiththe CIRES ratings, produce weight factorsthat represent the energy capacity requirement of CIRESs. These factors are denoted as cm j,j∈N CIRES,m∈N RRL. Hence, (1) becomes: min  j∈NCIRES ( m∈NRRL bm∗ j·cm j)(12) VI. VALIDATION OF THE PROPOSED METHOD The validation of the proposed method is performed as follows: Initially, the system under study is presented, for which the parameter-acquisition stage is executed. Subsequently, the method is applied, deriving the optimized solution for the deployment of RRL control. Finally, the effectiveness of the derived solution in suppressing the RVCs of the system under study is tested via dynamic rms simulations. A. System Under Study Theproposed strategyisappliedto theIEEEEuropean LVtest feeder [36]. This distribution system consists of 906 buses and 55 loads. In this study, the loads are modeled as three-phase, constant-power, and ohmic, with a consumption of 3kW. The short circuit capacity of the upstream feeder is 63 MVA. Note that this distribution system is selected, instead of the CIGRE MV feeder of Section IV, as its extended size and complexity render it more suitable for demonstrating the effectiveness of the proposed method in allocating the RRL control among multiple CIRESs. To simulate scenarios of high CIRES penetration, 32 PV units are dispersed across the distribution system. For the same PV locations, two distinct scenarios are examined, corresponding to different rated power values per PV unit. The two scenarios, TABLE IV PV RATINGS PER SCENARIO TABLE V RVCSOFTHEBASE CASE (ΔUMAX AND ΔUSS IN %) alongwiththe topology of the distributionsystem,arevisualized in Fig. 12. As depicted, in Scenario S1, the 16 most remote PVs have a 20-kWp rating, while the 16 most proximal to the substation PVs have a 10-kWp rating. Scenario S2 regards the opposite condition. The rated power and the location of the PVs for both scenarios are mapped in Table IV. B. Execution of Parameter-Acquisition Stage Here, the parameter-acquisition stage is executed, as follows: Step 1: The volatility of the PV primary energy source is modeledbytheinputactivepowerprofileofFig.13.Specifically, to derive this profile, the ramp rates of the real, daily PV profile of [16] were analyzed, and the 600-s window with the most severe power variations was isolated. The corresponding input ramprates arealso depictedinFig.13. Theprofiles areexpressed in pu to apply to distinct PV ratings. Step 2: For the examined distribution system, Buses 619, 861, and 882 are selected as representative. As observed in Fig. 12, these buses are remotely located in distinct areas of the distribution system. Step 3: The voltage-to-power sensitivity factors are calculated through the DIgSILENT Powerfactory software. Step 4: Using Powerfactory, rms simulations are executed to investigate the induced RVCs under the profile of Fig. 13 at the selected buses. The simulation time step is 10ms. The induced RVCs for Scenarios S1, S2 are aggregated in Table V, as per the IEC Standard. Specifically, the total number of events, N, and the maximum values of ΔUmax and ΔUss, for those N events, are recorded. The lower value of th =1% is considered, to generate a base case with multiple RVC events. As shown, although the total PV capacity in both scenarios is 480 kWp, RVCs in Scenario S1 are more frequent and of higher intensity, due to the higher rating of the PVs at the remote buses. Overall, up to 9 RVCs are recorded, for a 600-s period, i.e., 1/6 of an hour. As a comparison, the IEEE 1547-2018 Standard limits the allowable number of RVCs to only 2 ≤N≤10 per hour.