A novel hybrid home energy management system considering electricity cost and greenhouse gas emissions minimization

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/ This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TIA.2021.3057014, IEEE Transactions on Industry Applications 1 A novel hybrid home energy management system considering electricity cost and greenhouse gas emissions minimization Sara Barja-Martinez, Fabian R¨ ucker, M` onica Arag¨ u´ es-Pe˜ nalba, Roberto Villafafila-Robles, ´ Ingrid Munn´ e-Collado and Pau Lloret-Gallego Electrical Engineering Department CITCEA - Universitat Polit` ecnica de Catalunya Av. Diagonal, 647, 08028, Barcelona, Spain Abstract—This paper proposes a multi-objective hybrid energy management system that minimizes both the electricity expenses and the household greenhouse gas emissions released due to consumption, considering the entire life cycle of the generation assets used to provide energy. The global warming potential indicator is used to decide if it is more sustainable to purchase electricity from the grid or use the household’s flexible generation sources like photovoltaic panels and the energy storage system. Results prove that it is possible to reduce greenhouse gas emissions without incurring expensive electricity bill costs thanks to the hybrid-based home energy management system approach. This method gives the end-users a more influential role in the climate change solution, allowing them to give more or less importance to the economic or environmental component, according to their preferences. Index Terms—flexibility, life cycle assessment, multi-objective optimization, batteries, greenhouse effect I. INTRODUCTION Global carbon dioxide emissions reached in 2019 an alltime high, despite coal fading [1]. Decarbonization is vital; for this reason, the electricity sector has already started moving from fossil-based to net-zero greenhouse gas (GHG) emissions thanks to the increasing number of renewable energy sources (RES) and the use of flexibility in the power system to enhance the grid integration and maximize renewable energies potential. A long-term forecasting study about the evolution of the global energy transition is conducted in [2]. Key findings predict a sharp drop in fossil-fuel use (around 75% by 2050) and warn that the Paris agreement will not be accomplished if no further decarbonization measures are taken. Focusing on Europe, current policies will reduce GHG emissions by 60% in 2050 compared to 1990 emissions levels. Nevertheless, the European Commission has increased its climate ambition through the European Green Deal by enumerating key transformative economic policies and measures to put Europe on track to achieve the goal of net-zero global warming emissions by 2050. Concerning the residential sector, buildings and households play a crucial role in the energy transition. In 2018, they represented 28% of the global energy-related carbon dioxide emissions [1]. Due to the still existing margin of improvement in the energy efficiency field, buildings are expected to be the fastest sector reducing the CO2emissions [3]. Thus, more robust strategy measures to decrease GHG emissions associated with residential electricity demand need to be addressed. This work proposes that intelligent home energy management systems (HEMS) contribute to achieve environmental targets. In literature, there are principally two HEMS approaches: price-based (PB) -most of the current workand incentive-based (IB). The PB program aims to minimize the end-user electricity bill by re-scheduling controllable flexible sources optimally, considering a dynamic pricing tariff. Several studies have applied multi-objective functions HEMS for optimal scheduling considering electricity cost and end-user discomfort minimization [4–12]. For instance, [8] minimizes electricity cost and the power profile deviation at the point of common coupling, while the author in [9] proposes a cost-effective HEMS considering thermal and electricity comfort. The IB program offers flexibility to a third electricity agent to exchange economic compensation for changing its baseline consumption. [13] assures minimum energy cost and supports the upstream micro-grid operation by minimizing the load profile deviation. Electric vehicle and electric water heater provide flexibility in [14] for PB and IB programs. A third environmental-based (EB) category has been proposed in [15], which focuses on minimizing the GHG emissions produced by the generation units that provide electricity to the household. The study carried out in [16] presents a multiobjective dispatching optimization model of an energy system focused on the energy production, conversion, and storage to maximize the operating revenue and to minimize operational risk and carbon emissions. Focusing on HEMS sustainability factors, the study [17] calculates climate effects by displaying carbon emissions at customers’ premises to motivate them to diminish their consumption. In [18], curtailment of on-site PV is penalized to maximize green energy consumption. This paper introduces a hybrid-based (HB) HEMS, which is a mixture of the already mentioned PB and EB approaches. Some HEMS programs found in literature are listed in Table I. The vast majority of the studies focus on PB programs, which is also stated by [14]. Some works propose multi-objective functions that consider both PB and IB programs [13, 14, 16]. To the best knowledge of the authors, no HEMS approach in literature focuses on both electricity bill and GHG mini- This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/ This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TIA.2021.3057014, IEEE Transactions on Industry Applications 2 TABLE I HEMS PROGRAMS AND THEIR SERVICES. HEMS program Description HEMS service PB [4–7, 13, 14, 18– 20] The objective is to minimize the endusers electricity bill. Time-of-use pricing, real-time pricing and peak shaving. IB [13, 14, 21, 22] Flexible sources are economically incentivized to be flexible by modifying their electricity use. Providing flexibility to a third energy agent. EB [15] Flexibility is used to minimize the GHG emissions of buildings associated to generators that produce the electricity they consume. Minimization of GHG. mization. The most significant contributions of this paper are summarized as follows: •The development of a multi-objective hybrid HEMS that minimizes the costs and the GHG emissions associated with generators that produce the electricity that the household consumes by assigning weights to the two competing objectives. •The implementation of a life cycle assessment (LCA) methodology to measure the impact that each generation technology has on the climate, from cradle-to-grave. Thus, all the emissions expected during its life cycle are considered, having a more realistic picture of the emissions released into the atmosphere. •The implementation of a degradation battery model focused on HEMS that considers calendar and cycling aging constraints. •The proposed HEMS enables the end-users to have a more influential role in the climate change solution by giving more weight to the multi-objective function’s environmental component. The following bullets enumerate the research gaps: •Controllable flexible electrical loads are not considered in this study. •Due to the previous point, the discomfort penalization is not in the HEMS objective function since the only flexible sources are battery and PV generator. •The case studies proposed are two extreme opposing energy mix scenarios, one with 83.07% of non-carbon emissions generation share, and the other with 29.18%. The intention is to highlight the HB HEMS potential. This paper is structured as follows. The LCA methodology and the time-varying global warming potential indicator of the generation technologies are explained in Section II. The mathematical formulation is expressed in Section III. Section IV explains the case studies for evaluating the three HEMS proposed, while the results and sensitivity analysis are presented and discussed in Section V. Finally, conclusions are drawn in Section VI. This work is an extension of previous study conducted by the authors [15]. II. LCA FOR ELECTRICITY GENERATION SYSTEMS:A TIME-VARYING GWP APPROACH This section estimates the real environmental impact considering the entire life cycle of an electricity generation system, using the LCA methodology whose evaluation goes from raw materials and fuel extraction to the gate of the generation plant, through materials processing, plant operation and maintenance, and plant infrastructure commissioning and decommissioning. Global warming potential (GWP) is the LCA impact category selected as the reference measure to quantify and assess the potential environmental impact of an electricity generation source. The GWP indicators for the electricity generation sources evaluated in this study are noted in Table II and were obtained from [23]. Power systems consisting of diverse generation sources have time-dependent GHG emissions; therefore, the GWP performance changes over time along with the electricity mix [24]. One of the objectives is to calculate the hourly kg CO2 equivalent (CO2−eq) in one kWh of the energy mix. The hourly average GWP impact of the electricity supply from the grid is expressed as GWPgrid t=X i∈I GWPavg i·GSt,i (1) where GWPavg iis the GWP average constant for each type of generation source iand GSt,i refers to the estimated generation in the day-ahead market at period tfor each type of generation source i. Table II shows the GWP indicator range for the overall life cycle stages for each electricity generation source type, according to [23]. This paper applies the average GWP value. TABLE II LYFE CYCLE EMISSION FACTORS FOR ELECTRICITY GENERATION SOURCES. EXTRACTED FROM [23]. Generation source GSi GWP range [kg CO2−eq/kWh] Average GWP [kg CO2−eq/kWh] Hard coal 0.66-1.05 0.855 Lignite 0.8-1.3 1.050 Natural gas 0.38-1 0.69 Nuclear 0.003-0.035 0.019 Biomass 0.0085-0.13 0.0693 Hydro-power 0.002-0.02 0.011 Photo-voltaic 0.013-0.19 0.1015 Wind 0.003-0.041 0.022 Battery - 0.0706 III. MATHEMATICAL FORMULATION This section covers the three HEMS objective functions approaches -PB, EB and HBand mathematical formulation of the optimization model for controlling and re-scheduling flexible household sources. A. HEMS objective function 1) Price-based program: this program focuses exclusively on the economic aspect. It aims to minimize the electricity bill (2a), considering the battery degradation cost Kcalendar tdue to calendar aging, where Pbuy tis the time-varying electricity This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/ This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TIA.2021.3057014, IEEE Transactions on Industry Applications 3 price, χbuy trefers to the energy purchased to the grid, and PV AT is the tax applied. Constraint (2c) ensures that the energy balance is always met, where ψpv tis the optimized PV generation output and Winflex tstands for inflexible household consumption. Finally, constraint (2d) avoids exceeding the contracted power Xmax,imp. To switch from power to energy units, Nhour is used, which refers to the number of periods per hour. The objective function is expressed as min χ, V z1=X t∈T (Pbuy tχbuy tPV AT +Kcalendar t)(2a) s.t. Kcalendar t= 0.019 ·Vt−0.0629,(2b) ψpv t+σdis t+χbuy t=σch t+Winflex t,(2c) χbuy t≤Xmax,imp/Nhour (2d) 2) Environmental-based program: this program attempts to minimize the carbon footprint occasioned by the generation sources that provide electricity to the household. This HEMS is presented in [15]. The objective function is formulated as follows min χ, ψ, σ z2=X t∈T GWPgrid tχbuy t+GWPpvψpv t+(3a) +GWPbatσdis t(3b) s.t. ψpv t+σdis t+χbuy t=σch t+Winflex t,(3c) χbuy t≤Xmax,imp/Nhour (3d) where GWPgrid tindicates the kg CO2−eq/kWh of the grid on average per period t. It is calculated with the hourly energy production mix, taking the values of scheduled generation in the day-ahead market for each technology described in Table II. 3) Hybrid-based program: this approach is a multiobjective problem (MOP) that combines the PB and the EB objective functions. It is a multiple criteria decision-making problem with no unique optimal solution, but a domain of feasible solutions that satisfy all constraints. Therefore, the result is a trade-off, a compromise between minimizing the energy costs and decreasing GHG emissions derived from the generation sources that provide electricity to the house. The HB objective function z3is formulated in (4a). Normalization of the objectives is required so that both competing objectives can be equivalent and compared at the same level. z∗ 1is the optimal solution of the PB objective function z1(2a), and z∗ 2is the optimal solution of the EB objective function z2 (3b). The linear MOP is formulated as min χ, ψ, σ, V z3=α·z1 z∗ 1 +β·z2 z∗ 2 (4a) s.t. Kcalendar t= 0.019 ·Vt−0.0629,(4b) ψpv t+σdis t+χbuy t=σch t+Winflex t,(4c) χbuy t≤Xmax,imp/Nhour,(4d) α+β= 1 (4e) where αand βare weighting factors for PB and EB objective function, respectively. The sum of these two factors must be one (4e). To conclude, Table III shows the formulation of the objective functions of the three HEMS programs. TABLE III HEMS PROGRAMS’OBJECTIVE FUNCTIONS. Objective function Mathematical formulation Price-based [MIN]z1=Pt∈T(Pbuy tχbuy tPV AT +Kcalendar t) Environmentalbased [MIN]z2=Pt∈T(GW P grid tχbuy t+GW P pvψpv t+ GW P batσdis t) Hybridbased [MIN]z3=αz1 z∗ 1 +β2 z∗ 2 B. Energy storage system Battery aging is formed by calendar and cycling aging. Calendar aging happens during the battery rest time, whereas cycling aging is caused directly by charges and discharges. According to [25], the Li-ion battery degradation due to cycling shows minimal aging for low current rates [26] and also when the battery is not charged to its real maximum state of charge (SOC) since there is a faster degradation when charging to 100% SOC. Therefore, the following constraints explained in Section III-B2 are added to the battery model to ensure that the storage unit works under conditions that minimize the cycle aging impact: •Equation (13) ensures that the battery charges and discharges at low current rates. •Equation (14) reduces large cycles at high SOC by limiting the maximum SOC allowed. Given the above, the battery calendar aging is formulated. 1) Battery calendar aging: Battery operating conditions have a significant impact on their performance and life time. The storage model applied in this paper considers calendar aging for a lithium-ion (Li-Ion) battery. This phenomenon leads to a decrease in usable battery capacity and an increase in the battery’s inner resistance over time, resulting in a depreciation cost. The calendar aging model formulation applied in this study is parametrized in [26] through accelerated aging tests. The capacity defined in (5) is a phenomenon where the volume of energy that a battery can operate at the rated voltage diminishes over time [27]. The variables that impact calendar aging, thus influence battery life time, are cell temperature and voltage. The loss of capacity is more prominent than the This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/ This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TIA.2021.3057014, IEEE Transactions on Industry Applications 4 resistance increase in the calendar aging function according to [26], as the end of the battery life is reached first due to the loss of capacity. For this reason, only the capacity is considered in the calendar aging formulation. For a Li-Ion battery cell, the capacity Cdue to calendar aging is expressed as C(t)=1−ψ(V, T)·t0.75 (5) where ψis an aging factor that describes the aging rate during period tand is formulated as ψ(V, T) = (a·Vcell t−b)·e−c/T (6) where temperature is a constant parameter T= 293 K in this study, a= 7.543 ·106V−1days−0.75,b= 2.375 · 107days−0.75 and c= 6976 K[26]. The relationship between the open-circuit voltage (OCV) and SOC is known and expressed by the non-linear equation shown in Figure 1. In order to avoid non-linear constraints, the dependence between the OCV of the cell and SOC is linearized. The minimum SOC by restriction is limited to 25%. Fig. 1. Linear and non-linear OCV and SOC dependence. As a result, the linear correlation is represented as Vcell t= 0.0076 ·SOCt+ 3.4287 (7) where variable SOCtindicates the percentage of energy stored per period t. The depreciation of the battery during each time step ∆tleads to Kcalendar tcosts Kcalendar t(L, ∆t) = Kinvestment L∆t(8) where Kinvestment is the acquisition cost of the 8 kWh LiIon battery, and it is set in 7500 e,Lis the life time of the battery and ∆t the time step. The end of life criterion is defined to be 80% of initial capacity C[27]. Therefore, the expected battery life can be calculated as C= 0.8 = 1 −ψL0.75, so equation (9) remains Kcalendar t(V, T, ∆t) = Kinvestment 0.2 ((a·Vcell t−b)·e−c/T )1/0.75 ∆t(9) A linear approximation to equation (9) is calculated to relax the constraint and implement a linear solving method. Figure 2 shows both non-linear and linear equations, proving that the functions’ behavior is practically identical. Fig. 2. Dependence of cell voltage on SOC for a cell temperature of 293 K and time step ∆t= 15 minutes. Therefore, the linearized Kcal,linear tper time step tis formulated as Kcalendar,linear t(V)=0.019 ·Vcell t−0.0629 (10) 2) Battery constraints: The battery model constraints are formulated. The variable σsoc tin equation (11) represents the battery SOC for each period. To represent the real behavior of the battery, the efficiency factors for storing ηch and delivering electricity ηdis are considered. The variables σch tand σdis t represent the amount of energy charged or discharged in each period. σsoc t=σsoc t−1+σch t·ηch −σdis t ηdis (11) To avoid over-optimistic results, the battery SOC must be the same at the beginning and the end of the optimization horizon. σsoc t=0 =σsoc t=final (12) As mentioned before, it is essential to ensure that the battery is not fully charged or discharged by limiting its maximum and minimum allowed SOC to a specific fixed value to avoid cycle aging. The equation (13) ensures that σsoc tis always between a minimum Omin and a maximum Omax to preserve and extend the battery life time: Omin ≤σsoc t≤Omax (13) Equations in (14) also help to minimize cycling aging by limiting the maximum power allowed for charging Qch and discharging Qdis. σch t≤Qch Nhour , σdis t≤Qdis Nhour (14) Equations in (15) make sure that the energy charged per period σch tis linearly decreased. Sch is the threshold in This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/ This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TIA.2021.3057014, IEEE Transactions on Industry Applications 5 charging process. The same happens with the discharging energy σdis t. The threshold to limit the energy output is Sdis. σch t≤−Qch 1−Sch (σsoc t Omax −1), σdis t≤Qdis Sdis σsoc t Omax (15) C. PV generation constraints The formulation of a reducible PV generation model is presented. The optimization variable PV scheduled generation ψpv tmust be between 0 and the PV baseline electricity generation Wpv t, which is the forecasted PV generation curve for the following day. 0≤ψpv t≤Wpv t(16) IV. CASE STUDIES The case studies presented in this section aim to analyze the three HEMS program’s performance -PB, EB, and HBto compare the electricity expenses and kg of CO2−eq related to a single-family household. Real consumption and PV generation data are taken from the Data Port database [28] and used as input to the HEMS programs. These case studies are located in Spain; therefore, the Spanish dynamic electricity tariff (Precio Voluntario Peque˜ no Consumidor tariff) and its electricity mix are used as input data. PV AT is set to 21%. The optimization horizon is 24 hours, divided into 96 time periods of 15 minutes, starting at 00:00h. The household contracted maximum power is 6 kW and is equipped with a 4.8 kW PV and an 9 kWh battery, whose SOC must be at least 50% at the beginning and end of the optimization horizon. The value of the parameters applied for all the case studies are listed in Table IV. The end-user does not sell back electricity to the grid; therefore, the PV is exclusively for self-consumption, and the excess of production can be stored in the battery for later usage. The HB multi-objective function weights are set to α= 0.3and β= 0.7, according to the end-user preferences that emphasize environmental aspects. The HEMS has been implemented in Python, using the Pyomo optimization library and the Gurobi solver. The optimal solution of the HB HEMS program was obtained with a computational time of 0.44 seconds on a Laptop with a processor core i7 at 2,60 GHz and 8 GB of RAM TABLE IV CASE STUDIES PARAMETERS. Parameters Value Units Battery maximum allowed SOC 8 kWh Battery minimum allowed SOC 2 kWh Battery SOC initial/final 4.5 kWh Battery maximum power charge/discharge 3 kW Battery efficiency charge/discharge 0.95 - Household maximum import capacity 6 kW PV maximum output power 4.8 kW Two opposite scenarios of the Spanish energy mix generation are proposed to examine the HEMS program’s performance. On the one hand, low penetration of renewables in the electricity mix and, on the other, high participation of sustainable generation sources. A scheme of these case studies is presented in Figure 3. The PB HEMS is run separately since its performance only depends on the dynamic pricing tariff, regardless of the energy mix composition, since its objective is to minimize the cost, not GHG emissions. Case studies have identical inflexible demand, PV generation, battery parameters, and hourly electricity prices to compare the results of the three HEMS programs. The only input parameter that changes is the electricity generation mix of the grid. Fig. 3. Scheme of the case studies proposed to test HEMS programs performance. For Scenario A with low RES penetration, data from November 20th 2017 is used, which percentage of non-fossil generation penetration is 29.18%. For Scenario B with high RES penetration, March 6th 2020 has been selected, with a daily average of 83.07% of electricity generation sources with zero emissions during their electrical grid operation. It should be noted that nuclear power is incorporated within the zero-emissions energy sources. The hourly share of each generation source -listed in Table IIin the Spanish energy mix for both scenarios is illustrated in Figure 4. The total generation curve is also represented to demonstrate that the selected generation types are primarily responsible for the overall generation and cover 96% of the total demand. The higher the RES penetration in the energy mix is, the lower the GWP grid value per energy unit. Combined cycle and coal generation dominate in Figure 4(a), while wind power does it in Figure 4(b). Fig. 4. Generation sources share considered for both scenarios. This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/ This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TIA.2021.3057014, IEEE Transactions on Industry Applications 6 V. RESULTS In this section, the case studies’ results are presented and discussed. For a better understanding of the graphical outcomes obtained (see Figure 5, for instance), it should be noted that negative energy values represent a generation source like PV and battery discharging. In contrast, positive values represent energy consumption, such as inflexible loads or battery charging. Therefore, due to the energy balance equation (2c) where generation must match consumption, the resulting plot is symmetrical if all the generation and consumption sources are summed up separately. A. Price-based scenario The PB case study analyzes the HEMS behavior under dynamic price tariff with the explicit aim of minimizing the end-user electricity cost. The PB objective function result is identical for high and low RES penetration scenarios since it only depends on the electricity tariff variation. However, the GHG emissions indirectly caused by PB optimization vary depending on the scenario. For clarification, the PB program is executed if the parameters are set to α= 1 and β= 0 in the HB MOP, as equation (4a) indicates. The PB HEMS optimization results are displayed in Figure 5. The upper graph shows the electricity price, while the lower displays the PB optimization result. Focusing in Figure 5(b), during periods of low prices (4-27), the consumed electricity is purchased directly from the grid, taking advantage as well to charge the battery (19-23, 91-94) to discharge it later during time intervals with more expensive costs (28-33, 66-81). PV allows self-consumption during most daylight periods, and the surplus energy is used to charge the batteries, reaching the maximum capacity of 8 kWh in period 66. The battery is charged again in the last low-priced periods (91-94) to meet the restriction of ending at least half of its SOC. The total cost of the objective function is 2.79 e. If the result is broken down, 52% belongs to battery degradation cost, while 48% corresponds to the price of buying electricity from the grid, including taxes. Fig. 5. Results of price-based HEMS under a Spanish price scheme. B. Scenario A: low penetration of RES in the energy mix The outcomes obtained in Scenario A for EB and HB HEMS programs are displayed in Figure 6. The hourly-varying GWP of the grid along with electricity prices are represented in Figure 6(a). Figure 6(b) displays the EB and Figure 6(c) shows the HB results. The EB program is fed from the grid during periods with moderate kg CO2−eq levels (0-27) compared to the daily GWP average. In periods of high grid GWP indicators, the battery is discharged (28-35) to avoid consuming from the network. PV generation is diminished (43, 46, 52, 54) as it is not possible to store the surplus energy in the battery due to its SOC limits. Besides, the sum of the kg of CO2−eq per kWh of PV and the battery charge has a higher environmental impact than purchasing straight from the grid in specific periods. To comply with the battery SOC restriction at the end of the optimization horizon, electricity is bought from the grid to charge the battery when the grid GWP levels are low (see periods 38-60). Concerning the HB HEMS program in Figure 6(c), it is discerned that compared with the EB HEMS, the battery discharges at moderately high prices compared to the next periods (0-3). In (28-30), the battery discharges due to the high GWP values in the power system, although less energy than the EB, as the prices for that period are more high-priced. Solar energy does not reduce its production and is used for self-consumption. Meanwhile, the surplus energy is used for charging the battery. The HB avoids buying during the most expensive intervals of the day (76-80). The energy needed to charge the battery until the SOC imposed (4.5 kWh) at the end of the optimization horizon is purchased from the grid at affordable prices (91-96). Fig. 6. Results of an environmental-based and hybrid-based HEMS in Scenario A under a Spanish price scheme. This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/ This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TIA.2021.3057014, IEEE Transactions on Industry Applications 7 C. Scenario B: high penetration of RES in the energy mix The electricity generation of the grid is composed of 83% on average by CO2free generation sources, including nuclear energy and RES. Grid GWP values are 4.5 times lower than in the previous Scenario A, while the price signals, inflexible household consumption, and PV remain the same. EB HEMS results are shown in Figure 6(b). Due to the high penetration of renewables in the electrical system, the EB purchases energy from the grid practically in its entirety, excluding daylight hours when the household is self-supplied, generating just the electricity required to meet the consumption. The EB program outcomes confirm that if a nation’s energy mix is highly renewable as in Scenario B, the usage of batteries is more polluting than purchasing directly from the grid. The HB HEMS outcomes are represented in Figure 6(c). It avoids purchasing from the grid during expensive periods (0-6) and intermittent periods (68-82). It uses the PV surplus (46-65) to charge the battery and discharge it later during the high prices (68-72). Fig. 7. Results of an environmental-based and hybrid-based HEMS in Scenario B under a Spanish price scheme. D. Sensitivity analysis The Pareto optimal front for the two HB normalized objective components are represented in Figure 8(a), choosing as weighting factors αvarying from 0 to 1 in steps of 0.02 and β= 1 −α. The blue and purple points sequences represent the border of the feasible solution region that satisfies all the restrictions imposed for each GWP scenario. Figure 8(b) shows the z3optimal solutions for each αfor high and low GWP scenarios. A sensitivity analysis is performed in Figure 9 to show how the HB objective function z3is affected based on changes in 0.0 0.2 0.4 0.6 0.8 1.0 EB (z2/z2*) [-] 0.00 0.20 0.40 0.60 0.80 1.00 1.20 PB (z1/z1*) [-] (a) Pareto front 0.0 0.2 0.4 0.6 0.8 1.0 alpha [pu] 1.00 1.02 1.04 1.06 1.08 1.10 HB (z3) [-] (b) HB objective function values High GWP Low GWP Fig. 8. (a) Pareto front and (b) HB solution values for high and low GWP scenarios. the following input variables: maximum allowed battery SOC in (a)-(b), PV generation output in (c)-(d), inflexible household consumption in (e)-(f), and electricity price average in (g)- (h) for low and high GWP scenarios. The legend shows the maximum allowed SOC, the total daily PV generation and consumption, and the average daily electricity price. Thanks to this sensitivity study, it is known how the variation of one parameter affects the HB HEMS optimal solution, reducing uncertainty. It is reminded that α= 0 corresponds to the EB HEMS program and α= 1 to the PB HEMS program. The continuous grey line refers to the case study optimal solution described in Section IV and displayed in Figure 8(b). The start and end of αtake the value one due to the normalization of HB objective function z3. Consumption is the variable that most affects the HB program. The lower the consumption, the lower the energy cost and emissions (see consum 14 kWh). On the contrary, the higher the consumption, the greater the cost and environmental impact (see 59 kWh). Consumption is followed by the electricity price, although when αequals to zero, this variable is insignificant since the EB HEMS does not consider the electricity price on its objective. The PV generation is more sensitive when it produces less (3 kWh) as this implies an increment in the electricity cost (α= 1) because more electricity needs to be bought, but it has barely any impact for high amounts of generation (see PV 25 kWh and PV 32kWh) because the household consumption is minimal compared with the PV generation. Finally, the input parameter that has the least influence is the maximum SOC allowed for the battery. E. Results comparison and discussion A comparative overview of the HEMS program results for both high and low RES penetration scenarios is presented in Table V, where the total GHG emissions and electricity costs are shown. Moreover, these HEMS programs are compared with the household baseline, which is the energy exchanged with the grid if no flexible resources were activated. In other words, the end-user buys all the power from the grid to meet the inflexible consumption. It is concluded that PB HEMS achieves the lowest electricity costs for both scenarios as expected, due to avoiding the purchase from the grid during expensive periods. However, in return, the PB HEMS has the most polluting emissions associated with its consumption, This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/ This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TIA.2021.3057014, IEEE Transactions on Industry Applications 8 Fig. 9. Sensibility analysis of different HEMS input parameters specifically releases 15.36% more emissions than the less polluting program -the EBin the high RES scenario. On the other hand, the PB produces 5.19% more emissions than the EB in the low RES scenario. On the economic side, the EB program pays 2.81 times more in the electricity bill than the PB in high RES situation and almost twice as much in a low RES scenario. As can be appreciated in Table V, the HB program has a balance between the PB and EB programs. For high presence of renewables in the energy mix, the difference between the HEMSs emissions does not exceed a range of 15%, while in the opposite RES scenario, the boundary is narrowed to approximately 5%. This is because during periods of high renewable generation in the energy mix, the EB purchases from the grid most of the periods, so the battery is not utilized for self-supply, neither is charged by PV surplus, resulting in a higher electricity bill when buying more energy from the grid. The baseline case only emits 2% more pollutant emissions than the EB program in the high RES scenario. The explanation is the same as in the previous paragraph: energy from the grid is cleaner than discharging batteries previously charged with solar PV surplus. On the contrary, on days with a high percentage of fossil-type generation in the grid, the kg of CO2−eq soared compared to the rest of the programs: 61% more emissions than EB, 58.16% compared to HB, and 53.44% compared to PB. The explanation is that the rest of the HEMS programs use the surplus PV generation to charge batteries for later use, avoiding to a great extent buying from the grid, which is much more polluting than the use of flexible resources. TABLE V RESULTS OF THE CASE STUDY α=0.3 β=0.7 GHG EMISSION AND ENERGY COSTS COMPARISON BETWEEN THE DIFFERENT ENERGY MANAGEMENT SYSTEM PROGRAMS. HEMS program GHG [kg CO2−eq] Electricity cost [e] High RES Low RES High RES Low RES Price-based program 5.33 14.39 1.33 1.33 Hybrid-based program 4.85 13.96 2.74 2.54 Environmental-based program 4.62 13.68 3.74 2.70 Baseline consumption 4.72 22.08 5.73 5.73 VI. CONCLUSION The present paper introduces a novel hybrid-based HEMS formulation that optimizes the operation of PV generators and distributed storage units behind-the-meter in order to achieve the best trade-off between electricity cost and GHG emissions minimization, considering a life cycle analysis of the generation sources used to meet the household demand. Two facing energy mix scenarios are proposed: high renewable energy participation and high fossil-type generation participation. The results confirm the reduction of GHG emissions in the HEMS containing the environmental component. The EB program achieves the lowest emissions, while the HB seeks a compromise between economic and environmental factors. By assigning weights to the HB multi-objective function, the end-user can modify its priority in a fast and flexible manner. The more renewable generation is in the energy mix, the lesser the difference between the HEMS programs’ emissions and the baseline case. However, electricity costs increase the more renewable energy is in the energy mix. Therefore, if the objective is strictly to minimize the environmental impact produced by the household consumption, it is concluded that if a nation regularly holds a very high penetration of renewable generation in its energy mix as it occurs in scenario B, it is more sustainable from the household point of view to buy electricity directly from the grid than using self-consumption with batteries, previously charged with PV surplus, for instance. In return for prioritizing and considering the polluting emissions minimization, the HB HEMS electricity expenses can be two times higher than the PB. On the other hand, for countries with low penetration of renewable generation sources in the energy mix, flexible resources such as batteries and PV panels significantly reduce GHG emissions and costs. Therefore, HB HEMS is an excellent option to encourage end-users to participate in the fight against climate change without causing high economic expenses. It should be noted This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/ This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TIA.2021.3057014, IEEE Transactions on Industry Applications 9 that, if second-life batteries were used for this purpose in the future, the value of the GWP parameter would be reduced, and the conclusions obtained in this study could vary. Finally, the sensitivity analysis carried out for a set of input variables indicates that household consumption is the input variable that most affects the HB objective function’s results, followed by PV generation, electricity price, and maximum allowed SOC, respectively. VII. ACKNOWLEDGEMENTS This research was funded by Ministerio de Ciencia, Innovaci´ on y Universidades under the project RTI2018-099540. The research of Sara Barja-Martinez and ´ Ingrid Munn´ eCollado has received the support of the Generalitat de Catalunya under the FI-AGAUR Pre-doctoral grants and scholarships, and has been co-funded by the European Social Fund (FSE). M` onica Arag¨ u´ es-Pe˜ nalba is a lecturer of the Electrical Engineering Department of UPC of the Serra H´ unter programme. We thank Pecan Street Inc. Dataport database for providing real household data. REFERENCES [1] International Energy Agency. Global CO2 emissions in 2019 – Analysis. Technical report, 2020. [2] DNV GL’s Group Technology & Research. Energy transition Outlook 2019. Waddenland Outstanding, pages 269–282, 2019. [3] IEA. Perspectives for the Clean Energy TransitionThe Critical Role of Buildings. page 117, 2019. [4] Yeongenn Kwon, Taeyoung Kim, Keon Baek, and Jinho Kim. Multi-objective optimization of home appliances and electric vehicle considering customer’s benefits and offsite shared photovoltaic curtailment. Energies, 13(11), 2020. [5] Jaclason M. Veras, Igor Rafael S. Silva, Pl´ acido R. Pinheiro, Ricardo A.L. Rabˆ elo, Artur Felipe S. Veloso, F´ abbio Anderson S. Borges, and Joel J.P.C. Rodrigues. A multi-objective demand response optimization model for scheduling loads in a home energy management system. Sensors (Switzerland), 18(10), 2018. [6] Reshma Reghukumar, S. Sambhu, and V. Ravikumar Pandi. Multi-objective optimization for efficient home energy management system using differential evolution algorithm. 3rd IEEE International Conference on Recent Trends in Electronics, Information and Communication Technology, RTEICT 2018 - Proceedings, pages 1157–1162, 2018. [7] Xiuwang Wang, Xinna Mao, and Hossein Khodaei. A multiobjective home energy management system based on internet of things and optimization algorithms. Journal of Building Engineering, 33(May 2020):101603, 2021. [8] M. C. Di Piazza, G. La Tona, M. Luna, and A. Di Piazza. A twostage Energy Management System for smart buildings reducing the impact of demand uncertainty. Energy and Buildings, 139:1– 9, 2017. [9] Gengqiang Huang, Junjie Yang, and Chunjuan Wei. Costeffective and comfort-aware electricity scheduling for home energy management system. Proceedings - 2016 IEEE International Conferences on Big Data and Cloud Computing, BDCloud 2016, pages 453–460, 2016. [10] Kornschnok Dittawit and Finn Arve Aagesen. Home energy management system for electricity cost savings and comfort preservation. IEEE International Conference on Consumer Electronics - Berlin, ICCE-Berlin, 2015-Febru(February):309– 313, 2015. [11] Nadeem Javaid, Fahim Ahmed, Ibrar Ullah, Samia Abid, Wadood Abdul, Atif Alamri, and Ahmad S. Almogren. Towards cost and comfort based hybrid optimization for residential load scheduling in a smart grid. Energies, 10(10):1–27, 2017. [12] Muhammad Waseem, Zhenzhi Lin, Shengyuan Liu, Intisar Ali Sajjad, and Tarique Aziz. Optimal GWCSO-based home appliances scheduling for demand response considering endusers comfort. Electric Power Systems Research, 187(January):106477, 2020. [13] Tohid Sattarpour, Daryoush Nazarpour, and Sajjad Golshannavaz. A multi-objective HEM strategy for smart home energy scheduling: A collaborative approach to support microgrid operation. Sustainable Cities and Society, 37(September 2017):26– 33, 2018. [14] Mohammad Ali Fotouhi Ghazvini, Jo˜ ao Soares, Omid Abrishambaf, Rui Castro, and Zita Vale. Demand response implementation in smart households. Energy and Buildings, 143:129– 148, 2017. [15] Sara Barja-Martinez, Ingrid Munne-Collado, Pau LloretGallego, Monica Aragues-Penalba, and Roberto VillafafilaRobles. A Novel Home Energy Management System Environmental-based with LCA Minimization. In 2020 IEEE International Conference on Environment and Electrical Engineering (EEEIC / I&CPS Europe), pages 1–6, Madrid, Spain, 2020. IEEE procedings. [16] Tong Xing, Hongyu Lin, Zhongfu Tan, and Liwei Ju. Coordinated energy management for micro energy systems considering carbon emissions using multi-objective optimization. Energies, 12(23), 2019. [17] Anzar Mahmood, Faisal Baig, Nabil Alrajeh, Umar Qasim, Zahoor Ali Khan, and Nadeem Javaid. An enhanced system architecture for optimized demand side management in smart grid. Applied Sciences, 6(5), 2016. [18] Xin Jin, Kyri Baker, Dane Christensen, and Steven Isley. Foresee: A user-centric home energy management system for energy efficiency and demand response. Applied Energy, 205(June):1583–1595, 2017. [19] Mohammad Shakeri, Mohsen Shayestegan, Hamza Abunima, S. M.Salim Reza, M. Akhtaruzzaman, A. R.M. Alamoud, Kamaruzzaman Sopian, and Nowshad Amin. An intelligent system architecture in home energy management systems (HEMS) for efficient demand response in smart grid. Energy and Buildings, 138:154–164, 2017. [20] Hieu T. Nguyen, Duong T. Nguyen, and Long B. Le. Energy management for households with solar assisted thermal load considering renewable energy and price uncertainty. IEEE Transactions on Smart Grid, 6(1):301–314, 2015. [21] Fei Wang, Biao Xiang, Kangping Li, Xinxin Ge, Hai Lu, Jingang Lai, and Payman Dehghanian. Smart households’ aggregated capacity forecasting for load aggregators under incentive-based demand response programs. IEEE Transactions on Industry Applications, 56(2):1086–1097, 2020. [22] Maria Luisa DI Silvestre, Pierluigi Gallo, Eleonora Riva Sanseverino, Giuseppe Sciume, and Gaetano Zizzo. Aggregation and Remuneration in Demand Response with a BlockchainBased Framework. IEEE Transactions on Industry Applications, 56(4):4248–4257, 2020. [23] Roberto Turconi, Alessio Boldrin, and Thomas Astrup. Life cycle assessment (LCA) of electricity generation technologies: Overview, comparability and limitations. Renewable and Sustainable Energy Reviews, 28:555–565, 2013. [24] Ingrid Munn´ e-Collado, Fabio Maria Apr` a, Pol Olivella-Rosell, and Roberto Villaf´ afila-Robles. The potential role of flexibility during peak hours on greenhouse gas emissions: A life cycle assessment of five targeted national electricity grid mixes. Energies, 12(23), 2019. [25] Eduardo Redondo-Iglesias, Pascal Venet, and Serge Pelissier. Modelling lithium-ion battery ageing in electric vehicle applications—calendar and cycling ageing combination effects. Batteries, 6(1):1–18, 2020. [26] Johannes Schmalstieg, Stefan K¨ abitz, Madeleine Ecker, and