Optimal midterm peak shaving cost in an electricity management system using behind customersÃ¢ÂÂ smart meter configuration

This paper analyses a local electricity system (LES) comprising photovoltaic production (PV), a connection to the distribution network, local loads and an energy storage system (ESS). Given the flexibility of the ESS, the LES can provide a peak shaving service (PSS) to the grid operator based on the actual monthly power tariff. This paper proposes a stochastic mixed-integer linear programming problem that maximises the expected operating profit of the LES midterm. Assuming a behind customers’ smart meter configuration, income is derived from selling the energy of prosumers to other external electrical areas. If the costs are higher than the income, the net profit will be negative, i.e. a net loss. The cost component of the objective function can be reduced through the management of local resources and by providing PSS to the distribution network operator to minimise the power cost of the monthly power tariff. The model is tested for 720 h (considering a month of 30 days) in three cases: (i) without PV and ESS; (ii) with PV and ESS, where losses are 0%; (iii) with PV and ESS, where losses are 18%. Due to the monthly power tariff, the net loss of the LES is reduced through the optimal management of local resources when the ESS losses are lower than 18%. To assess seasonal implications about the LES, the 12 months of the year are also tested. The month of October indicated the highest peak shaving, while the lowest peak shaving depended on the ESS losses.


Introduction
The electric power grid faces challenges related to the growing share of distributed energy resources (DERs (see the table of acronyms  Table 1)). As renewable energy sources increase their presence in distributed systems, energy storage systems (ESSs) and electric vehicles are rapidly penetrating markets. Hence, the impact of ESSs and electric vehicles on power grids can be lessened through local electricity management systems. Consumers and prosumers who choose to be flexible regarding their non-time-critical consumption production needs can help alleviate grid challenges. In particular, local energy consumers and prosumers can sell their flexible consumption/production loads to different interested actors within the power system. The flexible loads can be utilised for congestion management and grid operation, among others. In this way, local flexibility providers can make a profit on energy flexibility sales or reduce their electricity bills. Flexible consumption and production can be facilitated through well-designed

Literature review
Various options for managing energy resources locally are discussed in the literature. In particular, the focus has been on local markets, grid services and electricity management systems.

• Local markets
European regulation was analysed in [2] and enablers of local flexibility markets were identified. An optimisation problem (mixed-integer linear programming [MILP]) was proposed in [3] to satisfy the necessity for the required flexibility from distribution system operators. Furthermore, research in [4] presented the development of short-term decision support models for aggregators who sell electricity to prosumers and buy the excess of local generation. The two-stage stochastic mixed-integer linear programming problem (SMILP) allowed simulating the aggregator model over two months. The role of ESSs in local markets was addressed in [5]. The authors created an agent-based simulation study of local peer-topeer electricity trading in a community. The study reflected on two local market scenarios, i.e. with and without local community storage. The aim was to procure electricity directly from local renewable generation. Accordingly, [5] concluded an increased local market efficiency impact associated with the utilisation of community storage. In the study of [6], energy storage was used to reduce system congestion cost caused by the two peaks by sending cost-reflective economic signals, where a new charging and discharging strategy based on binary search method was defined. Hence, a locational marginal pricing method was created for energy storage. Although the work presented in [7] does not deal with local markets, two relevant (for the local markets' context) strategies of PV power producers without an ESS were presented, where the PV producers could participate in the day-ahead and intraday markets and the imbalance markets. The work presented by [8] envisioned a microgrid environment, in which consumers and prosumers trade at the local market to keep profits within the community. Within research on LEMs, studies have focused on different LEM configurations considering the distribution system operators (DSOs) and residential customers. As presented in [9], DSOs can be beneficiaries of contracts and services through local market structures. Furthermore, co-simulated distribution networks and peer-to-peer energy trading platforms were discussed in [9]. Residential customers were analysed in [10] via experimental analysis. Next, centralised and decentralised analysis of local markets were addressed in [11]. The centralised approach was used as a benchmark of [11], where the national market operator optimally scheduled the bids of generators. The decentralised approach was modelled using the Stackelberg leadership model, which, by clearing local market procedures can be solved in simultaneous or sequential modes. Peer-to-peer energy trading is currently a popular area of research. A review on peer-to-peer and community-based markets was introduced in [12]. Peer-to-peer trading for prosumers in a residential building in London was tested in [13] with a degree of system flexibility provided by battery storage. To manage the uncertainty, a peer-to-peer local electricity market for trading energy and uncertain power with demand having power flexibility was developed in [14], while a prediction-integration model was implemented in [15] for a prosumer in continuous double auction-based peer-to-peer. • Grid services and energy management systems The primary grid services are peak shaving, congestion management and frequency reserve. A techno-economic analysis was introduced in [16] for the peak shaving and frequency containment reserve that battery storage could provide in Norway. [16] conducted analyses considering balancing prices, grid charges, degradation of batteries and opportunity investment costs. Another approach for testing the ESS was adopted in the form of the robust optimisation of system congestion management [17]. An ESS in Madeira, Portugal, was presented in [18], where arbitrage, self-sufficiency, peak shaving and energy back up were co-optimised. The peak shaving service for grid operation was analysed in [19], which focused on vehicle-to-grid technology (V2G). A review of peak shaving was presented in [20], which focused on the importance of peak load shaving, peak shaving using ESS, peak shaving based on electric vehicles and peak shaving using demand-side management. A pivotal part of grid services is energy management, which is applied for DERs and loads. [21] studied DERs sharing in the context of peak shaving and load balance, where the aggregation of DERs was carried in energy and capacity markets and consumer contributions were evaluated through an energy sharing scheme. Other investigations have focused on specific modelling characteristics. Load smoothing and peak shaving based on load forecasting through artificial neural networks were analysed in [22]. A multi-agent system framework was applied in the peak shaving and valley filling potential of PV production with an ESS for a high-rise residential building [23].

Aims and contributions
This study provides a preliminary analysis of a new energy trading platform operated by a smart energy service provider (SESP) or a local system operator (LSO). In the core of the energy trade is a local energy system (LES) that includes PV production, an ESS, local load (LL) and a connection with the distribution network (NC) as shown in Fig. 1.   The E-REGIO LES is a facility located in the south of Norway. The actual installations of local resources are: stadium load, residential and commercial loads, rooftop PV panels in the stadium, an energy storage system (ESS), and the distribution network operator (Skagerak Energi AS). This paper proposes a mathematical model, stochastic mixed-integer linear problem (SMILP), to analyse the described LES. The analysis of the LES targets the energy management system of a centralised local market considering the ambitions of the E-REGIO project. In this context, electricity market proposed within E-REGIO is a local market concept for trading products and services such as peak shaving, local energy surplus, local renewable production and flexibility provided by the local ESS. The implications of the ESS are also tested regarding its efficiency, which is tested with a 0 and 18% of losses. The 18% of loss is associated with the actual loss of the ESS.
More specifically, this paper is focused on the smart energy management system of local resources of the E-REGIO pilot site in Norway. Hence, the smart energy management system is part of the business case for a local energy system behind the customer's smart meter (BCSM).
This model, together with the measured data, provides relevant information for practical future implementation of the E-REGIO market concept as a local smart grid and electricity market, in other European jurisdictions. Moreover, the analysis of peak shaving service requires data over a period of at least one month to resemble the actual power tariff of the distribution network. Hence, the time frame of the optimisation model was adjusted for 720 h (30 days multiplied by 24 h) as average month hours. This power limit tariff characteristic based on the actual Norwegian market serves as a clear contribution to the scheduling of local resources but also limits the SMILP problem. Accordingly, the contributions of this paper are threefold as noted below.
• A new smart energy management system, focused on managing local resources from an economic point of view. The main objective is to maximise the midterm net profits, while the midterm peak shaving service minimises the distribution network costs of the local energy system, thereby promoting a more independent LES. • A novel stochastic programming model that allows for evaluating the novel smart energy management system for the Norwegian E-REGIO pilot, following the actual tariff design. The proposed optimisation model provides relevant feedback for future business models of the E-REGIO LEM. • Realistic simulations of monthly operations for one full year. This annual analysis captures the stochastic nature of demand, PV generation and wholesale market prices. By capturing seasonal variations, the study provides pivotal information for the practical implementation of the E-REGIO LEM in the current and future Norwegian context.

Paper organisation
The remainder of this paper is structured as follows. The assumptions are introduced in Section 2. Section 3 describes mathematical formulation, while Section 4 presents data and results of the pilot study. Section 5 presents the main conclusions from the results provided in Section 4.

Assumptions
Several assumptions and considerations were included in the study as noted below.
• Any month of the year comprising 30 days was considered; this provided for 30 days multiplied by 24 h (720 h), which allowed for the comparison between months. • Power (kW) was considered energy per hour kWh/h. Moreover, the time step resolution of simulations was one hour, indicating the importance of kW and kWh/h. • The NC did not have power congestion problems, although the model included the energy per hour flow limit (presented by kWh per hour). • All loads considered in the paper were local loads, i.e. the local consumption that happens in the LES.
• Midterm optimisation (optimisation for a time frame of 720 h (one month) with hourly time step resolution) presented difficulty for formulating scenarios. Hence, months from the preceding six years were selected as scenarios classified according to month; e.g. January group included all the January months in the six preceding years. • The actual ESS showed losses equal to 18% and was compared with a case without losses. • The PV and ESS power capacity are not included in the model because the facilities are predefined by pilot set-up. • In the text, LL is used as equivalent to LOAD and NC to NET; LL and NC are used in the text, while LOAD and NET are used in the mathematical model.
Finally, with reference to the Norwegian pilot, the envisioned energy flows are illustrated in Fig. 2. The energy flows are as follows: (i) the ESS can inject energy into the LL and NC; the ESS can also charge energy from the PV producer and NC; (ii) the LL energy is only consumed by participants in the form of residential and commercial loads; (iii) the PV can send energy to the ESS, LL and NC; (iv) the NC allows energy to be injected into the LES and the distribution system.

Mathematical formulation
In this section, a novel optimisation model based on two-stage stochastic programming is introduced to determine the optimal profile of local resources considering midterm peak shaving cost.
The first-stage decision variable set represents here-and-now decisions. The first-stage variables represent the first interest. The secondstage decision variable set includes the wait-and-see decisions, which drive recourse decisions to a specific realisation regarding the random variables. This mathematical formulation follows a model formulated as node-variable; therefore, the formulation does not need nonanticipativity constraints. The first-stage decision variables do not have a scenario index s, while the second-stage decision variables depend on the scenario index s. The nomenclature used in the mathematical model is presented in Table 2. Fig. 3 shows the objective function of the optimisation problem, which maximises the profits, subject to seven block constraints.
The problematic constraints of this mathematical model are divided into seven blocks: expected profit, PV energy flows, ESS energy flows and SOE, LL energy balance, NC energy flow, peak shaving and energy flow directions.
The mathematical model needs to satisfy all the constraints to be feasible. The development of the mathematical model follows the energy flow concept, as shown in Fig. 4.
The interactions between elements follow the diagram of Fig. 4. The PV can inject energy into the LL, ESS, NC, and shed, while the ESS can inject energy into the LL and NC. The NC can inject into the LL and ESS. These interactions between the PV, ESS and NC will depend on the costs and incomes, trying to maximise the profit of the LES. There are two important decisions through the binary variables. These are charge or discharge the ESS through the binary variable and inject energy into the LES or NC through the , binary variable.

Objective function
The objective function is presented in (1), which maximises the total expected profit (2); hence, the total expected profit may be positive (net profit) and negative (net loss). The total expected profit comprises income and costs. The LES only has an income when it sells the local production to the wholesale market (WM) through the NC as a prosumer. Local production can be derived from PV production and ESS. However, costs are derived from energy injected into the LES, which is used by the LL and ESS. This energy injected by the NC into the LES has an equivalent cost to the prices of the WM. Due to lower A.A.S. de la Nieta et al. PV production compared with the LL and energy from the NC charged in the ESS, the total expected profit will be a net loss (costs are higher than the income). The objective function then maximises the income and minimises the costs, subject to all the constraints.

Expected profit constraints
The total operating profits (2) are defined as the expected profit minus the power cost for the time frame of one month.
The expected profits per scenario and period (3) are prosumer incomes minus the costs associated with the energy. Then, the prosumer incomes (4) are the wholesale prices plus the tariff as prosumer multiplied by the energy injected into the WM through the connection. This injected energy represents the energy flows from the PV and ESS. The energy flows of the LES are shown in Figs. 2 and 4. Regarding the costs of (5), these include costs associated with the WM and PV production shed costs. The costs associated with the WM (6) are the wholesale prices plus the grid tariff of the NC multiplied by the energy flow that goes from the NC to the ESS. Furthermore, a similar cost is set for the LL, with the wholesale prices and grid tariff of the NC multiplied by the LL as the second part of (6). A PV cost (7) is considered when it is shed. The LES uses the PV shed mechanism only when it is required for effecting an energy balance. The PV shed cost is the marginal cost associated with the PV shed.
Eq. (8) defines the power cost. This equation represents different blocks of power with different tariffs. Once the first energy block of maximum energy per hour is used, the second block starts being used. The first energy block has a higher cost than the second energy block. The formulation of this cost has two parts. The block cost is the block tariff multiplied by the binary variable. This variable shows whether the block was used and the total energy of the block. The second part of (8) is the cost of the last block activated, where part of the energy block was used. Then, the second part of (8) is the block power tariff multiplied by the energy used in that block. See Fig. 5.

Photovoltaic energy flows constraints
Eq. (9) evaluates the energy flow that goes from the PV to the NC, LL, ESS and PV shed. The summation of all PV energy flows is equal to the scenarios of PV production, being an uncertain parameter.

Energy storage system energy flows and state of energy constraints
Another type of energy flow is ESS energy flow (10), which sees energy from the ESS sent to the LL and NC. The summation of flows from the ESS to the NC and LL must be equal to or lower than the state of energy (SOE) in that period.
Eq. (11) is the SOE . It is equal to the SOE in period − 1 plus the energy charged in the ESS from the PV and NC. When the ESS discharges, this energy is injected into the NC and/or LL [24]. Moreover, this SOE considers losses, which are defined in (12) as a percentage of the SOE. Finally, this SOE has upper and lower limits (13).

Local load energy balance constraints
Constraint (14)

Network connection energy flow constraints
The network connection energy flow constraints, (15) and (16), show the energy flows and total energy associated with the NC, respectively. To avoid problems with the LL, the NC provides energy mitigating the uncertainty of the LES. Then, the energy injected into the LES from the NC as shown in Fig. 4 is decided with the , , and , variables, which depend on the scenario, index , for mitigating the uncertainty of the LES. Those two variables will define the maximum energy per hour (kWh/h) injected into the LES, being the . It is considered as the peak shaving concept because the cost of the power related to the tariff will reduce the kWh/h injected into the LES from the NC. The constraints of this section are related to quantifying the energy from the NC.

Peak shaving constraints
Peak shaving is pivotal to the current study. The equations consider the peak limit that is required to satisfy the power needed in the LES. The peak shaving is the difference between the scenarios of LL and the limit of peak shaving that satisfies all scenarios in the highest LL of one period with the conditions of PV and ESS. Fig. 5 shows how the power tariff is modelled and the meaning of the parameters and variables.
The peak limit Eqs. (17)- (19) define the blocks of power that the power tariffs are divided into. Moreover, these equations evaluate the block where the total energy injected into the LES from the NC occurs. We developed these block tariffs following the definition of blocks as in [25], with selected differences. Eq. (17) limits the power of the block to the maximum power of the block. The binary variable is activated only in one block; then (18) fixes the summation of the binary variable per block to 1. However, the previous blocks of the active block are considered for the power cost. Eq. (19) is needed to derive the power and cost of used blocks. This equation evaluates the power blocks that were activated. Then, the summation of the previous active power blocks known through the binary variable is equal to the activated power block multiplied by the summation of power blocks . Eq. (19) determines the binary variable used in (8) for the . The total power is evaluated in (20); it is the summation per block of the power required of the active block plus the summation of previous power blocks.

Energy flow direction through binary variable constraints
Eqs.

Data
The E-REGIO LES is a real facility located in the south of Norway. Herein, we introduce all the parameters needed for our simulations, representing data from the actual installations, the applicable distribution network operator (Skagerak Energi AS) tariffs and wholesale market prices from Nord Pool [26].
There are two types of parameters, i.e. certain and uncertain. The costs considered in the model are = = e0,0205/kWh from September to April and e0,0186/kWh for the remaining months of the year. To avoid problems when there is an excess of PV production that cannot be injected into the network, the PV production can be shed with a cost = e0,05/kWh. There is also a power cost , which is associated with peak shaving.
is a power tariff per blocks. Then, the limit of the network connection is =1 = = 3000 kWh/h. As such, there are only two blocks as it is shown in Fig. 6. A tariff for the income as a prosumer is equal to e0,0186/kWh.
The ESS has different parameters. Its power limit is 800 kW, it is = 800 kWh per hour and the ESS loss is = 0,18, meaning that it has 18% of losses or a round trip efficiency (RTE) of 82%. Hence, two cases were analysed for the ESS. These two cases involved PV production and an ESS with a loss of 18% (real losses of the ESS) and with PV production and an ESS with a loss of 0%. The maximum and minimum limits of the SOE of the ESS are = 1000 kWh and = 0,2⋅ kWh, which is 20% of the maximum SOE. We consider a = 1, which is 100% efficiency when the energy is charged or discharged in the ESS.
The uncertain parameters are the , , , , and , . The uncertainty of these parameters is considered using scenarios. Due to the time frame needed for deriving the power cost of 720 h, the scenarios were created using real data for each month of previous years. Then, the data from the preceding six years (from 2013 to 2018) of the LL and WM prices [26] were used as scenarios. However, the PV panel facilities were only recently installed (PV power capacity of 800 kWp) in Norway. Thus, we could only use simulations that included selected measures of the location. The scenarios for PV production were created using this information. Based on these simulations, a yearly simulation hour per hour was found for all 8760 h. Hence, each hour per month (from 1 to 24 h) was adjusted to a truncated normal distribution. Following on, simulations per hour of each month were run. As previously noted, six PV scenarios were selected. Due to using the monthly data of the six preceding years, the number of scenarios per uncertain parameter was decided as six. To illustrate these three    The three uncertain parameters were then combined in a tree. Hence, a tree was used to create the scenarios. The scenario tree, similar to [27], is composed of i) 6 scenarios of the LL, ii) 6 scenarios of the PV production and iii) 6 scenarios of the WM prices. This provided a total number of scenarios equal to 6 ⋅ 6 ⋅ 6 = 6 3 = 216. If there were 216 total scenarios, the probability per scenarios as equiprobable was = 0,0046.

Results
This Section 4.2 has two parts. The first introduces one month (July 2019) and the second the annual results of three cases.  To establish an analysis with a clear comparison of the results, the simulations of a reference case were conducted. This reference case, case 1, was the LES without PV production and an ESS. This reference case allowed for a comparison with the PV production and ESS cases. The ESS can have losses of 0% (case 2) and 18% (case 3). The case 3 is presented because the actual loss of the ESS is 18%. The case 2 and case 3 are equivalent to 100% of RTE and 82% of RTE, respectively.
The leading case study is the case 3, but the case 1 and case 2 allow comparing the actual case (case 3) with the reference of the previous situation to install the PV and the ESS (case 1) and if the ESS is perfect (case 2).

Case 1: LES without PV production and an ESS July 2019
This case 1 is how the LES has been working so far. Fig. 10 shows the maximum of energy imported, without peak shaving because there are not PV production and an ESS. Fig. 10 presents a maximum of total Net that will be the maximum kWh/h per month in the hour 209, equal to 1.666 kWh/h. Moreover, the maximum load is equal to the maximum of the total Net. Hence, the power cost is 1.666 kWh/h multiplied by e5,54/kW is equal to e9229,64.

Case 2: LES with PV production and an ESS (0% of loss) July 2019
This case 2 has three peaks in Fig. 11, in the 136, 211 and 640 h. The average of the total Net is more volatile than in case 1 because the PV and ESS change the profile of the energy imported from the NC.
A maximum of power was required to satisfy the peak consumption in 136, 209 and 640 h, where the maximum LL in the 209 h could be 1.666 kWh/h and the energy imported from the NC could be 1.467 kWh/h. Hence, peak shaving reduced the power in July by 199 kWh/h and multiplied by e5,54/kW, and this peak shaving reduced the power cost by e1.102.
Due to the ESS having loss of 0%, the maximum Net to the ESS in Fig. 12 has only peaks because the energy is only used to charge the ESS and does not need to compensate the loss of the ESS.
The PV production does not need to compensate the loss of the ESS, as shown in Fig. 13. The total PV profile follows the profile of the PV scenarios presented in Fig. 8.

Case 3: LES with PV production and an ESS (18% of loss) July 2019
The main results presented are related to the energy imported into the LES, maximum power kWh/h per month, LL, energy imported to charge the ESS and satisfy the LL and the PV production to be injected into the ESS and LL. Fig. 14 presents the behaviour of the model in July, 2019. The maximum kWh/h in July (720 h) was 1.467 kWh/h. This maximum was required to satisfy the peak consumption in hour 209, where the maximum LL in this hour could be 1.666 kWh. Hence, peak shaving reduced the power in July by 199 kWh/h. Multiplied by e5,54/kW, this peak shaving reduced the power cost by e1.102.
Additionally, the average of the total net (injected energy into the LES from the distribution network) was lower for the peak periods and higher for the off-peak periods. During this time, the injected energy considering the PV and ESS was flatter than without the PV and ESS. Fig. 15 shows the energy flows from the NC that can go to the LL and ESS. As shown in Fig. 15, the energy flow from the NC to the LL had a similar maximum values profile to the maximum total energy injected from the NC to the LES. The main differences arose from the peak and off-peak periods as illustrated in Fig. 14.
We also observed an energy flow from the NC to the ESS when there was no PV production as shown in Fig. 15. This energy flow happened to compensate for the losses of the ESS and the minimum SOE of the ESS as 20% of the maximum SOE. There were some peaks of the energy flow from the PV to the ESS that mitigated selected NC peaks. Moreover, the energy flow from the PV to the LL was higher Table 3 The profits, maximum kWh/h and PS of the three cases; case 1: without PV production and an ESS; case 2: with PV production and an ESS with losses = 0%; case 3: with PV production and an ESS with losses = 18%.   than to the ESS as shown in Fig. 16. This happened because the energy flow from the PV to the LL, unlike the ESS, did not experience losses. The operation of the LES with LL, PV production, ESS and a connection to the distribution network had a profit equal to e−21.881 and the maximum kWh/h was 1.467 kWh/h in July. Here, the profits were negative (net loss); this infers a cost that arises because of the energy imported from the NC and PV shed.

Annual results with and without energy storage system losses
Comparisons between all the months of the entire year, profits, maximum kWh/h and the peak shaving in the three cases are presented in Table 3.
During the summer months, the negative profit (net loss) were lower compared to the rest of the year. The summer months had lower consumption and higher PV production; however, some summer months also showed a higher maximum kWh/h use. This higher maximum also occurred in the spring months.
From the total annual results, the highest profits (the lowest net loss) were detected in the presence of PV production and an ESS with losses = 0%, where peak shaving was 3.008 kWh/h for the entire year. However, higher losses reduced profits (higher net loss) and peak shaving. Reduced profits infer higher costs; consequently, the LES with PV production and an ESS with losses = 18% increased the total costs of the system. Furthermore, the peak shaving was lower, from 3.008 kWh/h to 2.766 kWh/h. These higher costs were derived from the ESS, which had a minimum level of energy for the entire day, i.e. the minimum SOE constraint. The lowest negative profit (net loss) in all cases was observed for August, while the lowest max. kWh/h occurred in December. Following this trend, the highest peak shaving happened in October. However, the lowest peak shaving occurred in different months, depending on the losses incurred. For a perfect ESS (losses = 0%), this was the case in June, while with losses of 18% this was in January.
The total energy injected into the LES from the distribution network is presented in Fig. 17. The lowest annual energy imported happened when the loss of the ESS was 0%. When the ESS had losses equal to 18%, more energy was imported than in cases without PV and ESS during the months of autumn and winter, while the energy imported was lower in summer and spring periods as shown in Fig. 17. The total energy imported for the three cases without PV and ESS was 3.658.657 kWh; for PV and ESS with losses = 0% this was 3.303.969 kWh and for cases with PV and ESS with losses = 18% total energy was 3.634.560 kWh. Notwithstanding the higher energy imported in the autumn and winter periods, the case with PV and ESS with losses = 18% had the lower total annual energy imported than the case without PV and ESS.

Conclusions
A two-stage stochastic programming model was presented in this paper. The optimisation model scheduled the resources of a local energy system, such as PV production and an ESS behind a connection with a distribution network. Moreover, this paper considered the minimisation of the power-dependent connection cost for which a peak shaving service was provided. Hence, our conclusions are as follows.
• Local resources introduce more flexibility in the LES. This flexibility supports higher independence from the distribution network. An LES is a clear example of a new business approach for companies interested in managing local resources and providing services to the grid operator. • The efficiency of the ESS is pivotal for providing flexibility and improving the profitability of the business case. • The peak shaving service is not sufficient for achieving a profitable PV-ESS system based on the current pricing and regulatory framework. Hence, more research is needed to enable the provisioning of other ESS services. • The PV local resource is the most beneficial local resource for the LES because the PV power produces local and low-cost electricity.
The oversized ESS capacity also allows to install more PV power capacity and provide more services, as needed by the distribution grid.
As the ESS has been installed with the purpose of future research projects, future work will be devoted to multi-ownership of local resources in LES and new configurations different to BCSM. An interesting extension will be to mix multi-ownership, new configurations, multi ESS services, and mid-term forecasting time series.

Declaration of competing interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.