Enhancing Renewable Energy Integration and Implementing EV Charging Stations for Sustainable Electricity in Crete’s Supermarket Chain

Abstract

In current times, sustainability is paramount, and businesses are increasingly adopting renewable energy sources (RESs) and electric vehicle (EV) charging infrastructure to minimise their environmental impact and operational costs. Such a transition can prove challenging to multi-location businesses since each chain store functions under different constraints; therefore, the implementation of a corporate policy requires adaptations. The increased electricity demand associated with EV charging stations and their installation cost could prove to be a significant financial burden. Therefore, this study aims to investigate and develop strategies for effectively incorporating RES and EV charging stations into the operations of a supermarket chain in Crete. Monthly electricity consumption data, parking availability, and premise dimensions were collected for 20 supermarkets under the same brand. To achieve a more tailored approach to custom energy system sizing, the integration of energy storage coupled with a photovoltaic (PV) system was investigated, using the Moth–Flame Optimiser (MFO) to maximise the Net Present Value (NPV) of 20 years. The algorithm managed to locate optimal solutions that yield profitable installations for all supermarkets by installing the necessary number of PV units. Manual exploration around the solutions led to the optimal integration of energy storage systems with a total upfront cost of EUR 856,477.00 and a total profit for the entire brand equal to EUR 6,426,355.14.

1. Introduction

The predominant theme in recent decades has undeniably been the emphasis on sustainable development as the strategic solution to global environmental deterioration [1]. In recent years, in the context of global environmental issues and within the European Union energy and climate targets framework, there has been an increased interest in RES and sustainable transportation systems [2]. Particular interest is also reported for the use of EVs as a more environmentally friendly means of transportation [3], as the penetration of RES in local energy communities, such as commercial retail stores, and the installation of EV charging stations can offer multiple benefits [4]. On the one hand, the input of RES can lead to the achievement of environmental targets by reducing harmful emissions, minimising the dependence on fossil fuels, stabilising electricity costs, and creating a sustainable business scheme for retail stores. On the other hand, the installation of EV charging stations can attract more customers to the store, permitting browsing on the premises while their vehicle’s battery is being charged and hence are more likely to use the retail store facilities. This increased level of activity and the potential synergies between the charging station business and supermarkets could provide valuable revenue without being a source of profit alone [5]. Consequently, with their widespread presence and significant energy demands, supermarket chains present a prime opportunity to lead this change.

Additionally, as the world is motivated by increasing grid parity and decreasing manufacturing costs of PV components, regulatory incentives and legislative control agencies are increasing efforts to incorporate residential, industrial, and grid-level systems. Energy storage (ES), whether on-grid or off-grid decentralised applications, is increasingly integrated into systems, as excessive energy consumption can be significantly reduced, thus contributing to greater sustainability [6]. The residential and industrial sectors represent a clear trend in the installation of grid-connected systems and the integration of ES. Because of the temporal separation between generation and power demand, daytime surplus energy produced by systems can be shared with other areas [7,8]. Also, integrating ES with PV energy management can protect against the sudden fluctuations in solar energy [9], maximise the use of solar power [10], improve smoother power transmission [11], and increase the economic and technical benefits of distributed generation systems [12]. ES and its use can also be managed in cases of emergency, contributing to the resilience of the system. These could be easily evaluated regarding their Technology Readiness Levels (TRLs) through appropriate methods [13], thus verifying their applicability and effectiveness.

The ability to store energy at times of sufficient solar irradiance and then to dispatch the power as needed offers significant opportunities for grid operators to manage increasing generation peaks, reduce the costs of other ES methods [14] and carbon emissions [15], reduce the need for peaking power plants, and increase demand response [16,17]. Off-grid systems with ES guarantee continuous power for temporal imbalances between generation and consumption throughout every month of the year. Moreover, incorporating ES in systems has the potential to mitigate the impacts of market instability and volatility in the long term [18]. In addition, with the implementation of net metering policies, the large amount of energy exchanged between customers and the grid will undoubtedly lead to overvoltage and power quality problems. Systems and ES will enable voltage regulation, as well as grid support services, with subsequent advancements in control and communication technologies and the transition to advanced distribution management systems [19,20]. Within this context, advancements in battery technologies are revolutionising ES [21].

Soon, the energy obtained from traditional sources will decrease. Nonetheless, incorporating RES into large complexes [22], such as commercial or public buildings [23], involves technological integration [24,25], economic implications, and operational strategies [26]. Thus, renewable energy technologies, such as PVs, are expected to maintain their increasing trend, due to their abilities [27]. However, common solar resource problems, such as cloudy and rainy days or inclement weather, have attracted worldwide attention. In order to prepare for the stochastic production originating from natural sunlight, storage systems are used for electric energy through an energy management system, so that solar PV systems can transmit power to the grid. Thereupon, batteries are used for the storage of generated energy to be used in peak periods, thus improving energy conservation, providing emergency energy requirements, improving energy management, scheduling, and balancing power fluctuations. With the growth of air conditioning systems and the price of solar cells declining, the installation of PV systems with ES is expected to grow rapidly in the next few years [28].

Although the benefits of such investment projects are considerable, the selection of the optimal renewable energy investment size and quality, as well as the optimal location of charging stations, depends significantly on the existing infrastructure, energy network configuration, and business standards of the commercial retail store [29]. According to recent research, the average annual energy consumption of buildings in Greece varies by sector: 406.8 kWh/m² in hospitals, 273 kWh/m² in hotels, 187 kWh/m² in offices, 152 kWh/m² in commercial buildings, and 93 kWh/m² in schools [30]. This energy use encompasses a wide range of activities, including heating, cooling, lighting, and powering various systems within residential, commercial, and industrial buildings.

Within this context, supermarket chains, which consume a significant amount of energy daily, possess a distinct chance to pioneer sustainable practices by embracing RES and establishing EV charging stations. Combining demand management with solar PV systems can significantly mitigate the impact of EV charging on a retail building’s peak demand and energy consumption, reducing peak demand by up to 38% [31]. Hence, the investment of PV panels to fulfil an important fraction of the supermarket’s actual annual electricity consumption is worth considering. Subsequently, under the hypothesis that some segments of the supermarket parking area are used as potential charging station locations to serve the customers during their stay, the combination of PV with battery energy storage system (BESS) units can also be implemented. By incorporating RES into supermarkets’ energy infrastructure, their dependency on non-renewable energy sources can be greatly mitigated, which will minimise emissions and energy expenses [32]. The scalable installation of solar and wind energy systems on expansive rooftops and parking lots further enhances the potential of supermarkets to serve as exemplary pilot cases for the energy transition. According to the literature, supermarkets can use on-site renewable generation to offset a sizable amount of their energy requirements [33]. Supermarkets may run on renewable energy even during non-solar hours by using battery storage systems, which also guarantee a steady energy supply by storing extra energy produced during peak production hours. Nowadays, supermarkets increase their capacity of customer convenience and promote sustainable energy practices due to the rise of EVs. As mentioned above, installing EV charging stations can accomplish two goals: attracting eco-aware clients and becoming a hub for renewable energy [34,35]. According to researchers, higher in-store sales may result from longer customer-dwelling durations at sites with charging stations available [36]. Additionally, the need for easily accessible charging infrastructure is increasing along with the introduction of EVs, giving retailers a useful strategy to stand out in competitive markets.

Notwithstanding the obvious advantages, supermarkets can face problems with the upfront costs associated with installing RES and EV charging stations [37]. However, government subsidies, incentives for the adoption of renewable energy, and collaborations with energy corporations can lessen these financial constraints. As grid integration and renewable ES technologies advance, supermarkets could develop into micro-energy centres that dispatch excess electricity back into the grid, further advancing sustainable development objectives.

The contribution of metaheuristic algorithms in optimising such complex systems has been extensive. Anila Yasmeen et al. [38] have deployed the Firefly Algorithm to minimise the operating cost of the supermarket under study, investigating distributed energy generation under standalone and grid-connected modes. Nickyar Ghadirinejad et al. [39] conducted a similar study for a supermarket located in Sweden, using the Particle Swarm Optimiser (PSO). Pavitra Sharma et al. [40] optimised a simulation of an energy system of a supermarket in India, focusing heavily on BESS introduction, relying on the PSO as an optimisation method. Incorporating RES and EV charging stations necessitates a combination of strategic planning, technology investments, and stakeholder involvement. Qingyuan Yan et al. [41] address such challenges by incorporating a pelican algorithm–XGBoost hybrid to predict PV production, and PSO to optimise the dispatch in a district area. Jun Yamasaki et al. [42] optimise the installation of a PV and battery storage system into a retail store with integrated EV quick chargers, aiming to maximise the effect of peak load cut, minimising the associated cost at the same time through the use of a genetic algorithm.

Within this context, this study’s initiative not only aligns with the current environmental directives but also offers economic and branding benefits to the retail sector. Additionally, the economic metrics and business environment characteristics employed in this research could also be helpful for similar business enterprises that would like to exploit another potential source of profit for the local economy through an energy synergy with electric vehicles’ owners or users. Specifically, this study aims to optimise the configuration of a PV-BESS system in commercial settings, by executing hourly energy profiling and evaluating economic profitability through the calculation of the investment’s NPV. Based on the need for effective energy management, an integrated optimisation model for distributed energy systems in supermarkets is presented. The deployment of the MFO aims to maximise the NPV, to outline the optimisation of both the capacity and operation of PV and battery systems. A case study of a major supermarket chain in the island of Crete is presented identifying which technologies can minimise total energy costs compared to conventional system setups. The methodology relies on the store’s monthly energy bills and the components’ properties, considering the current Greek legislature concerning charges and tariffs. Distinct from previous studies, which often focus on isolated components or single-site implementations, the inclusion of a sample of retail stores varying in structural properties, traffic, and weather conditions under the same brand, our approach holistically optimises energy systems under real-world operational and regulatory constraints. By prioritising economic feasibility through long-term analysis and addressing space limitations, regulatory compliance, and demand variability, this research offers a scalable and practical model for commercial energy system integration in regional markets.

2. Materials and Methods

To calculate the optimal set of PV and BESS units for each store, the simulation of the energy system is implemented. For the simulation, the hourly estimation of the electricity load and photovoltaic production is required, accompanied by the hourly PV production on an annual basis, yet the store brand’s limited availability of monthly consumption data and the related costs for each month of 2022 calls for an estimation of the required values. After estimating the initial hourly load, a normalisation procedure is applied using reference data to refine the load profiles. The estimated EV charging demand is then incorporated, completing the comprehensive load profile for each supermarket.

The characteristics of the PV and BESS systems are defined as parameters within the optimisation framework. The problem is solved using the MFO, which iteratively searches for the optimal configuration. During this process, financial metrics (NPV) are calculated to assess the performance of each potential solution.

Upon completion of the final iteration, the optimal solution is selected to guide informed decision-making regarding the installation of PV and BESS systems at each store location.

Figure 1 presents a visual summary of the methodology’s key steps.

2.1. Estimation of Hourly Consumption

As a first step of the optimisation process, the hourly consumption of electrical energy was conducted using the recorded monthly values provided by the company. Based on the provided data, Figure 2 showcases the consumption levels of four supermarkets located separately within the island, operating under different weather conditions for all months of 2022, along with their related costs.

As expected, the highest energy consumption was recorded in summer, particularly in August, with increased cooling demands, except for the store located in Tympaki. Monthly costs tend to correlate with the monthly consumption, reaching thousands of euros in operational expenses. The four regions are spread across the island, with solar irradiance distributed as seen in the map [43] of Figure 3.

This section’s methodology is heavily influenced by [44]. The authors adopted the Procida City Hall in Italy as a case study. Both public buildings like Procida Hall and supermarkets exhibited distinct energy consumption peaks during business hours, with relatively lower usage during off-peak hours. This similarity allowed for a reasonable approximation of energy demand patterns for retail stores operating in Crete. This specific case study was also based on the lack of detailed hourly energy consumption data from the supermarkets under study, necessitating the use of an alternative dataset. Procida Hall’s data, available from the existing literature, provided a reliable basis for estimating hourly profiles. Furthermore, Procida Hall’s energy consumption patterns reflect fluctuations between workdays, weekends, and holidays, analogous to variations observed in commercial retail settings. The study accounted for realistic daily and seasonal demand changes by leveraging these variations.

The study divided each day into workdays, weekends, and holidays. Workday energy consumption was separated into three periods: peak (F1) from 08:00 to 18:00, mid-level (F2) at 07:00 and 19:00 to 22:00 on workdays and 7:00 to 22:00 on Saturdays, and off-peak (F3) from 00:00 to 6:00 and 23:00 to 24:00 on workdays and Saturdays, and for all 24 h of Sundays and holidays. The reference utilised the hourly power consumption for each category based on the recorded energy data for each month. Due to the lack of similar data, the reference’s records of energy consumption for each period F1 (${\mathrm{E}}_{\mathrm{F}1}$), F2 (${\mathrm{E}}_{\mathrm{F}2}$), and F3 (${\mathrm{E}}_{\mathrm{F}3}$) of each month were used to adapt Procida Hall’s data to the Cretan supermarkets. A normalisation process was employed, ensuring the hourly profiles derived reflected the total monthly energy consumption recorded for each store. This adaptation ensured the data were contextually appropriate despite geographic and operational differences. Then, the power consumption for each month’s category was calculated using (1)–(3):

$${\mathrm{P}}_{{\mathrm{F}1}_{\mathrm{m}}}={⌊{\displaystyle \frac{{\mathrm{E}}_{\mathrm{F}1}}{{\mathrm{t}}_{\mathrm{F}1}}}⌋}_{\mathrm{m}}$$

(1)

$${\mathrm{P}}_{{\mathrm{F}2}_{\mathrm{m}}}={⌊{\displaystyle \frac{{\mathrm{E}}_{\mathrm{F}2}}{{\mathrm{t}}_{\mathrm{F}2}}}⌋}_{\mathrm{m}}$$

(2)

$${\mathrm{P}}_{{\mathrm{F}3}_{\mathrm{m}}}={⌊{\displaystyle \frac{{\mathrm{E}}_{\mathrm{F}3}}{{\mathrm{t}}_{\mathrm{F}3}}}⌋}_{\mathrm{m}}$$

(3)

The symbols ${\mathrm{t}}_{\mathrm{F}1}$, ${\mathrm{t}}_{\mathrm{F}2}$, and ${\mathrm{t}}_{\mathrm{F}3}$ symbolise the number of hours for each time period during the month $\mathrm{m}$. In Figure 4, the resulting time series is depicted where the division between the F1, F2, and F3 periods is clear.

To make further adjustments for a more accurate load profile, [45] was used along with the representative normalised known load profile (KP) of a medium retail store, as depicted in Figure 5.

Each month’s hourly KP value (${\mathrm{Hourly}}_{\mathrm{KP}\left(\mathrm{i}\right)}$) was calculated by multiplying the monthly maximum value by the seasonal hourly load factor. As a final normalisation step, the mean consumption value for the F1 period for each season was calculated using (4):

$${\mathrm{M}}_{\mathrm{F}1,\mathrm{WD}}={\displaystyle \frac{\sum _{08:00}^{18:00}{\mathrm{Hourly}}_{\mathrm{KP}\left(\mathrm{i}\right)}}{{\mathrm{n}}_{\mathrm{t},\mathrm{F}1}}}$$

(4)

with ${\mathrm{n}}_{\mathrm{t},\mathrm{F}1}$ symbolising the number of hours in the F1 period. The proportionality between ${\mathrm{M}}_{\mathrm{F}1,\mathrm{W}\mathrm{D}}$ and ${\mathrm{P}}_{{\mathrm{F}1}_{\mathrm{m}}}$, which have different heights for the same bases, was transferred from KP to the study cases for each month $\mathrm{m}$, using (5):

$${\mathrm{K}}_{\mathrm{F}1,\mathrm{m}}={\displaystyle \frac{{\mathrm{P}}_{\mathrm{F}1,\mathrm{m}}}{{\mathrm{M}}_{\mathrm{F}1,\mathrm{WD}}}}$$

(5)

And so, the estimated hourly value ${\mathrm{H}\mathrm{o}\mathrm{u}\mathrm{r}\mathrm{l}\mathrm{y}}_{\mathrm{S}\mathrm{C}\left(\mathrm{i}\right)}$ was calculated using (6):

$${\mathrm{Hourly}}_{\mathrm{SC}\left(\mathrm{i}\right)}={\mathrm{K}}_{\mathrm{F}1,\mathrm{m}}{·\mathrm{Hourly}}_{\mathrm{KP}\left(\mathrm{i}\right)}$$

(6)

Similarly for the other two categories, the mean consumption value was calculated for workday F2 (${\mathrm{M}}_{\mathrm{F}2,\mathrm{WD}}$) and Saturday F2 (${\mathrm{M}}_{\mathrm{F}2,\mathrm{ST}}$), along with their respective coefficient values ${\mathrm{K}}_{\mathrm{F}1,\mathrm{m}}$ and ${\mathrm{K}}_{\mathrm{F}2, \mathrm{ST},\mathrm{m}}$:

$${\mathrm{M}}_{\mathrm{F}2,\mathrm{WD}}={\displaystyle \frac{{\mathrm{Hourly}}_{\mathrm{KP},07:00}+\sum _{19:00}^{22:00}{\mathrm{Hourly}}_{\mathrm{KP}\left(\mathrm{i}\right)}}{{\mathrm{n}}_{\mathrm{t},\mathrm{F}2}}}$$

(7)

$${\mathrm{K}}_{\mathrm{F}1,\mathrm{m}}={\displaystyle \frac{{\mathrm{P}}_{\mathrm{F}1,\mathrm{m}}}{{\mathrm{M}}_{\mathrm{F}1,\mathrm{WD}}}}$$

(8)

$${\mathrm{M}}_{\mathrm{F}2,\mathrm{ST}}={\displaystyle \frac{\sum _{07:00}^{22:00}{\mathrm{Hourly}}_{\mathrm{KP}\left(\mathrm{i}\right)}}{{\mathrm{n}}_{\mathrm{t},\mathrm{F}2}}}$$

(9)

$${\mathrm{K}}_{\mathrm{F}2, \mathrm{ST},\mathrm{m}}={\displaystyle \frac{{\mathrm{P}}_{\mathrm{F}2,\mathrm{m}}}{{\mathrm{M}}_{\mathrm{F}2,\mathrm{ST}}}}$$

(10)

The F3 intervals concerning mean consumption on Saturdays (${\mathrm{M}}_{\mathrm{F}3,\mathrm{ST}}$), Sundays and holidays (${\mathrm{M}}_{\mathrm{F}3,\mathrm{SN}}$), and workdays between 23:00 and 07:00 (${\mathrm{M}}_{\mathrm{F}3,\mathrm{WD}}$) with respective normalisation coefficients ${\mathrm{K}}_{\mathrm{F}3, \mathrm{ST},\mathrm{m}}$, ${\mathrm{K}}_{\mathrm{F}3 \mathrm{SN},\mathrm{m}}$, and ${\mathrm{K}}_{\mathrm{F}3 \mathrm{WD},\mathrm{m}}$ for each month m are as follows:

$${\mathrm{M}}_{\mathrm{F}3,\mathrm{WD}}={\displaystyle \frac{\sum _{23:00}^{07:00}{\mathrm{Hourly}}_{\mathrm{KP}\left(\mathrm{i}\right)}}{{\mathrm{n}}_{\mathrm{t},\mathrm{F}3}}}$$

(11)

$${\mathrm{K}}_{\mathrm{F}3 \mathrm{WD},\mathrm{m}}={\displaystyle \frac{{\mathrm{P}}_{\mathrm{F}3,\mathrm{m}}}{{\mathrm{M}}_{\mathrm{F}3,\mathrm{WD}}}}$$

(12)

$${\mathrm{M}}_{\mathrm{F}3, \mathrm{ST}}={\mathrm{M}}_{\mathrm{F}3,\mathrm{WD}}$$

(13)

$${\mathrm{K}}_{\mathrm{F}3, \mathrm{ST},\mathrm{m}}={\mathrm{K}}_{\mathrm{F}3 \mathrm{WD},\mathrm{m}}$$

(14)

$${\mathrm{M}}_{\mathrm{F}3,\mathrm{SN}}={\displaystyle \frac{\sum _{00:00}^{23:00}{\mathrm{Hourly}}_{\mathrm{KP}\left(\mathrm{i}\right)}}{{\mathrm{n}}_{\mathrm{t},\mathrm{F}3}}}$$

(15)

$${\mathrm{K}}_{\mathrm{F}3 \mathrm{SN},\mathrm{m}}={\displaystyle \frac{{\mathrm{P}}_{\mathrm{F}3,\mathrm{m}}}{{\mathrm{M}}_{\mathrm{F}3,\mathrm{SN}}}}$$

(16)

2.2. EV Charger Integration

The adoption of EVs has progressed slowly on a national scale. However, the latest Energy Performance of Buildings Directive (EPBD) from the European Parliament and Council, particularly Article 14, mandates that new non-residential buildings with more than five parking spaces, as well as existing buildings undergoing major renovations, must include at least one EV recharging point for every five parking spaces. Additionally, 50% of the remaining parking spaces must have pre-cabling to enable future installations. By 1 January 2027, all non-residential buildings with more than 20 parking spaces will also be required to provide at least one recharging point for every 10 spaces [46].

To facilitate a smooth transition towards these goals, the number of available parking spots for all 20 supermarkets was collected to develop an installation scenario of EV charging stations equal to a percentage of the premises’ available parking spots. In Heraklion, only two out of the seven stores have available parking areas, since they are in a densely populated area. The number of maximum spots is met in rural parts of the island, reaching 72 in a retail store near Chania. Therefore, the installation of charging stations in 30% of available spots, regardless of the absolute number of available spots, is deemed appropriate for a smooth transition towards the objective of the EU.

For a realistic simulation, not all the available stations will be occupied at all times of operation. In Figure 6, a scenario of traffic is adopted where the occupancy rate peaks during 18:00–19:00, without reaching 100% occupancy rate.

The selected charging unit has a rated power of 7 kW, and the charging and installation fee is set at EUR 300 per unit, based on a quoted cost by an installation company. Each supermarket’s final hourly energy profile is the estimated demand as calculated in Section 2.1 plus the required power stemming from the station’s nominal capacity and adopted occupancy schedule.

2.3. Energy System Simulation

The proposed energy system’s simulation is a direct application of the methodology described in [47]. The export capacity was defined as each supermarket’s maximum hourly load value, including the additional load from charging stations. PV production was based on normalised hourly values from the PVGIS system, specific to each city in the study. Panels were positioned facing south, with a 35-degree slope, and operated at 22.2% efficiency, according to the quoted product specifications.

The sizes of the PV and battery units and each store’s total area were considered to create a more realistic model. A constraint was set, limiting the installation-occupied area to 30% of each building’s space. It was assumed that a significant portion of the available rooftop and premises space must remain unobstructed to support essential supermarket operations, such as inventory storage, HVAC equipment placement, and customer access. To avoid disrupting these operations, only 30% of the available rooftop area was allocated for PV panel installation.

Key factors, provided in

Table 1. System components’ properties and costs.

Properties	Value
Nominal Panel Capacity:	430 W
Panel Efficiency:	22.2%
Panel Dimensions L × W × H:	1865 × 1040 × 30 mm
Panel Cost:	600 EUR/kW
Energy Content/Rack:	154 kWh
Battery Efficiency:	90%
Rack Dimensions H × W × D:	2200 mm × 1000 mm × 938 mm
Rack Cost:	350 EUR/kWh
Inverter Cost:	100 EUR/kW of PV capacity
PV O&M Cost:	5% of PV initial cost
NPV Discount Rate:	3%
Battery O&M:	4 EUR/kWh of storage capacity

, are operational and maintenance (O&M) costs, system properties, and the discount rate used in the NPV calculation.

The project’s financial feasibility is assessed by factoring in installation costs for the PVs and battery racks, maintenance costs, import charges, and grid export gains, as dictated by the current legislature [48], along with a loan repayment plan spanning the first five years of a 20-year investment. Both the PV and the BESS systems have a 20-year lifespan, while the inverter system coupled to PV panels needs to be replaced every five years of operation.

Hourly production values were received by the PVGIS platform and normalised to estimate the output for any installed capacity and then multiplied by the total capacity for hourly production estimates. For the calculation of occupied area by PV panels, each panel array consisted of 100 panels, while 10-panel arrays were incorporated for supermarkets in dense urban areas with limited space.

To prevent shadowing effects, the distance $D_{PV}$ between arrays was calculated based on the available surface area (17):

$$D_{PV}={\displaystyle \frac{\mathrm{h}}{\tan \left(\mathsf{\omega }\right)}}$$

(17)

The symbol $\mathrm{h}$ represents the solar panel height, while ω refers to the solar elevation angle at noon during the winter solstice, approximately equal to 31.2° for all major locations of the island.

The total surface area for the PV system ${\mathrm{A}}_{\mathrm{PV}}$ was calculated using Equation (18). The panels were arranged side by side, with their width and individual panel area ${\mathrm{Ar}}_{\mathrm{p}}$ contributing to the overall length of the array.

$${\mathrm{A}}_{\mathrm{PV}}\left({\mathrm{n}}_{\mathrm{PV}}\right)={\mathrm{n}}_{\mathrm{PV}}·{\mathrm{Ar}}_{\mathrm{p}}+(\mathrm{round}\left({\displaystyle \frac{{\mathrm{n}}_{\mathrm{PV}}}{100}}\right) - 1)·D_{PV}·100·W_{PV}$$

(18)

where ${\mathrm{n}}_{\mathrm{PV}}$ is the number of installed panels, and $W_{PV}$ is the panel width.

To maximise PV production for load satisfaction and minimise grid imports, a BESS is proposed. The chosen technology is Lithium-Ferrum Phosphate (LFP) cells, stored in racks with specifications provided by the manufacturer’s datasheet.

The surface area occupied by each battery rack (${\mathrm{A}}_{\mathrm{Bat}}$) was calculated using Equation (19):

$${\mathrm{A}}_{\mathrm{Bat}}\left({\mathrm{n}}_{\mathrm{Bat}}\right)={\mathrm{n}}_{\mathrm{Bat}}·W_{Bat}·D_{Bat}$$

(19)

where ${\mathrm{n}}_{\mathrm{Bat}}$ is the number of racks to be installed, $W_{Bat}$ is the rack length, and $D_{Bat}$ is the length of each rack unit.

The battery is considered fully charged at the start of each iteration and supplies power when the PV output is insufficient and there is sufficient storage capacity. It charges with any excess PV production after meeting the load, while surplus energy that is unable to be stored due to insufficient storage is sold to the grid. If the export capacity is reached, excess energy is rejected.

The NPV is calculated as a financial index to assess the investment’s profitability. For this calculation, the initial expenses (CAPEX) for purchasing and installing components were considered. One-third of these costs were covered by a 5-year loan with a 5% interest rate. Annuity payments were calculated using Equation (20) for the project’s lifetime (20 years):

$$\mathrm{Annuity}={\displaystyle \frac{\mathrm{loan}·\mathrm{rate}}{{1 - (1+\mathrm{rate})}^{-\mathrm{years}}}}$$

(20)

Energy import and power fees are based on [46], including details on all applicable fees, tariffs, and carbon footprint taxes.

In this study, 30% of the available area ($\mathrm{A}$) is allocated for energy system installation in each scenario.

The methodology is designed to address the primary challenges retail store brands encounter when evaluating investments in renewable energy systems. As a result, the optimisation focuses exclusively on space allocation and financial profitability. Environmental considerations are limited to ensuring compliance with national and European climate regulations, without incorporating a detailed assessment of the systems’ environmental impacts. Therefore, the optimisation problem strives to maximise the NPV of the 20-year investment while adhering to these space constraints:

$${\mathrm{A}}_{\mathrm{Bat}}\left({\mathrm{n}}_{\mathrm{Bat}}\right)+{\mathrm{A}}_{\mathrm{PV}}\left({\mathrm{n}}_{\mathrm{PV}}\right)\le 0.3·\mathrm{A}$$

(21)

$${\mathrm{n}}_{\mathrm{Bat}},{\mathrm{n}}_{\mathrm{PV}}\ge 0$$

(22)

2.4. Metaheuristic Algorithm

For the optimisation process, the MFO algorithm, developed by Seyedali Mirjalili [49], was selected. This metaheuristic algorithm draws inspiration from the natural behaviour of moths, particularly their movement patterns around a light source, such as flames. In this approach, potential solutions are represented by moths, and their positions within the search space correspond to the variables of the problem. These positions can be organised in a matrix, where each row represents the coordinates (${\mathrm{m}}_{\mathrm{i},\mathrm{j}}$) of a moth (or agent) and each column corresponds to a dimension of the optimisation problem, i.e., the decision variables.

$$\mathrm{M}=\left[\begin{array}{ccc}{\mathrm{m}}_{1,1}& \cdots & {\mathrm{m}}_{1,\mathrm{d}}\\ ⋮& \ddots & ⋮\\ {\mathrm{m}}_{\mathrm{n},1}& \cdots & {\mathrm{m}}_{\mathrm{n},\mathrm{d}}\end{array}\right]$$

(23)

Here, n is the number of moths and d is the number of variables (dimension). The fitness value for all moths is stored in a separate matrix, calculated by evaluating the objective function:

$$\mathrm{OM}=\left[\begin{array}{c}{\mathrm{OM}}_1\\ ⋮\\ {\mathrm{OM}}_{\mathrm{n}}\end{array}\right]$$

(24)

Each moth’s fitness value (${\mathrm{OM}}_{\mathrm{i}}$) is determined by evaluating the objective function. Another critical component of the MFO algorithm is the set of flames, which are organised in a matrix structure similar to that of the moths:

$$\mathrm{F}=\left[\begin{array}{ccc}{\mathrm{F}}_{1,1}& \cdots & {\mathrm{F}}_{1,\mathrm{d}}\\ ⋮& \ddots & ⋮\\ {\mathrm{F}}_{\mathrm{n},1}& \cdots & {\mathrm{F}}_{\mathrm{n},\mathrm{d}}\end{array}\right]$$

(25)

For the flames, there is an array for storing the corresponding fitness values (${\mathrm{OF}}_{\mathrm{i}}$) as follows:

$$\mathrm{OF}=\left[\begin{array}{c}{\mathrm{OF}}_1\\ ⋮\\ {\mathrm{OF}}_{\mathrm{n}}\end{array}\right]$$

(26)

Both moths and flames represent potential solutions, but they are updated differently during each iteration. Moths serve as search agents exploring the search space, while flames represent the best positions discovered by moths so far. Flames act as reference points, guiding the moths in their search. Each moth searches around a specific flame and updates its position if it finds a better solution, ensuring it retains track of its best discovery. This process ensures that a moth always retains its best solution. Constraint handling involves checking the decision variables for any deviations from their lower and upper bounds at the beginning of each iteration. A penalty system is also implemented; if an agent violates any constraints, it is assigned a very low fitness value (in a maximisation problem), resulting in its drop to a lower position in the fitness ranking. To avoid getting trapped in local extrema, moths are initially placed randomly within the feasible area using the Mersenne Twister random number generator, ($\mathrm{Rnd}$), as shown in (27):

$${\mathrm{m}}_{\mathrm{i},\mathrm{j}}=(\mathrm{ub}-\mathrm{lb}+1) · \mathrm{Rnd}+\mathrm{lb}$$

(27)

In the equation, $\mathrm{ub}, \mathrm{lb}$ are the lower and upper bounds of the decision variables. To mathematically model their trajectory, the position of each moth is updated relative to a flame using (28):

$${\mathrm{M}}_{\mathrm{i}}=\mathrm{S}({\mathrm{M}}_{\mathrm{i}},{\mathrm{F}}_{\mathrm{j}})$$

(28)

The equation simulates the spiral movement of a moth around a flame, shown in (29):

$$\mathrm{S}({\mathrm{M}}_{\mathrm{i}},{\mathrm{F}}_{\mathrm{j}})={\mathrm{D}}_{\mathrm{i}}·{\mathrm{e}}^{\mathrm{b}·\mathrm{t}}·\cos \left(2·\mathsf{\pi }·\mathrm{t}\right)+{\mathrm{F}}_{\mathrm{j}}$$

(29)

Here, ${\mathrm{D}}_{\mathrm{i}}$ represents the distance of the i-th moth from the j-th flame, b is a parameter that defines the shape of the logarithmic spiral, and t is a random number in [−1,1]. The parameter $\mathrm{t}$ in the spiral equation influences how close the moth’s next position will be to the flame: $\mathrm{t}=-1$ represents the closest point, while $\mathrm{t}=1$ indicates the farthest. This creates a hyper-ellipse around the flame in all directions, within which the moth’s next position will fall. The spiral movement is a crucial aspect of the proposed method, as it dictates how moths update their positions relative to the flames. This approach enables a moth to navigate “around” a flame rather than simply between them, ensuring a comprehensive exploration and exploitation of the search space. The distance ${\mathrm{D}}_{\mathrm{i}}$ is calculated using (30):

$${\mathrm{D}}_{\mathrm{i}}=\left|{\mathrm{F}}_{\mathrm{j}}-{\mathrm{M}}_{\mathrm{i}}\right|$$

(30)

The iterative process of the MFO algorithm is illustrated in the flowchart of Figure 7:

3. Results

3.1. Final Load Estimations

Figure 8 presents the final estimated hourly energy profiles for the same stores presented in Figure 2. The differences between weekdays, Saturdays, and Sundays remain, with a smooth distribution on each day.

In the same figure, the final electricity demand after the EV charging stations is depicted. The increase in demand is significant, especially in locations with extensive parking areas, where the hourly demand has doubled. The operational burden for such cases could damage each supermarket’s financial planning.

3.2. Optimisation Results

Each supermarket consists of an optimisation problem with two decision variables. For all 20 optimisation problems, the decision variables were the number of panel and battery rack units installed. For the MFO implementation, the maximum number of iterations was equal to 50, and the number of initial search agents was 20. The algorithm was implemented in Microsoft Excel 2021 Version 2501 using a custom VBA script. The average convergence time was 32 min, conducted on a laptop with the following specifications:

• Processor: AMD Ryzen 5 4600H with Radeon Graphics, 3.00 GHz (6 cores, 12 threads)
• Installed RAM: 8.00 GB (7.37 GB usable)
• Operating System: Windows 10, 64-bit operating system, x64-based processor

In Figure 9, the convergence curve for all optimisation problems is depicted:

The most profitable investments are estimated to be in Chania, X2, and X3, while the rest seem to converge to a deviation of EUR 100,000, and the least profitable are retail stores in Heraklion with the least available surface area.

Table 2. Optimal decision variables and results.

Location	Store	PV Panels	BESS Racks	Occupied Area (m2)	Occupied Area (%)	CAPEX (EUR)	NPV (EUR)
Agios Nikolaos	AN1	100	0	158.87	15.81	30,100.00	337,208.78
	AN2	73	0	115.97	29.74	22,273.00	202,538.50
Chania	X1	100	0	158.87	23.00	32,500.00	332,976.67
	X2	199	0	499.81	27.16	65,599.00	669,628.50
	X3	165	0	445.80	29.92	56,265.00	574,382.94
Heraklion	H1	10	0	15.89	19.61	3010.00	35,470.38
	H2	99	0	157.28	29.96	29,799.00	310,023.99
	H3	99	0	157.28	17.59	30,399.00	321,437.20
	H4	99	0	157.28	28.60	29,799.00	242,966.24
	H5	99	0	157.28	26.21	29,799.00	265,996.68
	H6	92	0	146.16	29.53	27,692.00	243,590.52
	H7	50	0	79.43	29.86	15,050.00	164,161.89
	H8	99	0	157.28	26.21	29,799.00	250,978.01
	H9	99	0	157.28	18.29	32,799.00	361,064.22
Ierapetra	I1	99	0	157.28	15.73	35,499.00	350,892.33
Moires	M1	99	0	157.28	15.48	39,299.00	351,224.41
	M2	99	0	157.28	18.29	33,399.00	345,302.63
Neapoli	N1	99	0	157.28	19.42	30,399.00	283,395.66
Rethymnon	R1	99	0	157.28	23.47	32,499.00	341,824.75
Tympaki	T1	99	0	157.28	19.42	34,899.00	351,704.80
Total		1977	0			640,877.00	6,336,769.10

compiles the results after the final iteration, providing estimates for each store separately and the brand overall.

The algorithm determines that no battery storage is to be installed. The maximum area availability is to be reserved for photovoltaic installation.

The expected profit is a little more than ten times the initial expenditure.

Below is Figure 10 for the collective energy mix per region:

A common behaviour is observed in each habitable area under study. The contribution of PV systems ranges between 10.8% and 31%. The outlier of Neapoli is attributed to the fact that it accommodates only one store with similar space as the rest, yet its annual electricity demand is significantly smaller.

Figure 11 illustrates the multi-level RES production allocation for all regions:

Most of the PV production is allocated to load satisfaction, with exported production estimated between 3.29% and 22.37%. The figure illustrates the major contributor to the investment’s profitability, which is the cost reduction in electricity usage due to PV installation.

By manually exploring the decision variable space near the optimal solutions for each supermarket, some decision variable sets are adjusted, leading to the installation of BESS racks and increased estimated profits.

Table 3. Manually improved solutions.

Location	Store	PV Panels	BESS Racks	Occupied Area (m2)	Occupied Area (%)	CAPEX (EUR)	NPV (EUR)
Agios Nikolaos	AN1	100	0	158.87	15.81	30,100.00	337,208.78
	AN2	73	0	115.97	29.74	22,273.00	202,538.50
Chania	X1	100	0	158.87	23.00	32,500.00	332,976.67
	X2	199	0	499.81	27.16	65,599.00	669,628.50
	X3	165	0	445.80	29.92	56,265.00	574,382.94
Heraklion	H1	10	0	15.89	19.61	3010.00	35,470.38
	H2	99	0	157.28	29.96	29,799.00	310,023.99
	H3	99	0	157.28	17.59	30,399.00	321,437.20
	H4	99	1	158.22	28.77	83,699.00	271,795.71
	H5	99	1	158.22	26.37	83,699.00	284,485.85
	H6	92	1	147.10	29.72	81,592.00	260,173.91
	H7	50	0	79.43	29.86	15,050.00	164,161.89
	H8	99	1	158.22	26.37	83,699.00	276,662.02
	H9	99	0	157.28	18.29	32,799.00	361,064.22
Ierapetra	I1	99	0	157.28	15.73	35,499.00	350,892.33
Moires	M1	99	0	157.28	15.48	39,299.00	351,224.41
	M2	99	0	157.28	18.29	33,399.00	345,302.63
Neapoli	N1	99	0	157.28	19.42	30,399.00	283,395.66
Rethymnon	R1	99	0	157.28	23.47	32,499.00	341,824.75
Tympaki	T1	99	0	157.28	19.42	34,899.00	351,704.80
Total		1977	4			856,477.00	6,426,355.14

compiles the new solutions and the increase in NPV, the altered solutions have been highlighted in bold letters:

The manual exploration predicts an increase in profitability, despite the substantial differences in CAPEX. The BESS’s only positive contribution is to Heraklion’s location. This phenomenon is associated with the scale of the relationship between the consumption and PV production time series. In Figure 12, two stores, H2 and H4, have similar available space and the same number of PV panels, but H4 profits from a BESS unit:

Figure 12 emphasises the need for energy storage. For the same production, consumption in H4 is significantly less, meaning that for an extended period of the year, there is a surplus of produced energy, which, through cost reduction and/or export, pays back the cost of the storage unit installation and maintenance.

For an overall increase of EUR 215,600 in upfront costs, the profit over a twenty-year lifetime increases by EUR 89,586.04.

4. Conclusions

The findings of this study highlight the feasibility and potential benefits of integrating RES and EV charging stations within supermarket chains, particularly in regions like Crete where solar irradiance is high. To accomplish this, the monthly data regarding electricity consumption were collected for all retail stores under study. To estimate the hourly consumption profile, including the EV charging infrastructure, the available data of Procida Hall were used. While Procida Hall provides a valuable proxy, it is important to note that its energy use profile might differ slightly from that of supermarkets due to differences in building function and energy-intensive systems. However, these limitations were mitigated by normalisation techniques and sensitivity analysis during optimisation, ensuring the robustness of the results.

The optimisation results using the MFO suggest that PV and battery storage systems can provide substantial NPV estimates significantly surpassing initial capital expenditures, especially in supermarkets with larger available areas. Manual variable exploration around the solutions provided by the MFO yielded a better overall result, by introducing four additional BESS units across the supermarket chain.

This integration not only mitigates operational costs but also ensures the compliance of the company with national and European directives. By accounting for varying solar irradiation across the region and optimising for local conditions, this study provides a more tailored approach than those focusing on more homogeneous settings.

This study’s approach to scaling PV and battery systems based on individual store needs and constraints could serve as a model for similar retail operations worldwide, especially in regions with high solar potential and evolving EV infrastructure requirements. While the results suggest immediate economic viability, the continued decline in battery and PV costs may further enhance project feasibility, making large-scale adoption more accessible. However, grid support and energy distribution challenges may arise with widespread PV deployment; thus, future work could investigate the integration of advanced control systems to manage power flows between supermarkets and local grids.

Regarding optimisation, there are several promising directions for future research. One avenue would be exploring different metaheuristic optimisation techniques (such as PSO or Genetic Algorithms) to compare efficiency and results with the MFO. Additionally, assessing the impact of different government subsidies or incentives on project feasibility could provide deeper insights into financial risks and opportunities. Future studies could also evaluate integrating predictive analytics for optimising real-time energy distribution between PV, storage, and EV chargers based on forecasted energy demand and solar production, potentially enhancing energy savings.