The results present a comprehensive analysis of the simulation and optimization results for the Ultra-Fast Charging Station (UFCS). Three system configurations are evaluated—grid-only (baseline), PV-only, and PV integrated with a Battery Energy Storage System (BESS). The optimization of each configuration is conducted using a Genetic Algorithm (GA) to maximize the Net Present Value (NPV). Demand forecasting is carried out using a Gated Recurrent Unit (GRU)-based deep learning model, and performance is assessed through MATLAB-based simulations.
Baseline scenario: Grid-Only UFCS
The initial scenario involves operating the UFCS entirely using grid-supplied electricity. The forecasted average daily demand is 9669.52 kWh on weekdays and 7776.41 kWh on weekends, totaling 3,332,490.84 kWh annually. The first-year revenue generated from EV charging is estimated at €2,332,743.58, while the annual cost for purchasing electricity from the grid is €581,852.90. Over a 20-year analysis period, the baseline NPV is determined to be €27.29 million, serving as a reference for further evaluations.
Scenario 1: PV integration
In this configuration, PV energy is used to supplement grid power. The Conergy PowerPlus 300 Wp monocrystalline modules are employed for simulation. Using GRU-based forecasts, August 13 (weekday) and August 9 (weekend) are selected as representative days for evaluating PV performance. The results show:
-
Surplus PV energy: 90,462.30 kWh (weekday) and 78,070.07 kWh (weekend).
-
Grid energy purchased: 166,045.23 kWh (weekday) and 96,859.92 kWh (weekend).
Figure 10 illustrates the PV generation and load demand on August 13 (weekday), while Fig. 11 provides a similar comparison for August 9 (weekend). The actual PV output value of solar data has been taken from the National Solar Radiation Database at a specific location in Italy24. Figure 12 presents the cumulative surplus and grid energy purchases for both scenarios. These results highlight improved energy self-sufficiency, particularly on weekends, though considerable surplus PV energy remains unutilized.
Forecasted PV power output for August Weekday (13th Aug 2020).
Forecasted PV power output for August Weekend (9th Aug 2020).
Power Surplus of PV and Demand.
Scenario 2: PV and BESS integration
This hybrid configuration addresses PV surplus losses by incorporating a BESS to store excess energy for later use. Simulation outcomes reveal:
-
Battery-stored energy: 358,308.80 kWh (weekday) and 354,529.78 kWh (weekend).
-
Remaining surplus energy: 89,956.64 kWh (weekday) and 77,213.23 kWh (weekend).
-
Grid energy deficit: 165,598.63 kWh (weekday) and 96,103.16 kWh (weekend).
Figure 13 visualizes the distribution of surplus and stored energy, while Fig. 14 tracks the daily battery State-of-Charge (SOC). These results demonstrate enhanced energy utilization and lower grid dependency, particularly during periods of high solar output.
Weekday power surplus with PV and BESS.
Weekend power surplus with PV and BESS.
Energy reliability assessment
To evaluate the operational robustness and grid independence of the proposed UFCS system, two key reliability metrics were employed:
-
Energy Sufficiency Ratio (ESR): The ESR represents the proportion of total energy demand that is fulfilled by local renewable sources — namely, solar PV and the battery energy storage system (BESS). It reflects how much of the load is met without relying on the utility grid.
Weekday: 57.4%.
Weekend: 78.95%.
These results indicate that during weekdays, over half of the UFCS’s energy needs are supplied by the integrated PV + BESS system. On weekends, the ratio increases substantially due to lower demand and higher availability of surplus solar generation.
Weekday: 59.03%.
Weekend: 51.74%.
While ESR is higher on weekends due to reduced load, the AR is slightly lower compared to weekdays. This is because short periods of high demand or low PV output — even if brief — can require grid support and reduce autonomy, even though the overall energy consumption is low.
Optimization and economic result
This section consolidates the optimization outcomes across all the three system configurations – Grid-only, PV-only, PV and BESS and interprets their techno-economic implications. The Genetic Algorithm (GA) was implemented in MATLAB to maximize Net Present Value (NPV) over a 20-year horizon, considering forecasted energy demand (weekday/weekend), GRU-based PV generation, capital and operational expenditures, and system constraints. Key input parameters included PV and BESS capacities, financial data (CAPEX, OPEX, replacement costs), and forecasted demand.
The results of the GA optimization are summarized in Table 3. In the baseline scenario, the UFCS is entirely dependent on grid-supplied electricity. The forecasted daily demand amounts to 9669.52 kWh on weekdays and 7776.41 kWh on weekends, leading to an annual energy consumption of approximately 3.33 GWh. The first-year revenue from EV charging operations is estimated at €2,332,743.58. However, the annual energy purchase cost from the grid totals €581,852.90, a significant operational expense that escalates over time.
Over the 20-year life cycle, the system yields a total Net Present Value (NPV) of €27.29 million. This scenario serves as a reference benchmark and illustrates the economic limitations of a grid-only UFCS model, which is highly exposed to electricity price volatility, grid congestion, and future carbon pricing. The lack of energy autonomy also poses risks related to grid failures and peak-hour tariffs. While simple to implement, this model lacks sustainability and long-term economic competitiveness.
PV-Only configuration
In this configuration, the UFCS incorporates an optimized photovoltaic (PV) system to offset grid dependency. The simulation, guided by GRU-based solar forecasting, identifies an optimal PV sizing of approximately 1000 kW. Under this configuration, the system generates a higher NPV of €33.48 million, reflecting a profit gain of €6.19 million over the baseline grid-only scenario.
PV integration contributes significantly to reducing grid electricity purchases, especially during daytime hours. This results in considerable cost savings and a cleaner energy profile. However, due to the temporal mismatch between solar generation (peaking midday) and EV charging demand (spread throughout the day and evening), a substantial amount of generated PV energy is unutilized. On representative days (August 13 and August 9), surplus PV energy exceeds 90 MWh and 78 MWh respectively, highlighting inefficiencies. Despite these energy losses, the PV-only system offers a compelling economic improvement over the baseline. However, the findings also reveal the need for energy storage to manage surplus generation and further improve system utilization.
PV and BESS configuration
The PV + BESS configuration integrates battery storage to address the limitations of the PV-only model. The GA-based optimization recommends a PV capacity of 1199.9 kW and a BESS capacity of 500.057 kWh. The battery stores excess solar energy generated during low-demand hours and discharges it during evening peaks or cloudy intervals, thereby smoothing the demand-supply gap.
The simulation results show that on weekdays, 358,308.80 kWh of PV energy is stored, while 354,529.78 kWh is stored on weekends. Remaining surplus energy is significantly reduced to 89,956.64 kWh (weekday) and 77,213.23 kWh (weekend), indicating a substantial improvement in renewable energy utilization. Grid electricity demand is also reduced to 165,598.63 kWh (weekday) and 96,103.16 kWh (weekend).
Financially, this scenario achieves a total NPV of €33.97 million, yielding an additional gain of €0.49 million over the PV-only configuration and €6.68 million over the baseline. The integration of BESS leads to enhanced energy reliability, lower curtailment, and greater autonomy. Figure 15 illustrates the GA convergence trend, confirming effective algorithm performance and rapid fitness convergence. Although BESS increases capital cost, it improves both technical and financial resilience, especially under dynamic solar availability.
Sensitivity analysis
To evaluate the robustness of the proposed system under evolving market conditions, a sensitivity analysis was conducted. The cost parameters for sensitivity analysis are given in Table 4.
It considers anticipated reductions in PV and battery costs based on industry trends. The new economic assumptions are as follows:
With these parameters, the GA optimization identifies a slightly adjusted optimal configuration of 1199.87 kW for PV and 505.47 kWh for BESS. This cost scenario results in an NPV of €34.05 million—an increase of €6.76 million compared to the baseline grid-only system and €80,000 above the current-cost PV + BESS configuration.
Figure 16 displays the enhanced convergence behaviour under the revised cost structure, confirming the continued effectiveness of the GA in identifying profitable configurations. These results demonstrate that as component prices decline, the economic attractiveness of hybrid renewable UFCS systems will continue to grow, making them a sustainable long-term investment strategy.
Convergence Curve of GA for sensitivity analysis.
Discussion
The results of this study demonstrate the technical and economic benefits of integrating renewable energy systems into Ultra-Fast Charging Stations (UFCS) through a deep learning–driven optimization framework. Three configurations were evaluated: grid-only, PV-only, and PV + BESS, each optimized to maximize long-term Net Present Value (NPV). In the grid-only baseline, the UFCS achieves an NPV of €27.29 million over 20 years. This scenario serves as a reference, highlighting full reliance on grid energy and the associated high operational cost. While simple to deploy, it exposes the system to grid price volatility and peak load stress.
Scenario 1 (PV Integration) shows a notable improvement, with NPV rising to €33.48 million. This gain of €6.19 million confirms that PV integration significantly reduces grid dependency and improves cost-effectiveness. However, a large portion of the generated solar energy remains unutilized due to the mismatch between generation timing and demand, especially on weekdays. This underutilization presents a clear opportunity for storage optimization.
Scenario 2 (PV + BESS Integration) addresses this gap. By incorporating a 500.057 kWh battery, surplus solar energy is stored and used during peak demand or low-generation periods. This results in a higher NPV of €33.97 million — an additional gain of €490,000 over the PV-only case. The system benefits from smoother demand-supply balance, lower curtailment losses, and reduced energy imports from the grid.
The Energy Sufficiency Ratio (ESR) and Autonomy Ratio (AR) further validate the benefits of hybridization. ESR reaches 57.4% on weekdays and 78.95% on weekends, showing that most energy needs are met internally. AR values of 59.03% (weekday) and 51.74% (weekend) demonstrate that the UFCS can operate grid-independently for over half the time. These metrics confirm the technical viability of a renewable-powered UFCS, especially during periods of lower demand such as weekends.
Sensitivity analysis reveals that with projected reductions in PV and BESS CAPEX (to €900/kW and €380/kWh respectively), the optimized system achieves a peak NPV of €34.05 million. This result reflects a gain of €6.76 million over the baseline and indicates that the system becomes even more economically attractive as component costs continue to decline.
Overall, the PV + BESS configuration emerges as the most robust and financially viable option. The integration of deep learning–based solar forecasting (GRU), tailored weekday/weekend demand modeling, and evolutionary optimization (GA) enables a realistic, resilient, and profitable system design for UFCS — directly addressing both grid stress and sustainability objectives.
Comparative analysis and novelty
The proposed framework demonstrates significant advancements over existing studies on renewable-powered Ultra-Fast Charging Stations (UFCS). While previous works have explored PV and BESS integration or generic optimization methods, they often rely on static demand profiles, traditional forecasting models, or simplified economic analysis. In contrast, this study delivers a holistic, data-driven approach that combines machine learning, system design, and financial modelling.
Key differentiators of this work include:
-
Forecasting-Integrated Optimization: Unlike conventional models that assume average solar output, this study employs Gated Recurrent Unit (GRU) neural networks to generate high-resolution PV forecasts. These forecasts are directly integrated into the sizing optimization, ensuring that system capacities are aligned with temporal solar variability.
-
Dual Demand Profile Modeling: This study is among the first to distinctly model weekday and weekend EV charging demand based on probabilistic vehicle arrival distributions. This differentiation reflects real-world usage patterns and enables more accurate energy planning for UFCS deployments.
-
Economic Optimization Using NPV: The optimization framework maximizes Net Present Value (NPV) over a 20-year horizon by considering detailed CAPEX, OPEX, battery degradation, and energy trading dynamics. This contrasts with prior studies that often focus on short-term costs or energy efficiency alone.
-
System Reliability Metrics: To assess operational independence and resilience, the study introduces Energy Sufficiency Ratio (ESR) and Autonomy Ratio (AR). These metrics quantify the system’s ability to operate without grid support—an aspect rarely addressed in comparable works.
-
Forward-Looking Sensitivity Analysis: A parametric cost analysis evaluates how future reductions in PV and BESS costs affect profitability, offering valuable insights for long-term infrastructure planning and investment decisions.
Compared to earlier literature, this work provides a comprehensive and adaptive design strategy that reflects both technical and economic realities. By uniting deep learning-based forecasting, realistic demand modelling, and economic optimization, the proposed framework presents a replicable solution for cost-effective, high-performance UFCS design. Collectively, these innovations advance the state of the art in intelligent EV infrastructure design and planning.
