Optimizing Energy Storage for Solar-Powered EV Charging Stations
As the world accelerates toward a low-carbon future, electric vehicles (EVs) have emerged as a cornerstone of sustainable transportation. With global EV adoption surging, the infrastructure supporting this transition—particularly charging stations—must evolve to meet growing demand while aligning with environmental and economic goals. In this context, photovoltaic energy storage charging stations (PSCS) are gaining prominence as a key solution, integrating solar power generation, energy storage, and EV charging into a single, efficient system. A recent study published in Zhejiang Electric Power presents a novel approach to optimizing the energy storage capacity of PSCS by accounting for real-world variables such as user charging behavior and photovoltaic (PV) uncertainty.
The research, led by Jiang Yu from the School of Electric Power Engineering at Nanjing Institute of Technology, introduces a comprehensive model that enhances the economic viability and environmental performance of PSCS. Co-authors include Lü Ganyun, Jia Dexiang from State Grid Energy Research Institute Co., Ltd., Yu Xiangyi, Shan Tingting, Yu Ming, and Wu Qiyu from State Grid Jiangsu Electric Power Co., Ltd. Their work addresses a critical challenge in the deployment of renewable-powered charging infrastructure: how to balance cost, reliability, and sustainability in the face of unpredictable solar generation and fluctuating user demand.
The Rise of Photovoltaic Energy Storage Charging Stations
The push for carbon neutrality has placed immense pressure on energy systems worldwide. In China, the government’s commitment to achieving peak carbon emissions by 2030 and carbon neutrality by 2060 has catalyzed investments in clean energy technologies. Among these, PSCS represents a strategic convergence of two major trends: the electrification of transport and the decentralization of power generation.
Unlike conventional charging stations that rely solely on grid electricity, PSCS harnesses on-site solar panels to generate power, stores excess energy in batteries, and supplies it to EVs during periods of high demand or low solar output. This integration reduces dependence on fossil-fuel-based grid power, lowers operational costs, and enhances grid stability by mitigating peak load pressures.
However, the effectiveness of a PSCS hinges on the proper sizing of its energy storage system. Oversizing leads to unnecessary capital expenditure, while undersizing compromises reliability and revenue potential. Traditional optimization models often rely on average or typical PV output data, which fail to capture the inherent variability of solar irradiance due to weather fluctuations, cloud cover, and seasonal changes. Similarly, they frequently overlook the dynamic nature of EV user behavior, which is influenced by factors such as daily travel patterns, temperature, and pricing incentives.
Jiang Yu and his team recognized these limitations and set out to develop a more realistic and robust framework for energy storage capacity planning.
A Data-Driven Approach to Modeling Solar Variability
One of the key innovations in the study is the use of an improved K-means clustering algorithm to generate representative PV output scenarios. Rather than relying on a single “typical day” profile, the researchers analyzed a full year of historical solar generation data from a city in southern China. By applying advanced clustering techniques, they identified three distinct PV output patterns, each associated with a specific probability of occurrence.
The improvement over standard K-means lies in two critical modifications: the selection of initial cluster centers based on maximum distance (to avoid poor initialization), and the use of a Gaussian kernel function instead of Euclidean distance to measure similarity between data points. This allows for a more accurate grouping of days with similar solar profiles, even when the raw numerical values differ slightly due to noise or minor fluctuations.
The result is a probabilistic model that reflects the true range of solar conditions—from sunny days with high output to overcast days with minimal generation. These scenarios form the foundation for a multi-scenario optimization model, ensuring that the energy storage system is designed to perform well under a variety of conditions, not just ideal ones.
To evaluate the quality of the clustering, the team used the silhouette coefficient, a metric that quantifies how well each data point fits within its assigned cluster compared to others. The improved K-means method achieved a silhouette score of 0.628, outperforming both the traditional K-means (0.478) and K-means++ (0.598) algorithms. This higher score indicates tighter, more distinct clusters, leading to more reliable scenario representation.
Incorporating Realistic EV Charging Behavior
While solar variability is a major source of uncertainty, user behavior introduces another layer of complexity. The timing, duration, and intensity of EV charging are influenced by a multitude of factors, including vehicle type, driver routines, ambient temperature, and electricity pricing.
The researchers categorized EVs into three main types—buses, taxis, and private cars—each with distinct usage patterns. Buses follow fixed routes and schedules, requiring predictable charging windows. Taxis operate for extended hours and typically charge twice a day, during midday breaks and overnight shifts. Private car owners, on the other hand, exhibit more flexible behavior, often charging after returning home from work or on weekends following leisure trips.
To model these behaviors accurately, the team incorporated data on daily mileage, departure times, and battery specifications. For instance, private cars were assumed to travel an average of 70 kilometers per day, with a per-kilometer energy consumption of 0.149 kWh under standard conditions. However, the study went further by accounting for the impact of temperature on energy efficiency.
It is well known that extreme temperatures—both hot and cold—reduce EV range due to increased energy demand for heating and cooling. The researchers used a cubic polynomial function to describe the relationship between ambient temperature and energy consumption, based on empirical data. This allowed them to adjust the required charging energy dynamically depending on seasonal and daily weather variations.
Additionally, the model considered the effect of air conditioning load during driving, which varies with speed and climate. By integrating these physical and behavioral parameters, the simulation produced a more accurate representation of daily charging demand.
To handle the stochastic nature of individual driving and charging decisions, the team employed the Monte Carlo method. This involved running thousands of simulations, each time randomly sampling departure times and daily mileages according to statistical distributions (e.g., log-normal for mileage, normal for return time). The aggregated results provided a probabilistic load profile that captured the variability inherent in real-world operations.
Leveraging Demand Response for Grid Optimization
A crucial aspect of the study is the incorporation of demand response (DR), a strategy that uses price signals to influence consumer behavior. In many regions, electricity tariffs vary by time of day, with lower rates during off-peak hours (e.g., nighttime) and higher rates during peak periods (e.g., late afternoon).
The researchers focused on how time-of-use (TOU) pricing could be used to shift private car charging away from peak hours, thereby reducing strain on the grid and improving the economics of the PSCS. They introduced a price elasticity matrix to quantify how changes in charging prices affect load distribution across different time slots.
For example, if the price during peak hours is increased, some users may choose to delay charging until rates drop, leading to a decrease in peak demand and an increase in off-peak consumption. This phenomenon is captured by self-elasticity (the responsiveness of demand in a given period to price changes in the same period) and cross-elasticity (the responsiveness of demand in one period to price changes in another).
By optimizing the TOU pricing structure, the PSCS operator can incentivize users to charge when solar generation is high or grid prices are low, thus maximizing self-consumption of PV energy and minimizing reliance on expensive grid power.
The study found that when private car owners responded to dynamic pricing, the overall load curve became flatter—peak demand decreased while off-peak demand increased. This “peak shaving and valley filling” effect not only reduces operational costs but also enhances grid stability and improves the utilization of renewable energy.
An Economic Model That Values Carbon
Beyond cost and reliability, the researchers emphasized the importance of environmental impact. To reflect the true cost of electricity consumption, they included carbon emissions in their economic model. When the PSCS draws power from the grid, it indirectly contributes to CO₂ emissions, depending on the generation mix at that time. Conversely, using solar power or stored energy reduces emissions.
The model assigns a carbon cost based on the grid’s emission factor, which varies by hour to reflect changes in generation sources (e.g., more coal at night, more renewables during the day). This cost is factored into the total daily expense, encouraging the PSCS operator to minimize grid purchases during high-emission periods.
The objective function of the optimization model seeks to minimize the total daily cost, which includes:
- Grid electricity purchase cost
- Carbon emission cost
- Operation and maintenance expenses
- Revenue from selling excess solar power to the grid
- Revenue from charging EVs
The trade-offs between these components are complex. For instance, investing in a larger battery may increase upfront costs but reduce long-term expenses by enabling more solar self-consumption and avoiding high peak tariffs. Similarly, participating in demand response may slightly reduce per-kWh revenue but improve overall profitability by aligning load with supply.
Simulation Results and Practical Implications
To validate their approach, the researchers conducted a case study on a residential-area PSCS with a 350 kW PV installation and a maximum charging power of 100 kW. Using MATLAB and the CPLEX solver, they compared five different configurations:
- No PV, no demand response
- Typical PV, no demand response
- Typical PV, with demand response
- Multi-scenario PV, no demand response
- Multi-scenario PV, with demand response
The results were striking. The optimal energy storage capacity under the most advanced scenario (multi-scenario PV with demand response) was calculated to be 461 kWh, yielding a daily profit of 4,334.17 yuan. In contrast, systems using typical PV data required either more storage (496 kWh) or delivered lower profits (4,218.21 yuan), highlighting the inefficiency of oversimplified models.
More importantly, the carbon footprint was significantly reduced. The scenario with multi-scenario PV and demand response achieved the lowest daily CO₂ emissions—8,218 kg—compared to over 9,300 kg in the baseline case without solar. This demonstrates that accurate modeling of uncertainty and user behavior leads to both economic and environmental benefits.
The study also revealed that demand response plays a pivotal role in enhancing system performance. Even with the same PV model, enabling price-based load shifting increased daily revenue by over 100 yuan and reduced emissions by nearly 300 kg. This underscores the value of engaging end-users as active participants in the energy transition.
Toward Smarter, Greener Charging Infrastructure
The findings of this research have broad implications for policymakers, utility companies, and charging station operators. As cities expand their EV charging networks, they must move beyond one-size-fits-all designs and embrace data-driven, adaptive planning.
The integration of machine learning (for scenario generation), behavioral modeling (for load forecasting), and economic optimization (for capacity sizing) represents a new paradigm in energy infrastructure development. It shifts the focus from static design to dynamic resilience, ensuring that systems are not only cost-effective today but also adaptable to future uncertainties.
Moreover, the success of demand response suggests that financial incentives can be a powerful tool for shaping sustainable behavior. Rather than mandating charging schedules, operators can use pricing to nudge users toward greener choices—charging when the sun shines, rather than when demand is highest.
Looking ahead, the researchers suggest extending the model to include vehicle-to-grid (V2G) capabilities, where EVs themselves act as distributed energy resources. This would further enhance grid flexibility and open new revenue streams for both operators and vehicle owners.
In conclusion, the work by Jiang Yu and colleagues offers a blueprint for the next generation of smart, solar-powered charging stations. By embracing uncertainty rather than ignoring it, and by viewing users as partners rather than passive consumers, the energy system can become more efficient, equitable, and environmentally sound.
As the electric mobility revolution continues, studies like this will play a crucial role in ensuring that the infrastructure keeps pace—not just in quantity, but in intelligence and sustainability.
Jiang Yu, Lü Ganyun, Jia Dexiang, Yu Xiangyi, Yu Ming, Wu Qiyu, Shan Tingting, School of Electric Power Engineering, Nanjing Institute of Technology; Zhejiang Electric Power, DOI: 10.19585/j.zjdl.202405002