Smart Charging Strategy Boosts Renewable Integration at EV Stations
As electric vehicles (EVs) continue to gain momentum across global markets, one of the most pressing challenges facing urban infrastructure is how to manage their growing energy demands without overburdening power grids. With millions of new EVs hitting roads each year, uncontrolled charging behavior can lead to sharp spikes in electricity demand, especially during peak hours. This not only increases operational costs for charging station operators but also undermines efforts to integrate clean, renewable energy into the grid. Now, a team of researchers from China has introduced an innovative online optimization strategy that could transform how EV charging stations operate—making them smarter, more efficient, and better aligned with renewable energy availability.
The study, led by Zhou Zhuo of Jiangsu Wiscom System Co., Ltd., in collaboration with Lu Xiang, Liu Haitao, Qi Shenglong, Han Tao, and Wang Qing from institutions including State Grid Ningxia Electric Power Research Institute, State Grid Electric Power Research Institute, and Nanchang University, presents a real-time charging power optimization framework designed specifically for EV charging stations equipped with renewable energy sources such as solar photovoltaic (PV) and small-scale wind generation systems. Published in the February 2024 issue of Electrical Measurement & Instrumentation, the research introduces a novel approach that dynamically adjusts charging schedules based on current vehicle status, predicted renewable output, and grid conditions—all within a 24-hour planning horizon updated every 15 minutes.
What sets this strategy apart is its ability to balance multiple objectives simultaneously: maximizing the use of locally generated solar and wind power, minimizing reliance on the main power grid, reducing electricity costs through time-of-use tariff optimization, and smoothing overall load fluctuations to support grid stability. These goals are particularly critical in urban environments where commercial charging stations serve hundreds of vehicles daily, often without coordinated control mechanisms.
The core innovation lies in a two-stage optimization process that significantly improves computational efficiency and response speed—key requirements for any real-time control system. Traditional optimization models struggle with the sheer complexity of managing dozens or even hundreds of charging points, each with variable start times, required energy levels, and connection durations. The decision matrix for such a system can easily exceed thousands of variables, making real-time computation impractical.
To overcome this bottleneck, the team developed a state-dependent decision variable classification method. Before initiating the optimization algorithm, the system categorizes all charging points into three distinct groups based on their current operational state. The first group includes chargers that are either unoccupied or connected to fully charged vehicles—these are immediately excluded from further optimization since no charging action is needed. The second group consists of vehicles that require full-power charging throughout their entire parking duration to meet their next driving needs; these are assigned a fixed charging schedule and removed from the variable pool. Only the third group—vehicles whose charging requirements allow for flexible power modulation—is subjected to active optimization.
This classification dramatically reduces the dimensionality of the problem. Instead of optimizing all 60 charging points at a typical station, the algorithm may only need to compute optimal power allocations for a subset—sometimes as few as 20–30 units—depending on real-time occupancy and charging demands. Moreover, as vehicles charge and their remaining time decreases, the search space for each active charger shrinks further, enhancing both speed and precision.
Once the reduced set of decision variables is identified, the system employs a modified Differential Evolution Algorithm (DEA), a powerful evolutionary computation technique known for its robustness in solving complex, non-linear optimization problems. Unlike gradient-based methods that can get trapped in local optima, DEA explores the solution space through mutation, crossover, and selection operations across a population of candidate solutions. In this application, the algorithm is applied in two phases: first to evaluate and select the best combination of charging strategies across all active vehicles, and then to fine-tune individual charging profiles within that combination.
The objective function guiding the optimization is a weighted sum of three key performance indicators. The first measures how closely the total EV charging load matches the available renewable generation—essentially encouraging the system to “follow” the solar and wind output curve. The second component minimizes the cost of electricity drawn from the grid by shifting charging to off-peak periods when tariffs are lower. The third aims to stabilize the net power demand seen by the grid, avoiding sudden ramps that could destabilize distribution networks.
Field simulations based on one week of real-world data from a commercial charging station in China demonstrated the effectiveness of the proposed strategy. The test site featured 60 charging points serving a fleet of 300 EVs, with a combined PV and wind capacity capable of generating up to 200 kW under favorable conditions. Using historical weather and driving pattern data, the researchers simulated a full week of operations under three scenarios: uncontrolled charging (where vehicles charge at maximum rate upon arrival), partially controlled charging (where vehicles charge only to their immediate driving needs), and the proposed optimized strategy.
Results showed a 57% reduction in the mismatch between renewable generation and EV load compared to uncontrolled charging. This means that nearly twice as much solar and wind energy was directly consumed on-site, reducing curtailment and improving energy self-sufficiency. Furthermore, the total electricity cost for the charging station dropped by 23%, from approximately 24,800 RMB per week under uncontrolled charging to just 19,000 RMB with optimization. Even more impressively, the peak grid demand was reduced by over 30%, and the overall load profile became significantly smoother, with fewer sharp transitions that could stress transformers or trigger protective relays.
One of the most notable achievements was the computational performance. Despite the complexity of the problem, the average time to compute a new 24-hour charging plan was just 18.7 seconds—well within the 15-minute update interval. This represents a 28% improvement over conventional optimization methods, which took an average of 26.2 seconds. The fastest computation completed in under 10 seconds, while the longest took 25 seconds during periods of high vehicle turnover. These response times make the system viable for real-world deployment, where delays could result in suboptimal or outdated charging instructions.
The implications of this research extend beyond individual charging stations. As cities move toward decarbonization, integrating distributed energy resources like rooftop solar and EV batteries into a responsive, intelligent grid becomes essential. This study provides a practical blueprint for how charging infrastructure can act as a flexible load that supports—not strains—the power system. By aligning EV charging with renewable availability and grid conditions, the technology helps bridge the gap between intermittent generation and stable consumption.
Moreover, the approach aligns with broader trends in smart grid development, where digitalization, automation, and predictive analytics are transforming passive consumers into active participants in energy markets. In the future, such optimized charging stations could participate in demand response programs, offering grid services like frequency regulation or peak shaving in exchange for financial incentives. They could also integrate with building energy management systems, using surplus solar power not just for vehicles but also for HVAC, lighting, or battery storage.
From a policy perspective, the findings underscore the importance of supporting advanced charging management systems through regulatory frameworks and incentive programs. While many governments have focused on deploying charging hardware, less attention has been paid to the software and control systems that determine how those chargers are used. As this research shows, the intelligence behind the charger may be just as important as the charger itself.
For EV owners, the benefits are indirect but significant. Smoother grid operation means fewer blackouts and more reliable service. Lower operating costs for charging stations could translate into cheaper charging fees over time. And by maximizing renewable usage, the environmental footprint of EV ownership is further reduced—enhancing the sustainability argument for electrified transportation.
Industry experts note that while the technology is promising, several challenges remain before widespread adoption. One is data interoperability—ensuring that charging stations can reliably access accurate forecasts of renewable generation, grid prices, and vehicle availability. Another is cybersecurity, as any networked control system becomes a potential target for malicious attacks. Additionally, user acceptance is crucial; drivers must trust that the system will deliver a fully charged vehicle when needed, even if charging occurs slowly or during off-peak hours.
The research team acknowledges these limitations and emphasizes that their model serves as a foundational framework rather than a turnkey solution. Real-world deployment would require customization based on local climate, driving patterns, utility rate structures, and station design. Nevertheless, the core principles—state-based variable reduction, evolutionary optimization, and multi-objective balancing—are broadly applicable and scalable.
Looking ahead, the authors suggest several avenues for future work. These include incorporating vehicle-to-grid (V2G) capabilities, where EVs can discharge back to the station or grid during peak periods; integrating battery degradation models to extend vehicle lifespan; and expanding the optimization horizon beyond 24 hours to include longer-term forecasting. Machine learning techniques could also be used to improve prediction accuracy for both renewable generation and user behavior.
In conclusion, the study represents a significant step forward in the intelligent management of EV charging infrastructure. By combining advanced classification techniques with robust optimization algorithms, the proposed strategy enables charging stations to operate more efficiently, sustainably, and economically. As the world transitions to electric mobility, solutions like this will be essential for ensuring that the grid can keep pace with the revolution on the roads.
The research was conducted by Zhou Zhuo from Jiangsu Wiscom System Co., Ltd., Lu Xiang, Liu Haitao, Qi Shenglong from State Grid Ningxia Electric Power Research Institute, Han Tao from State Grid Electric Power Research Institute, and Wang Qing from Nanchang University. It was published in Electrical Measurement & Instrumentation, Volume 61, Issue 2, February 15, 2024, DOI: 10.19753/j.issn1001-1390.2024.02.015.