EVs and Microgrids Unite for Smarter Grid Control
The integration of electric vehicles (EVs) into the power grid is no longer a futuristic concept—it’s a present-day reality reshaping how energy is managed. As EV ownership surges and renewable energy sources become more prevalent, the strain on traditional distribution networks intensifies. Voltage instability, power losses, and unpredictable load patterns are emerging as critical challenges for grid operators worldwide. In response, researchers are turning to advanced control strategies that leverage the flexibility of EVs and localized energy systems like microgrids to stabilize the grid and enhance efficiency.
A groundbreaking study published in Automation of Electric Power Systems introduces a novel bi-layer coordinated control strategy that harnesses the combined potential of EVs and microgrids to optimize both active and reactive power flows within distribution networks. The research, led by Peixiao Fan, Jun Yang, Yuxin Wen, Song Ke, Xuecheng Liu, and Leyan Ding from institutions including Wuhan University and the Electric Power Research Institute of China Southern Power Grid Company Limited, presents an innovative solution that not only improves grid performance but also prioritizes user needs in the vehicle-to-grid (V2G) ecosystem.
The paper, titled “Bi-layer Coordinated Control Strategy of Distribution Network Considering Participation of Electric Vehicles and Microgrid,” proposes a framework that integrates real-time decision-making with intelligent learning algorithms. Unlike conventional voltage regulation methods that rely solely on fixed infrastructure such as capacitor banks or static var compensators (SVCs), this new approach treats EVs and microgrids as dynamic, responsive assets capable of providing both active and reactive power support.
At the heart of the proposed system is a dual-layer architecture. The upper layer manages the overall distribution network, making high-level decisions based on real-time data such as voltage levels, power flows, and equipment status. It sends control signals to key components including SVCs, EV charging stations, and microgrid units. The lower layer, embedded within each microgrid, receives these commands and executes localized coordination among internal resources—such as photovoltaic inverters, micro-turbines, and energy storage systems—to meet the requested power exchange while maintaining local stability.
What sets this model apart is its deep consideration of human behavior. The researchers recognize that EV owners are not passive participants in the energy system; their driving habits, charging preferences, and temporary usage needs directly influence the availability of vehicle-based energy resources. To account for this, the team developed a travel chain-based model that simulates the spatial and temporal movement of EVs across different zones—home, workplace, and public areas. This allows the system to predict when and where EVs will be available for grid interaction, ensuring that control actions do not interfere with essential user requirements.
For instance, the model distinguishes between three types of EV states: those requiring mandatory charging to meet departure expectations, those capable of bidirectional power exchange, and those already fully charged and unable to accept further energy input. By classifying vehicles in this way, the controller can avoid unnecessary discharging that would compromise user satisfaction—a common pitfall in earlier V2G schemes.
This human-centric design philosophy extends to the quantification of user demand loss. Instead of treating all EVs uniformly, the researchers introduced a metric that measures the difference between an EV’s maximum possible charging rate and its actual charging rate under grid control. When this gap is minimized, users experience little to no delay in reaching their desired state of charge, preserving trust in the V2G system. The study demonstrates that by incorporating this cost into the control algorithm, the system can maintain high user satisfaction without sacrificing technical performance.
To manage the complexity of this multi-objective optimization problem, the authors turned to artificial intelligence. They developed an enhanced version of evolutionary-deep reinforcement learning (EDRL), combining the global search capabilities of evolutionary algorithms with the adaptive learning power of deep reinforcement learning (DRL). Traditional DRL methods often struggle in environments with deceptive reward structures—situations where short-term gains lead to long-term inefficiencies. By integrating evolutionary guidance and novelty search mechanisms, the improved EDRL algorithm avoids local optima and converges faster to superior control policies.
In practical terms, the algorithm learns through repeated simulation cycles, adjusting its strategy based on feedback from the environment. The reward function is carefully designed to balance multiple objectives: minimizing voltage deviations, reducing network losses, and limiting user demand loss. Each of these factors is weighted according to its importance, allowing the system to make trade-offs that reflect real-world operational priorities.
The researchers tested their strategy on a modified IEEE 33-node distribution network, a standard benchmark in power systems research. The test scenario included realistic load profiles, solar generation data with noise injection to simulate forecasting uncertainty, and a fleet of 500 EVs distributed across residential and commercial charging stations. Control decisions were made every 15 minutes over a 24-hour period, reflecting the granularity needed for real-time grid management.
Results showed that the proposed strategy effectively maintained node voltages within the safe range of 0.95 to 1.05 per unit, even during peak load hours between 11:00 and 13:00. Network losses were reduced by 12.17% compared to a particle swarm optimization (PSO) baseline, while voltage deviation dropped by 65.68%. More importantly, user demand loss was minimized, demonstrating that grid stability and customer satisfaction are not mutually exclusive goals.
When compared to other machine learning approaches such as deep deterministic policy gradient (DDPG) and unenhanced EDRL, the improved algorithm demonstrated superior convergence speed and control accuracy. While DDPG failed to escape local optima and delivered poor performance, the enhanced EDRL achieved stable convergence within 20,000 to 30,000 training episodes—significantly outperforming its counterparts in both speed and solution quality.
One of the most compelling findings was the impact of explicitly modeling user demand. In a comparative test where the user cost component was removed from the reward function, the system caused a 4.1 times increase in demand loss with negligible improvement in grid metrics. This underscores a crucial insight: ignoring user needs leads to inefficient and socially unsustainable control strategies. A truly intelligent grid must respect the autonomy and convenience of its end users.
The role of microgrids in this architecture cannot be overstated. These localized energy clusters act as intermediaries between the main grid and distributed resources, enabling finer control and faster response times. When the upper controller requests a specific power exchange, the microgrid’s internal coordinator dynamically allocates the task among available assets. For example, if additional reactive power is needed, photovoltaic inverters can adjust their power factor; if active power support is required, micro-turbines or batteries can ramp up output. This hierarchical delegation ensures that control signals are translated into feasible actions at the equipment level.
Moreover, the inclusion of microgrids reduces the need for costly centralized infrastructure upgrades. By leveraging existing distributed generation and storage, utilities can achieve voltage regulation and loss reduction without installing new capacitors or transformers. This makes the proposed strategy not only technically effective but also economically viable.
From a policy perspective, the research aligns closely with recent national initiatives promoting grid-EV integration. In January 2024, China’s National Development and Reform Commission released guidelines emphasizing the importance of V2G technology in advancing energy transition and technological innovation. The study provides a concrete technical foundation for implementing such policies, showing how intelligent control systems can turn EVs from grid burdens into valuable assets.
However, the authors acknowledge limitations and outline directions for future work. While the current model captures key aspects of user behavior, it simplifies market dynamics and does not fully explore incentive mechanisms that could encourage broader participation. Additionally, the assumption of perfect communication between grid components may not hold in real-world deployments with latency or data loss. Future research will focus on refining user response modeling and incorporating market-based signals to create more holistic and scalable solutions.
The implications of this work extend beyond China’s power sector. As countries around the world accelerate their electrification and decarbonization efforts, the lessons learned from this study offer a roadmap for managing increasingly complex energy systems. Whether in urban centers with dense EV fleets or rural areas with isolated microgrids, the principles of bi-layer coordination, behavior-aware control, and AI-driven optimization can be adapted to diverse contexts.
For utility operators, the message is clear: the future of grid management lies in distributed intelligence. Rather than relying on top-down commands, next-generation control systems must embrace bottom-up adaptability, drawing on the collective flexibility of millions of connected devices. EVs, once seen merely as loads, are now becoming mobile energy resources. Microgrids, once viewed as isolated islands, are evolving into active participants in the larger energy ecosystem.
The success of this transition depends on interdisciplinary collaboration—between power engineers, computer scientists, behavioral economists, and policymakers. The research team exemplifies this collaborative spirit, bringing together expertise in power system dynamics, artificial intelligence, and energy policy. Their work demonstrates that technological innovation must be grounded in real-world constraints and human-centered design.
As the energy landscape continues to evolve, strategies like the one proposed by Fan, Yang, Wen, Ke, Liu, and Ding will play a pivotal role in ensuring reliability, efficiency, and equity in electricity delivery. The vision of a smart, responsive, and user-friendly grid is no longer a distant dream—it is being built today, one algorithm, one EV, and one microgrid at a time.
The study titled “Bi-layer Coordinated Control Strategy of Distribution Network Considering Participation of Electric Vehicles and Microgrid” was conducted by Peixiao Fan, Jun Yang, Yuxin Wen, Song Ke, Xuecheng Liu, and Leyan Ding from the Hubei Engineering and Technology Research Center for AC/DC Intelligent Distribution Network, School of Electrical Engineering and Automation at Wuhan University, and the Electric Power Research Institute of China Southern Power Grid Company Limited. It was published in Automation of Electric Power Systems, Vol. 48 No. 19, October 10, 2024, with DOI: 10.7500/AEPS20240203001.