A Hybrid Model‐Data‐Driven Scheduling Strategy for Vehicle‐to‐Grid Interaction Based on Virtual Power Plant Coordination

Electric vehicles (EVs) have experienced vigorous development in recent years. However, their large‐scale integration into the power grid presents challenges related to the ‘curse of dimensionality’ and uncertainties, making it difficult to balance rapid grid demand response with the interests of EV users. To overcome these challenges, this paper proposes a hybrid model‐data‐driven scheduling strategy for vehicle‐to‐grid (V2G) interactions based on virtual power plant (VPP) coordination. It adopts a sequential model‐based aggregated charging scheduling and data‐driven power division mechanism to mitigate dimensionality and uncertainty problems. Furthermore, the data‐driven division strategy is implemented in two stages: both the EV bidding process in Stage 1 and the VPP clearing process in Stage 2 are driven by a distributed deep reinforcement learning (DRL) framework with high sample efficiency. This design not only achieves rapid response to setpoints but also fully accommodates the interests of EV owners. The correctness and effectiveness of the proposed strategy are verified through simulations on a modified IEEE 33‐bus system integrated with massive VPPs and EVs.