Continuous Active Power Dispatch Optimization for Virtual Power Plants with Energy Storage Using DDPG
Continuous active power dispatch is a critical task for virtual power plants (VPPs) with energy storage, particularly under the increasing uncertainty of renewable generation and the growing demand for real-time coordination in smart energy systems. This study proposes an improved Deep Deterministic Policy Gradient (DDPG)-based optimization framework to enhance continuous dispatch performance for VPPs integrating wind power, photovoltaic generation, and battery energy storage. A multi-objective scheduling model is established by jointly considering dispatch cost, renewable energy utilization, and operational constraints, while priority experience replay, cross-attention mechanisms, gradient clipping, and cosine learning-rate decay are incorporated to improve training efficiency and policy stability. Simulation results demonstrate that the proposed approach significantly outperforms conventional methods by reducing average daily dispatch costs by 16.6%, achieving a renewable energy utilization rate of 96.8%, and maintaining rapid power balance recovery under highly volatile operating conditions. The optimized dispatch strategy further improves system robustness and dynamic adaptability in scenarios involving renewable fluctuations and sudden load variations. Beyond intelligent energy management, the proposed framework provides a practical optimization methodology for communication-enabled smart grids and distributed electromagnetic sensing infrastructures, where reliable information exchange and adaptive control are essential for coordinated operation of energy and wireless network resources.