Wind Power Prediction-Driven Optimal Dispatch of Microgrid
Year: 2026
Type: conference-paper
Abstract: To address the scheduling challenges caused by the stochastic nature of wind power and the partially observable environment in microgrids, this paper proposes a wind power forecasting-driven optimal scheduling method based on an improved deep reinforcement learning approach. First, a Res-ConvLSTM-Attention model is developed for accurate wind power prediction. The scheduling problem is then formulated as a Partially Observable Markov Decision Process (PO-MDP), and a residual-like Soft Actor-Critic (R-SAC) algorithm is designed to enhance feature propagation and policy learning in high-dimensional state spaces. Simulation results demonstrate that the proposed method effectively mitigates wind power fluctuations, improves operational economy, and increases wind power utilization. Compared with DDPG and SAC, the total cost is reduced by 17.1% and 8.7%, respectively, and the wind power utilization rises to 93.6%, while achieving superior convergence speed and stability, making it well-suited for microgrid optimal scheduling under uncertain conditions.
Authors: Qiang Lei, Xu Zhao, Jingwen Sun
Institution: Xi'an Polytechnic University
Language: English
Subfield: Control and Systems Engineering
Field: Engineering
Domain: Physical Sciences
Open Access status: hybrid
OpenAlex ID: W7204120499