Real-Time Scheduling of Wind-Solar-Hydro Complementary System Based on Deep Reinforcement Learning

In the context of carbon neutrality, the design of wind-solar-hydro complementary systems aims to enhance the integration of renewable energy. Real-time scheduling of the complementary systems is crucial for ensuring their efficient operation. However, traditional model-based scheduling methods heavily rely on accurate mathematical modeling, making it inflexible to adapt to the complex and dynamic operating environments of complementary systems. Therefore, this paper proposes a real-time scheduling approach based on deep reinforcement learning (DRL) for wind-solar-hydro complementary systems. Initially, a real-time scheduling model of the complementary systems is established. Subsequently, the scheduling problem is expressed as a Markov decision process, with real-time scheduling decisions implemented based on the soft actor-critic (SAC) algorithm. Finally, the proposed scheduling method is applied in a case study of a complementary system. Results indicate that the proposed approach achieves enhancements of 4.14% and 12.04% in cascade energy storage compared to deep deterministic policy gradient and genetic algorithms, with a shorter decision-making time.