LFC Strategy of Power System with Hybrid Regulation Resources: A Multi-Agent-Based Reinforcement Learning Algorithm

As renewable energy sources become increasingly prevalent, relying solely on conventional generators for load frequency control (LFC) in power systems proves to be neither cost-effective nor practical. Consequently, a variety of regulation resources (called hybrid regulation resources) will be integrated into the LFC. The heterogeneous characteristics of hybrid regulation resources pose a significant challenge to the LFC strategy design. In this context, this paper proposes a novel LFC strategy based on the multi-agent twin delayed deep deterministic policy gradient (MATD3) algorithm. We establish a mathematical model of the power system with hybrid regulation resources to create a reinforcement learning environment. Furthermore, we formulate a reward function consisting of the area control error (ACE) and the frequency deviation according to the control objectives of the LFC. Given that the hybrid regulation resources exhibit both continuous and discrete properties, a rounding function discretizes these actions tailored for the resources with discrete properties. Finally, the effectiveness of the proposed method is verified through simulation and comparison.