Multi-Agent DDPG-Based Load Frequency Control with Economic Dispatch for Low-Inertia Grids

Abstract

As the penetration of renewable energy sources (RES) increases, the power grid experiences a rapid decline in rotational inertia and an increase in overall stochasticity. Both of these factors make it difficult to maintain a tight balance between power generation and demand, thereby causing the frequency to swing outside its stable limits. While existing control strategies perform well, the need for an adaptive, intelligent controller is increasing with the current trend toward RES integration. This paper aims to solve the load frequency control problem along with the economic dispatch of load for a multi-area system by implementing a multi-agent deep deterministic policy gradient (MADDPG) algorithm, where each area is assigned a globally trained, local controller agent. The results show that the proposed MADDPG controller achieves effective frequency regulation and coordinated economic dispatch as system complexity increases.