Stability enhancement through reinforcement learning: Load frequency control case study

Abstract

A multi-agent based control architecture using reinforcement learning is proposed to enhance power system stability. It consists of a layer of local agents and a global agent that coordinates the behavior of the local agents. Load frequency control is chosen as a case study to demonstrate the viability of the proposed concept. Simulation results illustrate the effectiveness of this controller as an online automatic generation controller (AGC) for a two area system, with and without generation rate constraints (GRC).