Voltage control in power distribution networks via distributed reinforcement learning

Abstract

This thesis investigates voltage control in distribution networks through a multi-agent reinforcement learning framework. Using the IEEE 13-node feeder and OpenDSS as the simulation platform, a multi-agent voltage control environment is constructed in which multiple PV units act as controllable agents. The voltage regulation problem is formulated as a sequential decision-making task, and the MADDPG algorithm is adopted for controller training. The results show that the proposed method can achieve stable training convergence and learn an effective coordinated voltage control policy. Under the learned strategy, the monitored bus voltages remain within the allowable range, voltage violations are significantly reduced compared with the uncontrolled case, and the control actions remain smooth and moderate. The study verifies the feasibility of applying distributed reinforcement learning to voltage regulation in active distribution systems and provides a simulation reference for intelligent voltage control in distribution networks with high penetration of distributed energy resources.