Voltage Regulation in Active Distribution Network with Multiagent Deep Q-Learning Approach

Voltage and reactive power control via inverter-based distributed generators is necessary for distribution power network to mitigate voltage violations. Voltage control is vital to keeping distribution power system voltages within normal range, minimizing power network losses and reducing losses of voltage regulating devices. To deal with incomplete and inaccurate distribution power system models, voltage control problems are solved in a model-free way by using a multi-agent deep reinforcement learning algorithm(DRL). The voltage control problem has been transformed in power system into an deep Q-network(DQN) framework, which avoids solving a specific optimization model directly due to time-varying operating conditions. The designed reward function guides these agents to interact with the distribution power system in the direction of normal voltage interval.