基于鲁棒深度强化学习的考虑输电网电压波动的多馈线配电网分布式电压控制

Abstract In the multi-feeder distribution network , the power balance between photovoltaics generations and load demands across regions is more complex. To solve the above problems, this paper proposes a multi-agent distributed voltage control strategy based on robust deep reinforcement learning to reduce voltage deviation. The whole multi-feeder distribution network is divided into a main agent and several sub-agents, and a multi-agent distributed voltage control model considering the transmission network voltage fluctuations and the corresponding power fluctuations is established. Based on the information uploaded by sub-agents, the main agent models the uncertainty of the transmission network voltage fluctuations and the corresponding power fluctuations as a disturbance to the state, and a RDRL method is employed to determine the tap position of on-load tap changer. Furtherly, each sub-agent uses the second-order cone relaxation technique to adjust the reactive power outputs of the inverters on each feeder. The effectiveness of the proposed method has been verified in two real-world multi-feeder systems. The results show that the proposed method can achieve millisecond-level decision-making, with a voltage deviation only 1.28 % higher than the global optimal results, achieving near-optimal control. The proposed method also demonstrates robustness in handling transmission network uncertainties and partial measurement loss.