基于深度强化学习的分布式电源就地自适应电压控制方法

Adaptive Local Voltage Control Method for Distributed Generator Based on Deep Reinforcement Learning

高比例分布式电源的广泛接入引起了配电网电压波动问题,分布式电源换流器能够提供连续的无功功率支撑,是实现快速电压调节的潜在解决方案。针对高比例分布式电源自适应电压控制问题,文中提出了基于多智能体深度强化学习的分布式电源就地电压控制框架。配电网各区域通过构建深度强化学习智能体以实时感知配电网状态,制定分布式电源运行策略,自适应地应对电压波动。然后,考虑分布式电源换流器功率耦合问题,基于动态边界动作掩模机制设计以保证智能体动作的有效性。最后,采用IEEE 33节点及中国南方电网53节点算例验证了所提方法的可行性与有效性。

The large integration of high-penetration distributed generators(DGs)aggravates voltage fluctuations in distribution networks.DG converters are capable of providing continuous reactive power support and are a potential solution to achieve fast voltage regulation.Aiming at adaptive voltage control of high proportional DGs,this paper proposes a framework for local voltage control of DGs based on deep reinforcement learning of multiple agents.Each region of the distribution network is built with deep reinforcement learning agents to sense the state of the distribution network in real time,formulate DG operation strategies,and respond to voltage fluctuations adaptively.Then,considering the DG converter power coupling problem,a dynamic boundary action mask mechanism is designed to ensure the effectiveness of agent actions.Finally,the feasibility and effectiveness of the proposed method are verified by using IEEE 33-bus system and 53-bus system of China Southern Power Grid.