Treatment Method of Intercurrent Power Quality Problems in Low-Voltage AC/DC Hybrid Distribution Network

As the "carbon peaking and carbon neutrality" strategy and new power systems are progressively advanced, a significant portion of new energy sources and loads are integrated into low-voltage distribution networks. This integration makes it increasingly challenging to address concurrent power quality issues in real-time. Deep reinforcement learning, which benefits from "offline training and online deployment," presents a viable solution for managing real-time power quality issues in ADHDN (AC/DC hybrid distribution networks). This paper first establishes an MDP (Markov Decision Process) environment grounded in the Disflow calculation method. Subsequently, the "Deep Q-learning" reinforcement learning algorithm is employed to train an agent, enabling real-time management of concurrent power quality issues in low-voltage ADHDN. Finally, a case study involving a 21-node low-voltage ADHDN demonstrates the effectiveness of the proposed method in handling real-time concurrent power quality issues.