Self-Healing Method for Active Distribution Network Based on Hierarchical Reinforcement Learning

With the development of distributed generation (DG), fault self-healing technology in active distribution network (ADN)has also become a focus of research. This article proposes a fault self-healing method based on hierarchical reinforcement learning (HRL). Firstly, by obtaining the topology, network parameters, and DG parameters of the ADN, a simulation model of the ADN is constructed. Subsequently, a reinforcement learning framework for real-time control of ADN was established, and the self-healing strategy for distribution network faults was modeled based on a hierarchical decision-making mechanism. Afterwards, a layered objective reward function is modeled to generate reward signals based on the operational status of the distribution network; Using HRL algorithm to train in a distribution network simulation environment, a fault self-healing strategy model is obtained. Finally, based on the actual fault line information and the fault self-healing strategy model, a distribution network fault self-healing strategy is generated to achieve self-healing of the ADN. The self-healing method is evaluated on the IEEE 33-node system, which indicate its rationality and effectiveness fully verified.