Power Grid Restoration Partitioning Method Based on Reinforcement Learning

Reasonable partitioning is the core of parallel restoration for power systems. Reinforcement learning is a machine learning method that has been applied in solving a series of problems in power systems in recent years. Compared to traditional optimization algorithms, reinforcement learning can self-learn and continuously optimize the decision-making process through interaction with the environment, thus showing certain advantages in dynamic and complex power systems. This paper proposes a parallel restoration partitioning method based on reinforcement learning. The restoration partitioning process is modeled as a Markov process, and the Deep Q-Network algorithm is used to solve the problem. Based on this method, the agent is able to automatically learn and optimize the restoration partition strategy according to the operating status of the power grid, thereby improving the efficiency and stability of grid restoration. Simulation on the IEEE 39-bus system shows that the proposed method obtains reasonable partitioning result with high modularity.