Intelligent electric meter state evaluation method based on multi-agent reinforcement learning

The invention provides an intelligent electric meter state evaluation method based on multi-agent reinforcement learning, and belongs to the technical field of electric power system state evaluation. The problems of inaccurate state evaluation and poor model universality of a traditional intelligent electric meter are solved; comprising the following steps: preprocessing data of an intelligent electric meter; constructing a multi-agent reinforcement learning model: setting reinforcement learning agents, setting the state, action and decision of reinforcement learning, and setting a reward and punishment function of reinforcement learning; a decision matrix is initialized, proper actions are selected based on an epsilon-greedy algorithm, reward and punishment feedback of the environment to the agents is obtained, the states of the agents are updated through the reward and punishment feedback, and whether all the agents obtain optimal control strategies or not is judged; inputting the preprocessed data into the