A Multi-Objective Reinforcement Learning Active Distribution Grid Load Equivalent Modeling Method Based on Evidential Reasoning

With a large number of adjustable loads and new energy sources connected. Distribution networks are increasingly in the form of active distribution networks. This situation introduces more stochasticity and uncertainty into the active distribution network load equivalence modeling problem. At the same time, in the case of high stochasticity and time-varying, the method of constructing the load equivalence model needs to consider multiple influencing factors. Therefore, this paper proposes a multi-objective reinforcement learning active distribution grid load equivalent modeling method based on evidential reasoning. This method is divided into five steps. First, a multi-attribute analysis is performed to identify the set of evaluation indicators, the set of candidate solutions, the set of evaluation ratings, and the set of evaluation criteria set, candidate solution set, and evaluation level set. The confidence evaluation vector is subsequently obtained. In the second step, multi-evidence fusion is performed after obtaining confidence evaluation vectors of candidate solutions. The approximate degree of superiority of the candidate solutions can subsequently be obtained. In the third step, utility analysis is used to compare the average utility values of different candidate solutions and select the final candidate. In the fourth step, the active distribution network load equivalence modeling process is modeled as a Markov decision process. In the fifth step, the scenarios determined in the third step are brought into the fourth step to solve for the key parameters of the isotropic model. Simulation results show that the method proposed in this paper can obtain the required load equivalence model for active distribution networks.