Research on Distributed Renewable Energy Transaction Decision-making Based on Multi-Agent Bilevel Cooperative Reinforcement Learning

Abstract

With more and more distributed renewable energy connecting to the distribution network, the issue of regional transaction of distributed renewable energy has attracted more and more attention. In order to adapt to the complex transaction decision-making problem, this paper proposes a multi-agent bilevel cooperative reinforcement learning algorithm under the framework of bilevel stochastic decision-making model. By constructing a bilevel stochastic decision-making optimization model for distributed renewable energy trading, the uncertainties and fluctuations of distributed generation output are effectively solved. The objective of upper level planning is to maximize the profits of distributed renewable energy generators. The lower level planning is to optimize the dispatch of the whole regional market. The two layers are continuously iterated until the lower level planning is optimal, that is, the comprehensive benefit is maximized.After introducing multi-agent bilevel cooperative reinforcement learning, the algorithm can effectively carry out learning training, and after completing the training, it can quickly and accurately calculate the optimal results. Through the simulation of the model project of Guizhou Hongfeng area, the bidding decision algorithm has been verified, which can improve the profit of the power producer while taking risks into consideration, and at the same time maximize the comprehensive benefits.