Decentralized reinforcement learning collaborative consensus algorithm for generation dispatch in virtual generation tribe

Abstract

The article proposes a distributed reinforcement learning collaborative consensus algorithm for dynamic generation command dispatch of AGC in interconnected power grids under the framework of the virtual power generation tribes, in order to in response to the development of the EMS system in the Smart Grid from centralization to decentralized form. The simulation results of the Guangdong Grid show that: the algorithm can not only enhance the adaptive and dynamic performance of the system but also can reduce the adjustment cost as well as realizing the optimal allocation of automatic generation control.