•  
  •  
 

Keywords

improve NashQ, multi-agent reinforcement learning, multi-agent game, wind power dispatch, electric vehicle station

Abstract

Aiming at the problem of renewable energy gaming in power generation market, this paper studies wind power dispatching strategies in different scenarios, and proposes an improved NashQ wind power dispatching strategy. Firstly, a wind power optimal dispatch model under a game environment is established. In the dispatch model, the punishment for wind power forecast error, the environmental and economic benefits of wind power, and the cost of the curtailment of renewable energy are all considered. On this basis, the dispatch strategies are compared for the independent wind power operation mode, wind-vehicle operation mode, and wind-storage joint operation mode. Secondly, The Jensen-Shannon divergence is introduced for the learning rate of the intelligent agents. The convergence efficiency of multi-agent reinforcement learning is then improved. Finally, a microgrid model is constructed in Matlab for simulation. It is shown that the improved NashQ method has a significantly higher convergence speed than the NashQ and NETRL algorithms, and the wind-vehicle joint operation model has a better performance in the market games.

DOI

10.19781/j.issn.1673-9140.2022.06.007

First Page

62

Last Page

72

Download

Share

COinS