Abstract: Virtual power plant（VPP）can integrate distributed energy resource（DER）to participate in the

operation of power market and auxiliary service market，and provide management and auxiliary services for

distribution network and transmission network. Its operation and control have been widely concerned. Aiming at the

virtual power plant containing electric vehicle（EV）charging stations，the Stackelberg game model of VPP and EV

charging stations was constructed based on soft actor-critic（SAC）algorithm and twin delay deep deterministic

policy gradien（t TD3）algorithm. By training the network parameters of Stackelberg game，the strategy and solution in game equilibrium was calculated. The calculation example results show that the model proposed can effectively reduce the operating cost and smooth power of EV charging stations after the completion of training，and the SAC reinforcement learning method can integrate the internal DER of VPP and guide the orderly charging of EV. When VPP participates in day-ahead power market as price taker，it can also give optimal trading strategy. When there is Stackelberg game between VPP and EV，EV can reduce charging cost by using deterministic strategy algorithm，while VPP can improve revenue by using stochastic strategy algorithm.

      Key words: virtual power plant（VPP）；SAC algorithm；TD3 algorithm；electric vehicle（EV）；Stackelberg

game；real-time dispatch