In view of the characteristics of multi-agent and strong coupling of regional integrated energy system(RIES)

a multi-agent bi-level game interaction strategy consisting of energy suppliers

distribution networks

and users is proposed. The game strategy includes two aspects: dispatch and bidding. According to energy supplier's bidding electricity price

multi-energy loads are forecasted. With the aim of minimizing system costs

the dispatch department coordinates all adjustable resources which belong to different participants. The cooperative game realizes the purpose of multi-energy complementary. Under the assumption of incomplete information and bounded rationality

this paper designs the energy supplier bidding strategy and corresponding clearing mechanism. On this basis

according to the history of scheduling results and unit characteristics

the non-cooperative bidding strategy of day-ahead energy markets with cold

thermal

and electric energies is simulated to pursue their own profits. And the Q-Learning algorithm is used to optimize the multi-agent bidding strategy which is then reported to the dispatch department. The evolutionary law of dispatch-bidding bi-level game of RIES is studied using a case study

and the local Nash equilibrium of the strategy is analyzed.