Energy storage system（ESS）in the micro-grid can improve the reliability and economic in power grid.However

in the planned or unplanned islanding operation mode

due to the randomness and intermittent characteristics of the output of various types of distributed generators（DG）

reasonable use of multiple storage units within the micro-grid can optimize the power flow and economic cost. This paper is based on the multi-agent system controlling model

and studies the dynamic game model in the island within the micro-grid operation mode considering various distributed energy resources（DER）and load

while the micro-grid includes several ESSs. Including energy-price elasticity of the demand response

Agent electric quantity bidding model based on risk cost

and price bidding decision based on VRE reinforcement learning algorithm are established. In the end

the author combines the mentioned models and methods with the multi-agent simulation platform

to analysis of the validity of the proposed model.