Abstract: For the realization of large-scale air conditioning (AC) load to participate in demand response, a decentralized multi-agent cooperative control strategy based on leader-follower consistency was proposed. A hierarchical architecture that can effectively control large-scale ACs and a topology model of multi-agent related communication network were constructed. The demand side load reduction cost of the grid company was a quadratic function for the user's load reduction. A multi-agent leader-follower decentralized coordination algorithm based on incremental cost of load response was proposed to meet the requirement of precise load control for large quantity, location decentralized ACs. Finally, case studies were given to show that the proposed method can achieve the optimal state of consistency under different load conditions and agent topology. The factors that influence the rate of convergence were also discussed. In addition, an example was given to simulate the overall peak load regulation of large-scale ACs from top to bottom.