Abstract: As temperature control load, the air conditioning load can convert electric energy into heat energy and store it, and the demand response potential is huge. In this paper, the thermal comfort model is introduced to accurately quantify the thermal comfort of air conditioning users in the control process. Secondly, peak shaving is selected as the air conditioning demand response scenario, and a two-layer scheduling architecture of air conditioning load is proposed. The upper architecture proposes a power company scheduling decision model to analyze the impact of aggregator quotes on power company decision-making. The lower layer architecture considers the user 's thermal comfort behavior, establishes the economic optimization strategy model of the aggregator, and maximizes the actual interests of the aggregator.