As a typical flexible and adjustable resource, multiple constraints such as user comfort, policy norms, and market income need to be taken into account when air conditioning loads are utilized for power grid peak shaving. Therefore, the regulation preference and differen-tial compensation mechanism are introduced, and a multi-objective optimization decision-making method for peak shaving of air condition-ing clusters is proposed, in which the sensitivity of user regulation is considered. Firstly, an air conditioning cluster load model is construct-ed, where the characteristics of regulation sensitivity and the equivalent thermal parameter model are integrated. Then, the peak-shaving framework of air-conditioning clusters with both simulation accuracy and regulation duration is built: a differential compensation mecha-nism, in which temperature regulation amplitude and regulation duration are considered, is established in the economic dimension, and a quantitative characterization method of user satisfaction based on an S-function is proposed in the user behavior dimension. Finally, with the goal of maximizing the economic benefits of aggregators and achieving the highest customer satisfaction, an optimal decision-making model for peak shaving of air-conditioning clusters is established, which is solved by combining the Pareto frontier solution and the knee-point decision-making method. It is demonstrated by the simulation results that the profit of the aggregator's peak adjustment can be signifi-cantly improved by the proposed method under the premise that the user's comfort is ensured, and at the same time, the reasonable distribu-tion of the user-side subsidy is realized.