Considering the uncertainty of the initial state and the diversity of parameters of decentralized air-conditioning loads participating in regulation

a new clustering method and a dynamic optimization method of direct load control(DLC)applied to massive decentralized air-conditioning loads are proposed. In the process of clustering

the regulation potential of air-conditioning is set to be the clustering objective in order to fully utilize its regulation ability based on guaranteeing the operational space of DLC. During the DLC decision procedure

the scheduling error caused by model errors and temperature prediction errors can be corrected by rational selection of optimization cycle. Thus

the reducing power of air-conditioning load can perform more agreeably with the dispatching plan

while users' comfort requirements are ensured. Finally

the effect of the proposed method on fully releasing the regulation potential of decentralized air-conditioning loads is verified through simulation cases. This work is supported by National Natural Science Foundation of China(No. 51577028)and Fundamental Research Funds for the Central Universities(No. 2242016K41064).