Demand response (DR) resources in low-voltage power side are numerous and scattered, which lead to low accuracy of DR performance. Firstly, a generalized analysis and evaluation model of users’ DR potential is established. Furthermore, a decentralized cooperative appliance-level DR framework and its DR algorithms are proposed. The DR signal is composed of DR level and reduction ratio. The signal is generated and sent to each DR participant’s energy information gateway (EIG) by DR center, the EIGs then make decisions to maximize DR according to the received signal and user preferences. Based on this framework, algorithms for both DR center and EIG are proposed. Finally, in order to analysis the performance of DR control framework and control accuracy, a case with 100 EIGs DR group is studied. The case study shows that the proposed framework and algorithms are feasible and effective. © 2020, Chinese Institute of Electrical Engineering. All rights reserved.