Abstract: Air-conditioning load, as an important load in power grid, occupies a large proportion of load peak demand. With the capability to convert electrical energy into thermal (cold) energy stored indoors, the demand response potential of air-conditioning load in power grid has been explored. By analyzing the state-queue change of the traditional air-conditioning control strategy and considering consumers comfort, the limit downtime is calculated. On the basis of this, a strategy to control air-conditioning load by group is proposed, which reduces temperature setting workload of the control center in the case of massive air conditioners. In dealing with the “recovery peak”, the fluctuating peak value of load recovery is further reduced by changing the temperature interval length in different time intervals. Finally, the evaluation Indicators and method of air-conditioning control are put forward. The simulation results show that the improved method has better peaking curtailment performance, less fluctuation of rebound load and less computation cost of temperature set value.