Air conditioning load has become one of the main loads of summer electricity consumption in China, and chiller is the largest part of the central air conditioning system power consumption. For a system with multiple units operating together, each chiller often operates under partial load. Based on the user-side demand response, the quaternary polynomial regression model of COP-PLR is established by polynomial regression method. In order to minimize the operating energy consumption of the chiller, the objective function is established, and the control strategy of optimizing PLR of the chiller with improved hybrid genetic algorithm is studied.The overall COP of the chiller is improved by optimizing PLR of each chiller.Compared with the original control method, the typical daily energy consumption is reduced by 702 kW, and the energy saving rate is 8.48%, which verifies the feasibility of the optimization method.