The passenger flows of the metro station have significant impact on the energy-saving operation of the air-conditioning system. In order to improve the prediction accuracy of the personnel cooling load in the subway stations, a prediction model was proposed in this paper using the BP neural network(GA-BP) method optimized by the genetic algorithm, and the personnel cooling load of the subway stations was calculated dynamically. The initial weights and thresholds of BP neural network were optimized by using the genetic algorithm, and the nonlinear learning ability of BP neural network was improved. The model was validated by the actual operation data, and the simulated results were compared with the prediction results of the traditional BP neural network method. The results show that the proposed method can effectively improve the nonlinear learning ability of BP neural network and the accuracy and stability of the hourly personnel cooling load prediction. By comparing with the traditional BP neural network method, the average daily personnel cooling load prediction error of the GA-BP model is reduced by at least 10%, and the fitting correlation coefficient value of daily hourly personnel cooling load prediction is increased by at least 0.1.