In this paper, the heating system is used as a research object, and the prediction accuracy is not high due to temperature factors, random factors, and factors of the building itself in the central heating heat load. Therefore, this paper proposes to use the BP neural network algorithm for prediction, which has a good control effect on models with nonlinearity and can learn itself. However, due to the large fluctuation of BP neural network, local optimization phenomenon is relatively easy to occur.Therefore, on the basis of using BP neural network, the BP neural network and genetic optimization algorithm are improved and combined to make up for the deficiency of BP neural network. Finally, simulation experiments show that the error of heat loads prediction is greatly reduced,the prediction accuracy is improved, and reasonable heat supply is realized.