The accurate prediction of heating load in winter is of great significance to the safe and stable operation of the power grid and the peaking and frequency regulation of the power system. In order to fully exploit the implicit information in winter heating load data, a multivariable winter heating load forecasting method based on Bayesian network is proposed. Firstly, the multivariable data in implicit information data is divided into observable data and implicit data. Bayesian network structure is built based on the influence mechanism between variables, and observable data information is trained by EM (Expectation Maximization Algorithm) to obtain hidden data distribution, and then realize heating load forecasting based on observable data and implicit data. Using the measured data of winter heating load in an area provided by Beijing Power Grid to verify, the results show that the use of Bayesian network for heating load forecasting has higher prediction accuracy.