Accurate load forecasting is of great significance for improving the level of grid planning and accurately guiding investment. In view of the shortcoming of over-fitting in the combined forecasting model of empirical risk minimization, a combined forecasting model based on social learning multi-objective particle swarm optimization algorithm was proposed in term of partial least squares regression model, support vector regression model and grey prediction GM (1, 1) model. The uncertainty function information entropy of weight was introduced to represent the expected risk, and the empirical risk and expected risk were comprehensively considered in the model. The simulation results show that the proposed method has higher prediction accuracy than the single forecasting model and the other two combined forecasting models, and the social learning multi-objective particle swarm optimization algorithm has stronger global search ability and optimization performance.