Abstract:Power system load forecasting is the main basis of power system planning and economic policyformulation.However, the existing power system load forecasting combined forecasting methods, the prediction accuracy and efficiency are low; Aiming at this problem, this paper proposes a combined load forecasting method based on the deep belief network.First deep belief network training model is established, the nonlinear function relation between the combination data and the actual load data is applied to the training model.Through data training, deep belief network layer and the parameters areoptimized,making trained of the combineddeep belief network has the ability to predict. Using actual history data, the accuracy of the combined load forecast is calculated,.And the experimental results show that the proposed combination forecast method compared with traditional forecast methods, has high prediction accuracy, and its computational complexity is low.