Load Demand Prediction based on Improved Algorithm and Deep Confidence Network

Abstract

To address the issue of inaccurate load forecasting amidst the advancing smart grid technology and the widespread integration of various demand-side resources like controllable loads, distributed energy sources, and energy storage, the author proposes a deep confidence network based on improved algorithms for load demand forecasting. Firstly, the VMD algorithm is used to decompose the load data into different intrinsic mode functions (IMFs), Then combine the DBN network to predict each IMF, Finally, overlay the prediction results of each part to obtain the prediction results of the VMD-DBN model. The experimental results indicate that: The PSO-DBN model has good prediction results and fast convergence speed in power load forecasting. The MAPE is 1.03%, and the RMSE is 9.35MW, which verifies that the method has good prediction accuracy. Compared to the single use of DBN method and the combination of Empirical Mode Decomposition (EMD) DBN method, the proposed method by the author has a significant improvement in prediction accuracy.