Short-term load forecasting method with variational mode decomposition and stacking model fusion

Abstract

An accurate load forecasting method is critical to the distribution networks to match the supply with demand. In this paper, Variational Mode Decomposition (VMD) and Stacking model fusion are composed to obtain a short-term load forecasting method for real-time power dispatch. Firstly, a VMD algorithm decomposes the load series into dissimilar intrinsic mode functions (IMF), and the Approximate Entropy (ApEn) of each IMF is calculated to produce corresponding new components. Secondly, eXtreme Gradient Boosting (XGBoost), Long Short-Term Memory (LSTM), Support Vector Machines (SVM) and K-Nearest Neighbour (KNN) are used as basic models to predict each IMF. Thirdly, the data fusion problem, which is always oversimplified treated, is solved under the Stacking integration framework. The final prediction results of these basic models are obtained by an ensemble learning method. The prediction results of each component are superposed and then a weighted fusion method is carried out. It is demonstrated that the component estimation is well fused using the Stacking model fusion method. In comparison with the experimental results of XGBoost, VMD-XGBoost, and KNN methods, the proposed method can significantly improve the accuracy.