WRL: A Combined Model for Short-Term Load Forecasting

摘要

Load forecasting plays a vital role in economic construction and national security. The accuracy of short-term load forecasting will directly affect the quality of power supply and user experience, and will indirectly affect the stability and safety of the power system operation. In this paper, we present a novel short-term load forecasting model, which combines influencing factors analysis, Wavelet Decomposition feature extraction, Radial Basis Function (RBP) neural networks and Bidirectional Long Short-Term Memory (Bi-LSTM) networks (WRL below). The model uses wavelet decomposition to extract the main features of load data, analyzes its correlation with influencing factors, and then constructs corresponding adjustment factors. The RBF neural networks art; used to forecast the feature subsequence related to external factors. Other subsequences are input into Bidirectional LSTM networks to forecast future values. Finally, the forecasting results are obtained by wavelet inverse transform. Experiments show that the proposed short-term load forecasting method is effective and feasible.