Multiple Wavelet Convolutional Neural Network for Short-Term Load Forecasting

Although the accuracy of load forecasting has been studied by many works, actual deployability a model is rarely considered. In this work, we consider from four aspects: 1) prediction performance model; 2) robustness 3) dependence on external data; and 4) storage size model. From these aspects, propose multiple wavelet convolutional neural network (MWCNN) for forecasting. On two public data sets, verified MWCNN. The MWCNN only uses data, 497 kB, which shows that good deployability. addition, our results are interpretable. experimental show can effectively capture periodic characteristics data.