Multi-channel LSTM-CNN power load multi-step prediction based on feature fusion

Abstract

In order to fully consider the influence of uncertain factors such as complex and changeable weather conditions and social events, a multi-channel LSTM-CNN (Long Short Term Memory-Convolutional Neural Network) power load multi-step forecasting model based on feature fusion is proposed for the first time, by modeling the sequential data with different time scales and using the neural network model composed of multiple LSTM networks in parallel, the multi-scale time feature representation is learned, the output time step of each LSTM is convoluted by CNN, and its output characteristics are extracted.