Power load prediction method and system based on combination of transfer learning strategy and multiple channels

The invention discloses a power load prediction method and system based on combination of a transfer learning strategy and multiple channels, relates to the technical field of power load prediction, and is characterized in that a multi-channel CNN-BiLSTM model is combined with an improved hierarchical transfer learning strategy. According to the multi-channel CNN and BiLSTM combined model, the advantages that the CNN extracts data local features and the BiLSTM hidden layer weight back propagation gradually optimizes training weights are fully utilized, and compared with a traditional single-channel model, a multi-channel model has the advantages that convolutional layers connected in parallel use convolution kernels of various sizes to learn features of different details; the influence of characteristics and the time sequence change of data are fully considered. The improved hierarchical transfer learning strategy can improve the prediction precision of the model when small-scale data samples are insufficient