CNN-TCN day-ahead load prediction method based on novel decoupling thought

The invention discloses a CNN-TCN day-ahead load prediction method based on a novel decoupling idea. The load prediction method comprises the following steps: decoupling power load data into a load per unit curve, a daily average load and a starting point load after rotation alignment; a CNN-TCN day-ahead load prediction model based on the novel decoupling thought is constructed, and the prediction model comprises a convolutional neural network module, a time convolutional network module and a full connection layer; extracting shape features of the load per unit curve through the convolutional neural network module, and extracting time sequence features of the starting point load and the daily average load through the time convolutional network module so as to train the prediction model; and inputting external data into the trained prediction model to output a load curve of the predicted day. Compared with other prediction models, the method has the advantages that prediction errors are effectively reduced, a