A power load forecasting method based on convolutional neural network (CNN) and long short-term memory (LSTM) networks is proposed. On the premise of completing data preprocessing including normalization processing and K-means clustering, first use the CNN network to extract the feature vector of the continuous feature map constructed by the load influencing factors, and then use the extracted feature vector to establish an LSTM model for load predict. Finally, through a variety of different types of load data calculation examples, the analysis verifies that the proposed CNN-LSTM hybrid network model has better load prediction performance than the BP neural network, the CNN model alone or the LSTM model alone.