The invention discloses a multi-factor electrical load prediction method based on deep learning, and the method comprises the steps: firstly completing the acquisition and storage of data, including electrical load data and environmental influence data; preprocessing and standardizing the data based on abnormal data detection, autoregression interpolation and sequence data normalization of a k-proximity algorithm and an improved DBSCAN algorithm; then, propsing an improved CNN-LSTM electrical load prediction model, and firstly using a CNN feature extraction module to learn local features of input data; inputting the input data into an LSTM sequence learning model, and extracting sequence feature information of the input data; meanwhile, introducing a self-attention mechanism into the LSTM for learning features of a hidden layer of the LSTM, and extracting key features by distributing different attention weights, so that the final prediction precision is improved; and finally, predicting the electrical load. According to the invention, digital upgrading of a power grid can be promoted, personalized requirements of users are met, and industry correlation analysis, power generation dispatching, power consumption trend prediction, work and production resumption guidance and the like are realized.