Electric power spot price prediction method, device and system

The invention discloses an electric power spot price prediction method, device and system. The method comprises the following steps: acquiring to-be-predicted data; a price prediction result of the to-be-predicted data is obtained through a pre-trained price prediction model, and the price prediction model is obtained after a neural network is trained to be converged through multiple sets of preprocessed data; and obtaining an electric power spot price prediction result according to the price prediction result. According to the method, targeted data enhancement is carried out on peak and valley area data of historical electric power spot prices, so that the accuracy of electric power spot price prediction, especially the peak and valley area electric power spot prices, is effectively improved; meanwhile, in combination with holiday and festival features and trend features, data dimension increase, fusion and dimension reduction processing is performed on training data of the price prediction model, and the ca