Abstract: In order to fully mine the effective information contained in a large number of collected data and improve the accuracy of short-term load forecasting, a short-term load forecasting method is proposed based on a hybrid model of convolutional neural network (CNN), bidirectional gated recurrent unit (BiGRU), and fully connected neural network (NN). The massive historical load data, meteorological information, and date information are taken to construct feature maps according to time sliding windows. Firstly, the CNN is used to extract valid information from the feature maps to construct feature vectors. And then, by taking the feature vectors as the inputs, the BiGRU-NN network is used to make short-term load forecasting. The load data in the test question A of the Ninth National Electrical Mathematics Modeling Contest held in 2016 are taken as an actual computation example, and the experimental results show that this method has higher accuracy in short-term load forecasting than GRU neural network, DNN neural network, and CNN-LSTM neural network.