Short Term Power Load Forecasting of Large Buildings Based on Multi-view ConvLSTM Neural Network

In this paper, the Multi-view ConvLSTM Neural Network based on multi view convolution is proposed to solve the problems of low accuracy and poor stability of power load forecasting in large buildings. The historical data of multi-dimensional feature sequence is selected and standardized. The Multi-view convolutional neural network is used to extract features, and then the ConvLSTM neural network is used to further extract features from the feature map. The pre-processed historical data are trained and the trained model is used for forecast. The Multi-view ConvLSTM Neural Network can not only extract the information along the time series, but also extract the implicit semantic information between different feature sequences. Moreover, it is very little affected by the amount of data, so as to improve the stability and forecast accuracy of the model. In the process of model training, a variety of overfitting inhibition strategies are used, including setting the timely stop callback, learning rate attenuation, Dropout, batch normalization, etc.