Short-term load forecasting method based on deep learning under digital driving

Abstract

Relying on the background of power grid digital drive, this paper uses the improved deep learning network model to analyze and predict the load energy consumption of complex systems. In order to provide a complete and reliable sample data set for the multi-layer network model, this paper uses normalization, mutual information and other methods to preprocess the data set, reduce the correlation among different data; At the same time, based on the error reciprocal method, the bidirectional long and short term memory network is combined with the XGboost network model to reduce the calculation error of the model. The simulation experiment is used the actual data set of a city in southern China. The result proves that the index MAPE of the Bi LSTM XGboost forecasting method is 6.15, which can realize the accurate load forecasting of the actual complex system.