Research On Short-term Load Forecasting Of Power System Based On Deep Learning

With the rapid development of the power Internet of Things,the power system is transitioning to a more intelligent and flexible interactive system.Power system load forecasting will play an increasingly important role in future grid planning,power demand side management,and power enterprise operations.Accurate short-term load forecasting(STLF)can effectively guide the combined dispatching of generator sets,formulate maintenance plans,reduce power generation costs,increase economic benefits,maintain the safe and stable operation of the power system,and make reasonable arrangements Electricity market operation.At the same time,environmental pollution and energy crises have become more and more serious,making my country’s power grid urgently in need of innovation in dispatching and operation.And accurate short-term load forecasting is a necessary condition for the optimization of power grid dispatching.Therefore,how to improve the accuracy of short-term load forecasting is the main content of this paper.This article considers the internal laws of historical load itself and the external influencing factors of historical load.On the one hand,the Pearson correlation coefficient is used to analyze the correlation between the external factors of historical load,and the relevant characteristics are screened out.Then,the BIGRU model is introduced to construct a deep learning model that considers the multi-factor BIGRU network to improve the load forecast accuracy.The experimental results show that BIGRU,the model has a better prediction effect than MLR,SVR,LSTM and GRU models.On the other hand,by introducing variable modal decomposition technology to fully mine the inherent information of historical load,and combining the powerful feature extraction capabilities of convolutional neural networks,a short-term load forecasting method based on VMD-CNN-BIGRU hybrid network is proposed.The experimental results show that the prediction method based on the VMD-CNN-BIGRU hybrid network has a further improvement compared with the BIGRU model in terms of accuracy,prediction effect or computational resources.