Short-term Power System Load Forecasting Method Based On Improved Deep Recurrent Neural Network

Short-term load forecasting is of great significance to the safe,reliable,and economical operation of the power system.Improving the accuracy of load forecasting can provide a better basis for operation,planning,and protection of the power system.At present,the load status of the power system has become more and more complicated with the increase of load diversity,and the random volatility has gradually improved.In a modern society that requires higher and higher forecasting accuracy,the problem of limited accuracy of traditional load forecasting algorithms has become more significant.This article focuses on the research work of the short-term power system load forecasting model based on the improved deep recurrent neural network.Taking the public data sets of Australian energy market operators as the research object,by comparing with other models,we will discuss its applicability,and mainly focus on the following two issues:First of all,in order to make up for the shortcomings of traditional time series algorithms in prediction accuracy and give full play to the advantages of traditional algorithms for rapid prediction,this paper combines the improved generalized autoregressive conditional heteroscedasticity algorithm with the long and short-term memory network from the perspective of algorithm complementarity,and obtains a fast algorithm for short-term load forecasting.In the process of constructing the combined algorithm,the optimal parameters are obtained by grid search.The results show that the proposed model has strong generalization ability and stability,and the prediction accuracy is greatly improved compared with the model before the combination.In addition,this article also addresses the problem that a single recurrent neural network cannot extract the time series information of the power system well,it proposes that the variational mode decomposition can be combined with the improved multi-dimensional gated recurrent unit to give full play to the excellent information extraction of the variational mode decomposition.In the parameter selection of the model,this part is jointly determined by a combination of experience and grid search.The results of the calculation examples show that the proposed model effectively improves the accuracy of short-term power system load forecasting compared with common single algorithm and other combined forecasting methods.The two forecasting models proposed in the article and the prospects based on them have positive exploratory significance for constructing better short-term load forecasting models.