Short-term Power Load Prediction Based On SSA-LSTM Model

Power load reflects the development level of both the country and industrial technology.It can not only plan the future power distribution,but also give early warning of regional power load,of the fully forecasting and analyzing power load data;which effectively reducing the waste of power resources.Therefore,it is of great significance to carry out power load forecasting.In view of the influence of various factors such as meteorological reasons,seasonal reasons,and algorithm optimization reasons on short-term power load forecasting,this paper proposes a short-term power load forecasting model based on an improved longshort-term memory neural network,as well as verifies and compares according to various of factors.The research contents of this paper are as follows:(1)Pre-processing the actual power load data.The directly collected data may have various abnormal states such as missing data and abnormal data.Data pre-processing includes detection of missing data,filling of missing data,and data normalization.Afterwards,the complete power load data and meteorological data are correlated and screened out for appropriate characteristic variables as input data.(2)The LSTM model is selected as the prediction algorithm in this paper.After comparing and verifying LSTM,Least Squares Support Vector Machine(LSSVM)and Ellman Neural Network(ELMAN)under the same conditions,LSTM with smaller error was selected.The principles of the three models are introduced,and the final data results and image results are summarized to visually find the gap.(3)The optimization method based on LSTM model is determined.Aiming at the problem of optimal parameters of LSTM,the sparrow search algorithm(SSA)is selected as the parameter optimization algorithm of LSTM model,and a short-term power load forecasting model based on SSA-LSTM is constructed.Two optimization methods of sliding window width adjustment and vertical sampling days adjustment,are carried out for the SSA-LSTM prediction model,and the optimal width and sampling days are obtained through analysis.According to the short-term power load forecast results in different seasons,the influence of seasons on short-term power load is analyzed.Finally,the error of the optimized SSA-LSTM model is the smallest by comparing and verifying the optimized SSA-LSTM model under the same conditions.Based on the above data pre-processing,model selection and verification comparison,parameter optimization algorithm selection,and model optimization,the prediction model in this paper is selected,which has the characteristics of higher accuracy and smaller error.Taking into account the relationship between meteorological factors and load data,as well as being able to display the change of load in each season and analyze it to give corresponding conclusions.Therefore,the power load forecasting model selected in this paper can complete the short-term power load forecasting much better.Figure[55] Table[17] Reference[43]...