Research On Domestic Power Load Forecasting Based On EEMD And Optimized LSTM

With the rapid development of smart grid and the continuous expansion of power supply scale,power load forecasting has become an essential work to ensure the development of power system planning.Accurate prediction is of fundamental significance for the economic and stable operation and efficient dispatch of power system.Scientific prediction is the theoretical basis for reasonable planning and correct decision-making of power consumption.Therefore,it is very important to find an effective short-term load forecasting method to improve the forecasting accuracy.In the work of domestic power load forecasting,due to the existence of various influencing factors,it will lead to the fluctuation and instability of load data series.The EEMD algorithm is used to decompose the historical load data into a finite number of IMF and a residual component.LSTM has the advantages of fitting complex data and time correlation analysis of complex data.The data of domestic electricity load is affected by many complex factors,which has strong randomness.Using a single prediction model for prediction,the prediction accuracy is low and the prediction time is long.In order to solve the problem of high-dimensional data,AE automatic encoder is used to extract the features of influencing factors,obtain their high-order feature representation,and mine their deep feature factors.In order to optimize LSTM model,reduce data dimension and data interference,an optimized LSTM model of automatic encoder is established.This paper presents an EEMD-AE-LSTM algorithm combining EEMD,Auto Encoder and LSTM for short-term load forecasting of residential electricity consumption.Input the influencing factors of domestic power load data,and extract the features to get the feature sequence.Different training matrices are formed by integrating the data components of EMD.The LSTM model is used for prediction,and the final prediction result is obtained by superposition prediction.Finally,through experiments,the EEMD and AE optimized LSTM combined forecasting model is compared with LSTM forecasting model,AE optimized LSTM forecasting model and EEMD and LSTM combined forecasting model,and the performance indexes of MAPE and RMSE are used to verify the analysis.MAPE decreased by 7%,5.4% and 4.1%,while RMSE decreased by 14.52%,8.19 and 6.88%,respectively.Compared with the other three prediction models in this experiment,it has a great improvement.The experimental results show that the combined optimization algorithm is effective and the accuracy of load forecasting is significantly improved.The combination method of domestic power load forecasting proposed in this paper has strong generalization performance.At the same time,it has the ability of deep data mining,which is a more effective short-term load forecasting method.