Application Of Improved Grey Model In Load Forecasting Of County Distribution Network

Electric load forecasting is the basis and boundary condition for the research of power grid planning and the important basis for investment decision of power grid.Throughout the development of electricity and load in China,it has not only the stability of increasing year by year,but also the instability of random fluctuation.It is a typical grey system,which can be predicted by grey model.However,with the continuous connection of various distributed generations and new energy sources,it is possible to use grey model to predict power load.In order to achieve the goal of relatively simple,high precision and high practicability,it is necessary to improve the traditional grey prediction model.In this paper,the basic principle and data processing method of grey theory are introduced in detail,and the principle,modeling process and error analysis of GM(1,1)model based on grey theory are introduced.The advantages and disadvantages of the model are analyzed,and the improvement ideas of the model are given.In view of the weakness of the original model in noise reduction,an improved method combining data pretreatment with residual correction based on Fourier transform is studied.The prediction error is optimized by using the strong noise reduction ability of original data smoothing and Fourier transform.In view of the weakness of the original model in medium and long term load forecasting,the data location using equal dimension innovation information is studied.Logically,the oldest data is replaced by the new predicted data,and the dimension of the data is maintained unchanged.The predicted data are repeated until all the predicted results are obtained.Finally,based on the original data of Longkou's total social electricity consumption from 2006 to 2015,the forecast values of 2016 to 2018 are given by the above methods,and the prediction accuracy of the improved model is verified by the actual values of 2016 to 2018.Finally,the forecast values of 2019 to 2025 are given by using the recommended method based on the data of the whole social electricity consumption from 2006 to 2018.