Study On Refined Model Of Electricity Forecast Based On Industrial Electricity Characteristics

With the industrial transformation of China,the fluctuation of consumption in various industries has intensified,which has led to an increase in the prediction error of many models with the total social electricity consumption as the forecast object.At the same time,the development of society requests higher demands on the accuracy of electricity forecasting.However,the existing industry division standards cannot meet the requirements of electricity forecasting,and need to be redesigned on the basis of existing industries.This paper studies on three levels with four chapters from the optimization of electricity industry classification,optimization of existing prediction algorithms and optimization of prediction results to achieve refined electricity prediction.Because the existing domestic and international industry classification standards do not meet the electricity forecast by industry division,an electrical characteristics index system for electricity consumption prediction is designed.The indicator system is based on the historical electricity data of the existing industry,and is classified based on three aspects: peak-valley characteristics,monthly electricity time series and annual electricity time series.Compared with the current industry division results,the new classification results can effectively classify industries with similar electricity consumption characteristics with a smaller number of classifications.This classification scheme can be applied to the electricity forecasting in the following sections.In order to realize the refinement prediction of the electricity in each sub-sector,an electricity combination forecasting model based on the electrical characteristics of industries and the improved entropy weight method is proposed.The model is based on the classification scheme of the previous chapter.At the same time,the improved entropy weight method is applied to correct the problem that the entropy weight disorder result from the distribution of multi-entropy value tends to 1.The weights of the multi-error index evaluation index and the single algorithm of combined model are re-allocated to establish the optimized combination forecasting model for various types of electricity industry.The example shows that the model has good prediction accuracy.Further consideration is taken on the refinement prediction with the influence of external factors,a electricity prediction model based on EMD method and improved GM(1,N)algorithm is designed.The model selects the indicators used to establish the GM(1,N)model,and decomposes the industrial electricity curve that is affected by the complex factors into a number of relatively simple,regular,and low-frequency components.And an improved GM(1,N)model is established based on a linear correlation coefficient and a gray action coefficient,which corrects the defects of the traditional model in terms of modeling principle,parameter setting and model structure,making it more reasonable and stable;Superimposed component prediction results are used to obtain refined prediction results considering various external factors in various types of industries.The example shows that the model can complement the model of the third chapter,and the industry sensitive to external factors can further improve the accuracy of electricity forecasting,thus improving the forecasting accuracy of total power.For the problem that the total forecast results and the sum of component forecast results are inconsistent in the industry dimension and time dimension,the multi-level coordination model is used and a two-level coordination model of the two dimensions is constructed.The feasibility and effectiveness of the multi-level coordination model are verified by an example,and the prediction results are optimized to achieve refined prediction.