Medium And Long Term Power Load Forecasting Based On Fractional Grey Model

Power load prediction is concerned with people’s livelihood of economy.Reasonable power load prediction can save national expenditure and have a positive effect on the long-term development of Chinese power system,improve investment efficiency of state-owned enterprises,meet the demand of various development of electricity,and is of great significance to ensure the demand of electricity in production and life of urban and rural residents.First of all,this paper introduces the basic principle,development and research status of grey model and fractional theory.Grey model conforms to the law of medium and long term load forecasting and is a common algorithm in the field of medium and long term load forecasting.High-precision prediction model is of great significance for load prediction.However,the fitting mode of traditional gray model is only first-order differential equation,and the fitting mode is relatively single,and the prediction accuracy is not high,which cannot meet the load demand.Therefore,this paper carries out a series of optimization and improvement on this basis.Then this paper combines the theory of multivariable grey model with the theory of fractional order,and proposes a fractional order multivariable grey prediction model by replacing the original cumulative operation with fractional order cumulative operation.In addition,the fitting ability is improved by adding the initial value selection process,and the differential evolution algorithm is used to optimize the order and find the optimal order,so as to improve the prediction accuracy of the model.Finally,the constructed fractional multivariable gray model is compared with the traditional first-order multivariable gray model,neural network model,SVM model and other models,which further reflects the superiority of the model construction method in this paper.Finally,the wavelet decomposition is combined with the fractional multivariable gray model of this paper,and the load data and related influencing factors are decomposed into multiple sequences through wavelet decomposition,and the fractional multivariable gray prediction is carried out for each group and each layer of the sequence.Then,through the reconfigurable advantage of wavelet decomposition,the prediction sequence is reconstructed to obtain a set of final prediction sequence,and its relative residual value is calculated.Through comparison,the best wavelet function and the number of layers are found,so as to achieve further optimization of the model.Finally,the constructed wavelet fractional multivariable gray model is compared with the wavelet neural network model,wavelet SVM and other models,so as to highlight the advantages of the model in this paper and complete the construction of the model.