Font Size: a A A

Research On Holiday Electricity Forecasting Algorithm Based On Historical Sample Enhancement

Posted on:2022-02-13Degree:MasterType:Thesis
Country:ChinaCandidate:Z M XiaFull Text:PDF
GTID:2512306722486434Subject:Electrical engineering
Abstract/Summary:PDF Full Text Request
Short term load forecasting plays a key role in the work of the whole power dispatching and power selling companies participating in the spot market.It provides basic data for basic generation plan,system security analysis,day ahead market quotation,etc.The accuracy of load forecasting method affects the economy and security of power grid operation.Due to the influence of the external weather,its own characteristics,the limitation of the number of samples and the rest scheme,it is difficult to achieve good results in the accuracy of the traditional forecasting method for holiday load.Based on the influencing factors of holiday load forecasting,this paper designs a new forecasting method to forecast the total power consumption of holidays every day.The main contents of this paper are as follows.A sample enhancement of holiday electricity forecast based on generative confrontation network is proposed.Based on the advantages of the generative adversarial network model in data enhancement,taking KL divergence as the objective function,the generator and discriminator are trained alternately,and new samples are obtained through training,so as to make up for the lack of electricity forecasting samples in holidays.A holiday load forecasting model based on GAN-GRU is proposed.Analyze the influencing factors of holiday load,and decompose it into meteorological components and reference components through Fourier decomposition;in view of the overlap of holidays caused by the adjustment of legal holidays,a prediction alternative method is designed to eliminate the influence of holiday overlaps on the prediction model.Impact;the enhanced samples are predicted by the two-layer GRU model.Finally,the prediction results are analyzed by calculation examples,and the results prove the superiority of the prediction model.Designed and developed a web application for forecasting electricity during holidays.Based on the Keras deep learning library of python language,designed and developed a set of holiday electricity forecasting software programs,which can be directly called on the command line under windows system;based on java web technology,designed and developed supporting holiday electricity forecasting web programs and electricity forecasting software.The program interacts to realize the visualization of the page.
Keywords/Search Tags:Holiday load forecast, GAN, Deep Learning, Sample enhancement, Fourier decomposition
PDF Full Text Request
Related items