Research On Very Short Term Load Forecasting Method In Smart Grid

With the development of smart grid,more and more new energy resources,such as solar,wind,etc.,have been involved to form a distributed power grid model.However,the power generation of these new energy sources is easily disturbed by the natural conditions,such as illumination and wind speed.Especially,with the increase of the new energy access,the stability of the smart grid is greatly affected.The veryshort term load forecasting has significant reference value in grid stability and real-time dynamic reactive power voltage optimal control.Ultra-short term load forecasting has the characteristics of short prediction time and high real-time requirement.At present,the ultra-short term load forecasting is still in research.How to make use of the large amount of time-series data in the smart grid and to excavate the potential information in the short-term load forecasting has become a hot research direction.Aiming at the stability of very short term load forecasting algorithm existing and ignoring user behavior similarity problem,we proposed the very short term prediction method based on virtual user model and prediction interval;and then combined with the power user data stream characteristics,we introduced data stream clustering into very short term load forecasting method to improved prediction speed.The main research of this paper is as follows.First of all,this paper summarizes the existing algorithms of very short term load forecasting.Secondly,to solve the problem of existing prediction algorithms without considering the similarity of user consumption behavior,this paper proposes virtual user model by analyzing the characteristic of load curve;thirdly,taking the user consumption behavior of the stochastic characteristics into account,this paper introduce prediction interval to improve the stability of prediction algorithm.Combined with virtual user model,a new method is proposed base on virtual user model and prediction interval;then,according to the timing characteristics of user load data in smart grid,the data flow model of virtual users clustering technology of very short term load prediction method was improved to increase the prediction speed of the algorithm;finally,by the experiment,this paper puts forward the very short term load forecasting based on virtual user model of accuracy of algorithm,which is better than other algorithms in the experiments,and the introduction of data stream clustering is also in the range of acceptable precision of prediction,the prediction speed has been significantly improved.