Research On Multi-time Scale Forecast For Schedulable Capacity Of EVs Based On Big Data

With the development of smart grid,the high penetration of intermittent renewable energy and the increasing use of electric vehicles make the randomness of power fluctuations stronger,thus adversely affecting the stability of a power system.Electric vehicles fleets(EVs)can also be used as distributed energy resources providing ancillary services to the grid through Vehicle-to-grid(V2G)and Grid-to-vehicle(G2V).Therefore,accurate and fast prediction of the schedulable capacity(SC)of EVs play an important role in providing ancillary services.Massive electric vehicle plug-in and randomness of human behaviors cause difficulties for accurate and fast prediction of SC of EVs.In this paper,focusing on the methods of accurate and fast prediction of SC of EVs,the main works and innovation are as follows:1.Considering the randomness and intermittency of SC of EVs,two time scales schedulable capacity forecasting(SCF)problems for EVs are addressed,including real-time scale and one day ahead scale.2.The innovative model of real-time SCF of EVs is proposed based on massive real-time charging or discharging data of individual EVs and EVs owner demand to ensure the reliability of forecasting results of real-time SCF of EVs.Besides,a distributed big data parallel analysis method is used to solve the massive data process and storage problems.3.Forecasting Models of one day ahead SCF of EVs are built based on three paralleled big data algorithms,including parallel random forest algorithm(PRF),parallel decision tree algorithm(PDT)and parallel k nearest neighbor algorithm(PKNN),which is used to analysis the large amount of historical data of the real-time SCF.Besides,the comparative analysis of the three models are done.4.Multi-time scale forecasting models for SC of EVs based on big data analysis is established in big data platform which cluster four PC by Hadoop and Spark.These proposed methods are tested by real-time operation data involving 521 EVs in the field.The test results show real-time SCF method has a high prediction precision,MAPE of real-time SCF can reach 3.07%.For one day SCF,both speed and accuracy of PDT are best,meanwhile,PRF and PDT are of similar accuracy.PKNN is always of the lowest accuracy and speed.