Multi-time scale forecast for schedulable capacity of Electric Vehicle fleets using big data analysis

Abstract

Electric Vehicle fleets can be used as distributed energy resources and providing ancillary services for the Grid through Vehicle-to-grid (V2G) technology. To enable this benefit, a key problem that should be addressed first is how to predict the schedulable capacity (SC) of electric vehicle (EV) fleets to meet different demands of power dispatch for utility. In this paper, two different time scales schedulable capacity forecasting (SCF) methods for EV fleets are proposed, including the real-time SCF method and one day ahead SCF method based on parallel random forests algorithm. The real-time charging or discharging data of each EV is collected to ensure the reliability of results by the real-time SCF method. The large amount of results of the real-time SCF is used as historical data for training one day ahead SCF model. A distributed parallel method of data processing for these big data is introduced to solve problems of the mass data processing and storage in the real-time SCF of EV fleets. The proposed methods are tested and analyzed by digital simulation calculation and compared with support machine (SVM) method by combining cloud computing technology, Hadoop and Spark platform. The simulation results show that proposed methods have the obvious advantages in speed and accuracy over SVM, and can meet the requirements of multi-time scale SCF.