Forecasting the EV charging load based on customer profile or station measurement?. (1st February 2016)

• Record Type: Journal Article
• Authors: Majidpour, Mostafa; Qiu, Charlie; Chu, Peter; Pota, Hemanshu R.; Gadh, Rajit
• Is Part Of: Applied energy. Volume 163(2016)
• Page Start: 134
• Page End: 141
• Publication Date: 2016-02-01
• Subjects: Electric Vehicle -- Privacy -- Time series -- Load forecasting; Power (Mechanics) -- Periodicals; Energy conservation -- Periodicals; Energy conversion -- Periodicals; 621.042
• DOI: 10.1016/j.apenergy.2015.10.184
• Languages: English
• ISSNs: 0306-2619

Abstract

Highlights: We compare the forecasting of the EV charging load based on two different datasets. Customer profile dependent dataset is prone to privacy invasion. Station measurement dataset is directly measured from charging outlets. Results show customer profile based prediction is faster due to less preprocessing. We found that both datasets yield comparable forecasting error. Abstract: In this paper, forecasting of the Electric Vehicle (EV) charging load has been based on two different datasets: data from the customer profile (referred to as charging record) and data from outlet measurements (referred to as station record). Four different prediction algorithms namely Time Weighted Dot Product based Nearest Neighbor (TWDP-NN), Modified Pattern Sequence Forecasting (MPSF), Support Vector Regression (SVR), and Random Forest (RF) are applied to both datasets. The corresponding speed, accuracy, and privacy concerns are compared between the use of the charging records and station records. Real world data compiled at the outlet level from the UCLA campus parking lots are used. The results show that charging records provide relatively faster prediction while putting customer privacy in jeopardy. Station records provide relatively slower prediction while respecting the customer privacy. In general, we found that both datasets generate comparable prediction error.