Electric Vehicle User Behavior Prediction Using Learning-Based ApproachesSara Khan; Boris Brandherm; Anilkumar Swamy
In: 2020 IEEE Electric Power and Energy Conference (EPEC). IEEE Electric Power and Energy Conference (EPEC-2020), November 9-10, Edmonton, AB, Canada, ISBN 978-1-7281-6490-8, IEEE, 2020.
One of the main barrier for electric vehicles to be successful in real world is the need for expensive charging infrastructures. The key aspect of EV is time required to charge the battery to full capacity is far less than the time duration for which the car remains available for charging. Smart charging system can leverage this aspect to efficiently manage the load demand, which in turn alleviates the need for more than necessary number of expensive charging infrastructures. EV user behaviour prediction is vital for building EV Adaptive Charging System. In the past there have been several statistical and ML methods that tries to predict EV user behavior. But with the influx of huge amount of EV user data and deep learning's (DL) ability to perform well on such large data enables us to build DL based methods to predict EV user behavior. In this paper, we predict EV user behavior using ML and DL methods and compare the results and infer the insights for difference in performance. By comparing at various settings between machine learning (ML) and DL methods, we found that K-Nearest Neighbours outperforms Neural Networks with a very minute difference of 0.031 in Mean Absolute Error metric.