Predicting Efficiency of On-Board and Off-Board EV Charging Systems Using Machine Learning

Prediction of charging energy and power profiles is crucial for optimal scheduling of different EV models. This paper proposes the idea of using advanced machine learning (ML) techniques to predict the operating efficiency of on-board (11 kW) and off-board (50 kW) EV charging systems. An experimental setup is developed to control the charging power, measure and collect datasets under several operating conditions using a real EV model. The collected dataset is used to train, validate, and compare ML techniques such as linear regression (LR), random forest (RF), artificial neural network (ANN) and conditional generative adversarial network (cGAN). The evaluation metrics used for this comparison are mean absolute error (MAE), mean square error (MSE) and the coefficient of determination (R<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup>). The results demonstrated that the RF and ANN performed better than the LR and cGAN models in both charging systems.