Huzaifa Rauf1, Muhammad Khalid2, Naveed Arshad3, and Michael G. Pecht1
1 Center for Advanced Life Cycle Engineering (CALCE) University of Maryland College Park, United States
2 Electrical Engineering Department, K.A. CARE Energy Research and Innovation Center King Fahd University of Petroleum & Minerals (KFUPM) Dhahran, Saudi Arabia
3 Department of Computer Science Syed Babar Ali School of Science and Engineering (SBASSE) Lahore University of Management Sciences (LUMS) Lahore, Pakistan
For more information about this article and related research, please contact Prof. Michael G. Pecht.
Battery cyclic loss is a key parameter to assess lithium-ion battery degradation in electric vehicles (EVs), while machine learning (ML) methods can be used in evaluating and predicting the degradation trend of battery health due to cyclic loss. The accuracy of ML methods is influenced by the input parameter selection of the model. This paper develops a feature selection strategy based on the utilization of a data pre-processing method, which extracts useful model input parameters from the battery data. To show the advantages of the method, eight widely used ML algorithms are applied to a case study and compared for battery cyclic loss prediction. The results show that the developed feature selection method has improved the prediction accuracy by at least 9%, in the case of LASSO regression The results also depict that the random forest (RF) regression, Gaussian Process Regression (GPR), and XGBoost methods, when applied in combination with the developed feature selection method, show an improvement of 44%, 48% and 52% in the prediction accuracy, respectively.
This article is available online here and to CALCE Consortium Members for personal review.