Fault diagnosis for cell voltage inconsistency of a battery pack in electric vehicles based on real-world driving data

• Cell voltage inconsistency of a battery pack is important for the safety of electric vehicle. • Density-Based Spatial Clustering of Applications with Noise (DBSCAN) is able to localize cell fault. • Least-Square Support Vector Regression (LS-SVR) predict the change of the monomer voltage. • Fault over a short time horizon based on voltage difference and monomer voltage are diagnosed. Cell voltage inconsistency of a battery pack is the main problem of the Electric Vehicle (EV) battery system, which will affect the performance of the battery and the safe operation of electric vehicles. In real-world vehicle operation, accurate fault diagnosis and timely prediction are the key factors for EV. In this paper, real-world driving data is collected from twenty all-electric buses for many years and divided into three driving fragments to analyze cell voltage inconsistency and summarize the voltage characteristics of the cell when an inconsistency fault occurred. A fault diagnosis method based on Density-Based Spatial Clustering of Applications with Noise (DBSCAN) algorithm is proposed for timely localization of the abnormal battery cell. It is found that the DBSCAN clustering algorithm has shown better effectiveness and accuracy as compared to K-means to locate irregular battery cells. A fault prediction method based on the Least-Square Support Vector Regression (LS-SVR) is developed to predict the change of the monomer voltage. The experimental comparison show that LS-SVR has better prediction accuracy than ordinary Support Vector Regression (SVR), and it can make short-term predictions based on the voltage difference and monomer voltage value for cell consistency failures and over/under voltage faults.