Electric vehicles performance estimation through a patterns extraction and classification methodology

• Published in: Journal of power sources Vol. 273; pp. 670 - 679
• Main Authors: BARRE, Anthony, SUARD, Frédéric, GERARD, Mathias, RIU, Delphine
• Format: Journal Article
• Language: English
• Published: Amsterdam Elsevier 2015
• Subjects: Aging; Algorithms; Applied sciences; Classification; Computer Science; Data Structures and Algorithms; Electric batteries; Electric power; Electric vehicles; Engineering Sciences; Exact sciences and technology; Extraction; Ground, air and sea transportation, marine construction; Methodology; Numerical Analysis; Pattern analysis; Representations; Road transportation and traffic; Performance evaluation; Automatic classification; Electric vehicles performance; Pattern classification; Estimation; Classification; Electric vehicle; Feature clustering; Pattern analysis; Shape detection; Electrical characteristic
• ISSN: 0378-7753; 1873-2755
• DOI: 10.1016/j.jpowsour.2014.09.098

Direct estimation of battery performance is a major challenge as ageing process is a complex phenomenon not directly measurable. In this work a new methodology is provided to estimate global battery performances under real-life electric vehicle use. Such performances are estimated through battery signals patterns extraction. These signals patterns are used to identify physical degradation behavior of batteries. The analysis framework is composed of patterns extraction, clustering algorithms, summarizing data representation in the feature space of cluster distances and classification algorithms. This methodology is then applied on datasets, acquired from batteries used on electric vehicles, without controlled environmental conditions. The classification algorithm accuracy is studied on the obtained real data. The results suggest that battery signals patterns analysis provides an innovative technique for online estimation of the battery performance level. A detection of dysfunctions caused by ageing is also made, only based on battery signals pattern extracted during real vehicle accelerations.