Classification of duty pulses affecting energy storage systems in vehicular applications

The study of traction batteries real-world usage in vehicular applications faces a handful of serious challenges. In this paper, we propose a new approach to evaluate real-world collected data of battery usage in EVs and HEVs. This automated method relies on K-means clustering technique and aims at classifying duty pulses according to their current and energy distributions. We present the way data must be prepared and we discuss the results of the clustering. We believe this method enables meaningful comparison of vehicles architectures, road conditions, battery management strategies or driving behaviours. Ultimately, it allows for the elaboration of battery ageing profiles which should be more representative of real-world usage.