State-of-health estimators coupled to a random forest approach for lithium-ion battery aging factor ranking

Abstract Electrified vehicles users may expect their vehicle to have a steady autonomy range and available power throughout the lifetime of their cars. The health assessment of Lithium-ion batteries (LIBs), in that regard, represents a critical point for performance evaluation and lifetime prediction. Reliable state-of-health (SoH) assessment is essential to ensure cautious and suitable use of LIBs. To that end, several embedded solutions are proposed in the literature. In this paper, two new aging indicators are developed to enrich the existing diagnosis-based (DB-SoH) solutions. These indicators are based on collected data during charging (CDB-SoH) and driving (DDB-SoH) events overtime. The data are comprised of variables such as distance, speed, temperature, charging power, and more. Both solutions produce reliable state-of-health S o H assessment with a significantly good estimation error. Additionally, a data-driven battery aging prediction using the random forest (RF) algorithm is introduced using actual users’ behavior and ambient conditions. The proposed solution produced an S o H estimation error of 1.27%. Finally, a method for aging factors ranking is proposed. The obtained order is consistent with known aging root causes in the literature and can be used to mitigate fast LIB aging for electrified vehicle applications.