Random Forest Regression of Charge Balancing Data: A State of Health Estimation Method for Electric Vehicle Batteries

Recently, Electric Vehicles (EVs) are becoming more widespread. However, their mass adoption is hindered by the limited capacity of their Energy Storage System (ESS). Nowadays mainly Lithium-ion (Li-ion) technology is used for mobile applications, however, their energy density and cost put a hard limit on the maximum size of viable EV battery packs. Therefore, it is crucial to use existing technologies as effective as possible. To efficiently use a battery pack over its entire lifetime, the State of Health (SoH) of the cells needs to be taken into account. In this paper, we propose a novel SoH estimation method, based on the battery pack's behavior during Active Charge Balancing (ACB). From this behavior we are deriving a metric and show that it strongly correlates with the SoH. We use this metric, together with other cell parameters, to train a Random Forest (RF) regression estimator. To gather the training data, we implemented a modular simulation framework, that is able to reproduce the charging and discharging cycles, the charge balancing processes, as well as the aging behavior of battery packs over their entire lifetime. Besides showing a strong correlation between balancing behavior and SoH, we are able to estimate the cells' SoH with an accuracy of 1.94 % for the capacity and 4.28 % for the resistance, respectively. Our capacity SoH estimation outperforms state-of the-Art machine learning approaches, while we are among very few to even provide an estimate for the resistance with a high accuracy.