Estimation of the state of health (SOH) of batteries using discrete curvature feature extraction

Abstract

Lithium battery applications in a variety of engineering sectors must be safe and reliable while maintaining a high level of energy efficiency. An accurate assessment of the battery's state of health (SOH) is critical in battery management systems (BMS). In recent years, it has been proved that machine learning is effective at estimating SOH. This work proposes a novel approach of health indicator (HI) extraction based on the U-chord curvature model, based on a complete analysis of battery aging data. In contrast to previous approaches for feature extraction, our method splits the discharge process into various phases based on the curvature of the discharge curve and extracts many HIs with a high correlation to battery SOH in the discharge platform stage of the discharge curve. To demonstrate the superiority of the proposed model, several well-known machine learning algorithms are employed to estimate SOH using extracted attributes. Long short-term memory (LSTM) and artificial neural networks (ANNs) are examples of these techniques. Accuracy, reliability, and robustness of the proposed model are evaluated using three publicly available data sets. According to the data, the model appears to be capable of accurately calculating the battery's SOH, with a mean absolute error of less than 1.08% and a root mean square error of less than 1.46% for various battery types.