Robust Fuzzy Entropy-Based SOH Estimation for Different Lithium-Ion Battery Chemistries

Machine learning technologies have gained considerable attention for state of health (SOH) estimation of Lithium-ion batteries due to their advantages in learning the behavior of non-linear systems. The mapping between the features and the SOH can be established according to learning and optimization theory. However, the SOH features can become invalid under different conditions as the battery aging process is closely related to the operating conditions. In this work, the fuzzy entropy (FE) of the voltage, extracted from short-term current pulses, is proposed as a feature for support vector machine-based (SVM-based) SOH estimation. The robustness and effectiveness of the proposed methods are verified by extended experiments performed on the three most common Li-ion battery chemistries, i.e., NMC, LFP, and NCA. The obtained Pearson correlation coefficient, relating the FE feature to the SOH, returns values higher than 0.9. Finally, the proposed FE-based SVM model can estimate the SOH of the considered batteries with MAPE below 1.6% when the battery state of charge (SOC) is known and MAPE below 3.4% when the SOC is not known.