Data smoothing in Fuzzy Entropy-based Battery State of Health Estimation

To ensure the reliable operation of the batteries and maximize their service lifetime, it is important to have accurate knowledge of their state of health (SOH). Using data-driven methods to estimate the SOH is extensively studied and the feature data plays an important role in such methods. As fuzzy entropy (FE) can capture the variation of the voltage during the battery aging process, it can be used as a feature. In this paper, in order to reduce the noise from raw feature data, six smoothing methods are introduced to pre-process the FE. Furthermore, the relationship between the smoothed feature and SOH is established by support vector machine and Gaussian process regression. The comparison results show that adding a simply feature smoothing step before the model training can improve the SOH estimation performance. Finally, the effectiveness of the proposed method is verified by experimental results.