An Entropy-based Approach for Modeling Lithium-Ion Battery Capacity Fade

Batteries are key components of many electronic devices including instrumentations, vehicles, embedded systems, and medical devices. The malfunction of batteries may cause failure in operations of the entire system. Thus, the health management of the batteries, such as the determination of the operating conditions and replacement intervals, is essential in order to ensure the normal functioning and operations of the entire system. The battery health indicators built on effective health monitoring can be integrated into a prognostic model such that the batteries will be operating within the design limits to meet expected performance and safety requirements. Entropy, which originated as a concept in physics and thermodynamics, has been widely used to measure the regularities and uncover the uncertainties of stochastic processes. Different entropy measures have been introduced since Shannon presented the first definition of entropy, including Permutation entropy, Renyi entropy, Tsallis entropy, Approximate entropy, and Sample entropy. In this study, we assess various entropy measures of short voltage sequences of multiple lithium-ion batteries under different testing conditions. Then a Support Vector Machine (SVM) is employed to model the relationship between the battery capacities and various entropy measures of battery voltages. Numerical results reveal that the entropy measures are effective estimators of battery capacity loss and the proposed SVM-Entropy-based model is capable of predicting the battery capacity fade with high accuracies.