Prediction of Li-Ion battery State-Of-Health based on data-driven approach

The problem of State-Of-Health estimation of Li-Ion batteries is becoming of central importance in many critical fields of application. As of now, the majority of literature focuses on developing adequate tools in order to predict the future degradation of the battery capacity. However, in some cases, such algorithms require measuring several parameters as well as the use of several features and complex neural architectures, which make it difficult an implementation on a Battery Management System. This work builds on this gap by introducing a methodology for the estimation of the battery State-Of-Health (SOH) based on the prediction of residuals of an exponential fitting function. The features of the algorithm are only time-related variables, along with the number of endured cycles, which can be easily measured and stored in real-use applications. The proposed method has been tested and validated using a publicly available battery degradation dataset provided by the NASA prognostic research center. Early results are promising and an incentive for further developments.