A reliable approach of differentiating discrete sampled-data for battery diagnosis

Abstract Over the past decade, major progress in diagnosis of battery degradation has had a substantial effect on the development of electric vehicles. However, despite recent advances, most studies suffer from fatal flaws in how the data are processed caused by discrete sampling levels and associated noise, requiring smoothing algorithms that are not reliable or reproducible. We report the realization of an accurate and reproducible approach, as “Level Evaluation ANalysis” or LEAN method, to diagnose the battery degradation based on counting the number of points at each sampling level, of which the accuracy and reproducibility is proven by mathematical arguments. Its reliability is verified to be consistent with previously published data from four laboratories around the world. The simple code, exact fitting, consistent outcome, computational availability and reliability make the LEAN method promising for vehicular application in both the big data analysis on the cloud and the online battery monitoring, supporting the intelligent management of power sources for autonomous vehicles.