DaSoR: Data-driven State-of-Health & Remaining Useful Life Estimation of Li-ion Batteries in a Closed-loop system of Electric Vehicles

Understanding the State-of-Health (SoH) and the Remaining Useful Life (RUL) of the commonly used Lithiumion Batteries (LIBs) within the emerging sector of Electric Vehicles (EVs) is crucial for assuring their efficiency, stability, and safety. However, predicting SoH precisely is challenging due to the intricacies of electrochemical processes within the cells and especially the variability of real-world operating circumstances of an EV. As a result, numerous data-driven approaches for estimating SoH with resilient and adaptable characteristics have been advocated in the literature. However, there is a lack of real-time closed-loop systems, extensive algorithm analysis across multiple datasets, and efficient hyperparameter optimization in the existing works. Hence in this paper, we proposed a robust closed-loop system to report user/ OEM of the battery health, quantitatively compared the performance of several ML algorithms on both CALCE and NASA Ames PCoE Battery Datasets and used AutoML for optimal hyperparameter selection. Furthermore, time series modelling of the intrinsic parameters of the LIBs with Long Short Term Memory (LSTM) for estimating the SoH yields promising results in general, with a mean absolute error (MAE) of 0.031 on the NASA Ames PCoE Battery Dataset and a best MAE of 0.026 on the CALCE Dataset for MLP.