Augmented model-based framework for battery remaining useful life prediction

Traditional, model-based approaches for predicting the remaining useful life (RUL) of a rechargeable battery cell simply update and extrapolate mathematical model which describes evolution cell’s capacity fade trend. These are straightforward but tend to break down when trend changes over lifetime. To retain desirable properties prediction (uncertainty quantification, long-term accuracy, limited physical meaning) improve their overall accuracy in RUL prediction, we augment empirical with data-driven error correction. Our approach decomposes task into two steps: 1) Offline training models correction 2) Online prediction. The is evaluated on five datasets consisting 237 cells: three open-source datasets, one proprietary dataset, 3) simulated out-of-distribution dataset. Results show that effectively reduces root-mean-square-error by 40% mean uncertainty calibration 34% compared alone. proposed also shown be more conservative its estimates than purely approach. Special attention given ensure initial propagated through considered final enhanced quantification our makes it suitable deployment an online predictive maintenance scheduling framework.