Detection and Isolation of Small Faults in Lithium-Ion Batteries via the Asymptotic Local Approach

This contribution presents a diagnosis scheme for batteries to detect and isolate internal faults in the form of small parameter changes. This scheme is based on an electrochemical reduced-order model of the battery, which allows the inclusion of physically meaningful faults that might affect the battery performance. The sensitivity properties of the model are analyzed. The model is then used to compute residuals based on an unscented Kalman filter. Primary residuals and a limiting covariance matrix are obtained thanks to the local approach, allowing for fault detection and isolation by
$\chi^{2}$
statistical tests. Results show that faults resulting in limited 0.15% capacity and 0.004% power fade can be effectively detected by the local approach. The algorithm is also able to correctly isolate faults related with sensitive parameters, whereas parameters with low sensitivity or linearly correlated are more difficult to precise.