In this work, a new approach for state of health (SOH) estimation via an automated diagnosis method that could be implemented directly into a battery management system (BMS) is introduced. The approach proposed in this work is a multi-step method that improves the accuracy of the estimation and only requires a multidimensional look-up table to be embedded into the BMS. The embedded look-up table enables a robust diagnosis of the battery degradation with a low computing cost.
The first step of the methodology is the emulation  of a cell from its half-cell data. The second step is the selection of features of interest (FOI)  from its voltage response. Next, the variations of these FOIs due to different degradation paths are compiled into a multidimensional look-up table. Lastly, SOH diagnosis is performed by comparing the experimental FOI values to the look-up table. The novelty of this approach resides in the consideration of the variations of all FOIs concurrently by using detection in an n-D space for n FOIs. This simultaneous analysis of multiple parameters provides accurate predictions even when each FOI taken separately cannot provide a conclusive diagnosis.
Our study presents a method which was very light on calculation at the BMS level and was accurate for 90% of all the possible degradation paths for commercial GIC//LFP and LTO//NMC cells. Moreover, the area where the method was not accurate can be identified and correlated with the data resolution and thus reduced the probability of an erroneous diagnosis in the field. This poster will present the methodology and its limitations.
Figure 1: Diagnosis error as a function of the LLI/LAMPE/LAMNE split at 10% capacity loss for the GIC//LFP cell (top) and the LTO//NMC cell (bottom). Diagnosis error on LLI, LAMPE and LAMNEestimations are respectively in the left, middle and right columns.
 M. Dubarry, C. Truchot, B.Y. Liaw, J. Power Sources, 219(2012) 204-216.