Lithium-ion batteries have paved the road for electrified mobility that could help reduce the environmental impact of transportation. The expected lifespan of a battery pack (around 300.000 km) is still one of the main limitations for an all-off-grid electric transportation. Being able to understand the impact of the operational parameters on the lifespan of the battery, and as consequence its aging, could give the possibility to slow it down. Here, the main challenge consists in obtaining information from the cell having access only to voltage and current. Extended databases containing voltage, current, and temperature time series, as well as Electrochemical Impedance Spectroscopy (EIS) measurements at different voltages from many cells are open access and the researchers have thoroughly explored these databases with artificial intelligence approaches to determine state of charge and/or state of health. EIS information constitutes the most interesting and complete picture of the single processes, but it is measured traditionally under equilibrium conditions, during which the battery cannot be used for hours. Here we present Dynamic EIS (DEIS), which allows extracting the time-varying frequency analysis of the system while it is drifting through non-equilibrium states. The focus of our research is to correlate the time-varying impedance with the SoH of the battery to be used in battery management systems (BMS). We are collecting continuous time-varying impedance on lab scale cells cycling with a constant current constant voltage (CCCV) protocol until reaching its end of life. We implemented a data fitting routine through regression of Padé approximants combined with statistical tools. From the Padé approximants is then possible to reconstruct the degree of freedom of the system and their variation in time. This allows building a physical model of the battery. A schematic representation of the full procedure is reported in Figure 1.
The time-varying parameters are then employed in a feature extraction process to find its correlation with the state of health using common techniques like Principal Component Analysis (PCA), Linear Discriminant Analysis (LDA) and Non-negative Matrix Factorisation (NMF).
This approach will allow an on-line monitoring tool for the continuous estimation of SoH of the single cell in the pack through the BMS. The BMS could then be programmed to run the pack in the best state and prolong the lifespan of the battery using different power loads for each cell of the pack.