Battery health diagnostics with impedance-voltage cell data and metaheuristic optimisation

Lithium-ion batteries have gained prominence as a leading energy storage choice in consumer electronics, with a steadily increasing presence in the automotive industry. However, persistent safety and reliability concerns stemming from battery degradation remain significant obstacles to the broader acceptance of Lithium-ion battery technology, especially in the automotive sector. This research is focused on answering key research gaps identified at the cell level by providing a framework for accurate cell-level model parameterisation, both at pristine and aged stages of the cells, with subscale battery degradation insights to push the current technological advancements.

Among the several analytical techniques, physico-chemical battery models are capable of better simulation of Lithium-ion cell behaviour due to their more physically descriptive modelling approach using many physical parameters. However, the accuracy of simulations highly depends on the correct identification of model parameters. Preanalysis of parameter sensitivities to model outputs can greatly help in determining the best conditions and techniques for identifying many parameters with better accuracy. The first research objective of this thesis demonstrates a next step in improving the parameter identification accuracy by combined parameter sensitivity analysis of both discharge voltage and electrochemical impedance spectroscopy data. A commercially available NMC-graphite based Lithium-ion case study cell is modelled and 28 physical parameters are analysed for their sensitivity. The ranked analysis of the parameter sensitivities clearly demonstrates the ability to identify many discharge voltage-insensitive parameters using impedance measurements.

On the other hand, accurate and feasible model parameterisation with state-of-the-art experimental techniques is laborious, expensive and tied to inherent measurement and estimation errors. The second research objective presents a multi-step, multi-SOC voltage- and impedance-based data-driven hybrid parameter identification framework to explore the capabilities of such technique for accurate model parameter identification.

A methodical parameter identification framework based on parameter sensitivities is presented, coupled with metaheuristic optimisation. The proposed parameter identification framework is benchmarked under multiple cell operating conditions and achieves more accurate model simulations under constant current discharge and transient duty cycles compared to an experimentally derived parameter set from the literature. Battery degradation quantification is a heavily researched frontier by both academic and industrial sectors. Generic battery health metrics without internal degradation insights are not sufficient for correct interpretations of battery ageing due to path dependencies of battery ageing. Under the third research objective of the thesis, the proposed multi-step parameter identification technique is analysed with cycling-aged and calendar-aged case study cell data for assessing battery health diagnostics. The proposed technique is also benchmarked against a more widely used voltage profile model by analysing the degradation modes, which are widely accepted as representative of internal degradation mechanisms. The battery health analysis utilising the evolvement of tracking parameters of the proposed technique shows competitive results to both experimental data and voltage profile model estimations. More importantly, the proposed technique provides electrode level degradation insights, which can be treated as the next level of battery health diagnostics, yet not possible with the voltage-based technique.