The [formula omitted]-method: State of health and degradation mode estimation for lithium-ion batteries using a mechanistic model with relaxed voltage points

Abstract

Lithium-ion batteries exhibit path-dependent aging behavior. Degradation mode (DM) estimation is a first step towards accurate state of health (SOH) representations by clustering degradation mechanisms. Mechanistic models shift and scale pristine half-cell open circuit potential (OCP) curves of both electrodes to reconstruct the open circuit voltage (OCV) curve by minimizing the difference between measured and reconstructed OCV. Alignment parameters describe the shift and scaling of the OCPs and can be used to estimate SOH and DMs. This study introduces the ΔQ-method, which relies on relaxed voltage points and accumulated charge between these points. It is independent of current rates and applicable after almost every event. The optimization problem minimizes deviation between measured and reconstructed ΔQ. The method is developed with an automotive cell dataset and validated with real-world vehicle data from the BMW i3. The ΔQ-method achieves a mean absolute SOH estimation error of 2.52% and a mean absolute OCV reconstruction error of 7.19mV. Reliable estimations are ensured by predefined filters. The method remains effective with restricted state of charge (SOC) windows or limited data points. It is robust against variations in input data, solver choice, and optimization settings. Convergence is improved by constraining the solution space.