State-Of-Health Estimation Pipeline for Li-ion Battery Packs with Heterogeneous Cells

The internal condition of lithium-ion batteries, in particular State-of-Health (SoH), needs careful monitoring to ensure safe and efficient operation. In this paper, we propose a hybrid online SoH estimation pipeline for series-connected heterogeneous cells. Implementing a single cell parameter estimation scheme for a battery pack with hundreds to thousands of cells is computationally intractable. This challenge is solved in this work using feature-based adaptive polling of cells with "extreme" parameter values. Furthermore, the electrical parameters for the polled cells are estimated using online recursive least squares with forgetting factor. The key novelty lies in accounting for the uncertain state dependence of the parameters. We use sparse Gaussian process regression to obtain the parameter bounds as a function of both SOC and temperature. The pipeline is validated through a simulation study, using experimental data from Li-NMC cells.