A statistical assessment of the state-of-health of LiFePO4 cells harvested from a hybrid-electric vehicle battery pack

Second-life battery applications can reduce the environmental impacts and life-cycle costs of electrochemical energy storage. The feasibility of reusing battery packs is strongly influenced by their modules' and cells' stateof-health. This paper analyzes the statistical distribution of state-of-health of cells in a retired BAE-Systems ESSA123 hybrid-bus battery pack, consisting of 16 modules with LiFePO4 26,650 cells in a 12s8p configuration. We systematically sampled cells from the pack and collected several categories of state-of-health data. We provide three quantitative descriptions of the statistical distributions in SOH that can be used to model variation and uncertainties in 2nd-life feedstocks, including a linear regression between direct-current internal resistance (DCIR) and percent initial capacity (PIC) and fitted distributions for each variable. We found that SOH variation and mismatch within the modules is high in used HEV packs, making it infeasible to reuse most of them as-is. We illustrate with two feasible acceptance criteria adopted that the number of cells and modules that are accepted for second-life applications (90-99 % and 75-94 %, respectively) can vary drastically based on the criteria, illustrating the sensitivity of a 2nd-life business model to the definition of 'faulty' cells and modules. Finally, we found a key frequency (89 Hz) at which the correlations between AC impedance and constant-current cycling parameters are particularly strong, suggesting that cheap, fast EIS tests can potentially substitute longer diagnostic tests, thus lowering the cost of refurbishment.

Automated summary

This paper analyzes the statistical distribution of state-of-health of cells in a retired battery pack and provides quantitative descriptions for modeling uncertainties in 2nd-life feedstocks. It found that the variation and mismatch within the modules of used HEV packs are high, making it infeasible to reuse most of them as-is. The paper also suggests that cheap, fast EIS tests can potentially substitute longer diagnostic tests, thus lowering the cost of refurbishment.