Enhanced state of charge estimation through Cluster-Based Learning Model: Impact study on degradation and profitability of second-life electric vehicle batteries

The growing adoption of electric vehicles (EVs) presents an opportunity for repurposing end-of-life batteries for second life (SL) applications, such as energy storage systems. However, accurate estimation of the state of charge (SOC) remains critical for optimizing battery performance and extending operational life in these applications. This paper presents an in-depth investigation into the impact of advanced SOC estimation on the degradation and profitability of second-life EV batteries, utilising a Cluster-Based Learning Model (CBLM). An empirical degradation model is adapted to quantify how SOC estimation errors influence key battery health metrics, including capacity loss, State of Health (SOH), and energy retention. The study proposes the “energy advantage metric,” which quantifies the usable energy retained in SL batteries based on SOC estimation accuracy. Capacity loss analysis across various SL applications demonstrates that the CBLM model significantly reduces battery degradation compared to the Standard Long Short-Term Memory (S. LSTM) model, particularly under deep discharge cycles. These improvements in capacity retention are then translated into economic impact, revealing cost savings ranging from £339 in residential PV systems to over €200,000 in grid-scale energy arbitrage. t-Test confirmed significant differences in degradation performance between CBLM and S. LSTM models, with Cohen's d effect size showing a small but meaningful effect size for Loss of Lithium Inventory (LLI) (d = 0.24).