Machine Learning Framework for Optimising Regeneration of Lithium‐Ion Battery Cathode Materials

Abstract

The rapid growth of the lithium‐ion battery (LIB) market, driven by the rise of electric vehicles and renewable energy integration, has intensified the need to address battery degradation and the environmental impact of spent LIBs. Direct regeneration has emerged as a promising alternative to traditional recycling methods, offering a low‐energy, environmentally friendly, closed‐loop approach for restoring degraded cathode materials. However, identifying optimal regeneration parameters remains experimentally intensive and complex. This study explores the potential of machine learning (ML), particularly artificial neural networks (ANN), to optimise the regeneration process and to evaluate the battery's remaining useful life (RUL) and state of health (SoH) of regenerated cathode materials using solid‐state regeneration methods. A structured ML‐driven framework was developed and applied to two representative chemistries, lithium cobalt oxide (LCO) and lithium iron phosphate (LFP), using experimentally derived datasets. Across both systems, ANN consistently outperformed other algorithms, achieving high predictive accuracy ( R 2 up to 0.99) while identifying refined regeneration conditions that improved specific capacity, enhanced cycling stability, and reduced thermal input. The model further demonstrated strong generalisation capability by accurately predicting performance under unseen regeneration conditions. These results establish a transferable ML‐based optimisation framework for regenerated LIB cathodes and highlight its potential to reduce experimental workloads, improve energy efficiency, and support sustainable battery recovery.