Optimising Material Recovery and Lifecycle Management of Spent Lithium-Ion Batteries: AI-Based Separator, Repurposing, and Safe Discharge Solutions

This study presents an integrated approach for the collection, classification, characterisation, and repurposing of spent lithium-ion batteries (LIBs), leveraging low-cost embedded systems and safe discharge techniques. Spent LIBs were obtained from mobile phone repair shops near Kwame Nkrumah University of Science and Technology and classified using a machine learning-based separator deployed on an ESP32-CAM and an Arduino nano microcontroller. The system incorporated a conveyor mechanism driven by a NEMA 17 stepper motor and SG90 servo motors, with real-time image classification trained via the Edge Impulse platform. Battery characterization was conducted based on terminal voltage, using 2.5 V as the threshold for sorting. Cells with voltages below this threshold were recharged and evaluated for charge retention, while those above were subjected to controlled discharge tests using an electronic load tester. Recovered cells were then assembled into series and parallel configurations using a Battery Management System (BMS) and buck converter. The repurposed systems demonstrated effective performance in low-power applications such as LED desk lamps, Bluetooth headphones, and emergency power banks. However, the system was unsuitable for high-demand devices like smartphones due to current mismatches, which triggered device protection mechanisms. Additionally, chemical discharge using 50% and 60% brine solutions (NaCl) revealed that higher salt concentrations enhanced discharge rates but also introduced safety concerns such as hydrogen evolution and corrosion. The findings underscore the potential of repurposing end-of-life LIBs for secondary applications, the limitations of edge-device AI deployments, and the importance of safe, environmentally conscious discharge practices. This work contributes to the advancement of circular economy models and sustainable energy management.