Lingnan University's scholar award-winning project develops AI model to extend retired batteries’ lifespan and reduce environmental pollution

With the global rise in electric vehicles’ popularity, the demand for lithium-ion batteries has surged, leading to a considerable increase in discarded batteries. The issue of how to efficiently use “retired” batteries has become urgent, and to address this Prof Tang Xiaopeng, Assistant Professor of the Science Unit at Lingnan University, in collaboration with researchers from the University of Shanghai for Science and Technology, recently published a groundbreaking research paper titled "Lifespan-based Battery Classification towards Second-life Utilisation", which won the Best Paper Award at the prestigious 2024 IEEE 25th China Conference on System Simulation Technology and its Application (CCSSTA). It was one of only four papers out of 278 submissions to receive recognition.

The research team developed an innovative screening method based on a semi-supervised learning AI model for lifespan-based battery clustering, proposing a way to regroup retired batteries based on their service life. First they trained a convolutional neural network (CNN), a deep learning model considered core to AI technology—to classify batteries in different groups based on their lifespan, using data from the first three charge cycles. This AI model addresses mismatches caused by the inherent differences in retired batteries in conventional detection methods. Instead of predicting lifespan directly, it predicts whether two batteries have similar lifespans, forming an adjacency matrix. This novel approach allows batteries with similar lifespans to be grouped together, effectively enabling the classification of retired batteries based on insights from fresh cells.

The research team then tested 38 retired batteries. Of these, 22 reached the cut-off capacity, while the other 16 batteries showed a better lifetime performance, making them candidates for second-life usage although their lifespan remains unknown. The 22 batteries were further classified into two classes, shorter- and longer-life, as training data. The research team uses the shorter-life battery-aging data to train a data-driven machine to sort the longer-life batteries into different groups, based on their expected lifespan. They carried out a one-to-one comparison of unclassified batteries with the AI model, where batteries predicted to belong in the same category were collected to formulate a group. The model effectively enabled a structured classification of retired batteries.

Results show that even when the initial capacity and resistance of retired batteries are very similar, their lifespans can vary significantly. Using conventional capacity-resistance methods for battery evaluation leads to faulty battery grouping results, and the AI model proposed by the research team can reduce lifespan loss by at least 20 percent over traditional approaches. These findings offer new, reliable and accurate solutions for enhancing and predicting the second life of retired batteries, which helps extend their usable life, contributing positively to energy savings and emissions reduction throughout their lifecycle and, ultimately, improving battery management.

The research team explained that electric vehicle batteries must be retired when their capacity declines to 80 percent of the original, due to safety and range considerations. Repurposing retired batteries for less demanding applications such as energy storage and backup power can alleviate energy crises and reduce environmental pollution, especially because the production of lithium-ion batteries is very energy-intensive. However, retired batteries are inherently aged with lower consistency, and individual batteries with poorer performance can affect the overall performance of the battery pack during secondary use.

Prof Tang said: “The production and manufacturing of lithium-ion batteries is a very energy-consuming process. Therefore the batteries must have a sufficiently long lifespan to ensure their entire lifecycle can make a positive contribution to energy conservation and emission reduction. However, the bottleneck issue of the retired battery is its highly inconsistent service life. Unfortunately battery classification algorithms considering a battery’s lifespan are very limited. These findings break through the technical limitations in examining the effectiveness of retired batteries, which not only improves battery management practices but also contributes to a more sustainable energy future.”

He added that, by ensuring retired batteries are effectively used, this research offers new potential for advances in energy storage and environmental protection. This technology is currently under patent application in Hong Kong, and Lingnan University is actively recommending and exploring its feasibility in collaboration with the HKSAR government and other organisations.