Establishing Second-life Application Suitability Classification Criteria for Retired Batteries using Autoencoder and K-Means Clustering

Increase the lifespan and decrease the environmental impact of lithium-ion batteries, especially in electric vehicles and renewable energy sources, and improve the sustainability of lithium-ion batteries. There is a need to establish criteria for second-life applications and retired batteries to address battery recycling. A framework was presented in the paper to classify second-life using an autoencoder and K-means clustering to classify second-life situations. K-means clusters the latent space using the Autoencoder. An average silhouette score of 0.91 supports classification into second-life application and disposal phases. With an increase in clusters, the average silhouette score decreases. The clusters become more homogeneous as the cluster number increases, thus becoming more difficult to distinguish. The selected battery health parameters are voltage, current, charging, discharging, temperature, and cycle aging. Experimental results proved that a healthy battery performs at 96% of its rated capacity. Second-life applications are suitable for those operating 95-59%, while those below 58% need to be recycled or disposed of. Operating temperatures below 96% of the maximum indicate normal product life, while temperatures above suggest the disposal phase. Batteries that operate less than 96% of the maximum capacity of cycle aging and charging characteristics are considered suitable for second-life applications, while those outside these ranges may need disposal or recycling.