Accurate estimation of residual capacity for large-scale retired-LFP batteries with multiple aging pathways

Accurately estimating the residual capacity of large-scale retired batteries is a challenging task due to the existence of multiple aging pathways, which hinders the development of second-life applications. This study proposes an effective method for accurately estimating the residual capacity of large-scale retired batteries exhibiting multiple aging pathways. First, five clustering features that represent different aging pathways were identified through the analysis of capacity loss mechanisms and incremental capacity (IC) curves. A clustering algorithm combining the self-organizing map (SOM) and K-means (SK-means) algorithms was then developed. The integration of SOM addresses the challenges of determining the initial number of clusters and selecting the initial cluster centers in the K-means algorithm. The SK-means algorithm was subsequently applied to cluster retired lithium iron phosphate (LFP) batteries with multiple aging pathways, using a hybrid dataset from Yishui (YS_LFP) and EVE Energy (YWLN_LFP). Finally, an improved Gaussian process regression (IGPR) model was established by combining the Matern covariance function with automatic relevance determination (ARD) and the Rational Quadratic covariance function with ARD. Based on the clustering results obtained from the SK-means algorithm, distinct residual capacity estimation models were developed for each cluster. Furthermore, highly accurate residual capacity estimates were achieved using only the first 10 % of the data for each cluster, with root mean squared error (RMSE) values within 1.81 % for the first cluster and 0.13 % for the second cluster. Comparisons with various scenario assumptions and different estimation methods demonstrate that the proposed method achieves superior accuracy in estimating residual capacity for batteries with multiple aging pathways.