Rapid residual value evaluation and clustering of retired lithium-ion batteries based on incomplete sampling of electrochemical impedance spectroscopy

With the large-scale retirement of power lithium-ion batteries in electric vehicles, the appropriate disposal of retired batteries (RBs) has become an important concern. Evaluating the residual value and exploring secondary applications for RBs are considered promising technical approaches. However, existing residual value assessment techniques face challenges in balancing assessment accuracy and efficiency. To address this issue, a rapid residual value evaluation and clustering method for RBs based on incomplete sampling of electrochemical impedance spectroscopy (EIS) is presented. First, a neural network-based EIS reconstruction method that utilizes a limited number of EIS sampling points to reconstruct the full-frequency EIS is developed, significantly reducing the testing time. Next, the second-order fractional-order model (FOM) parameters are identified by an improved particle swarm filtering algorithm to investigate battery aging characteristics. Subsequently, the Gaussian process regression (GPR) algorithm is applied to estimate the state of health (SOH) of the battery based on the reconstructed EIS and FOM model parameters. Finally, soft clustering of RBs is conducted via a Gaussian mixture model (GMM) based on SOH and FOM model parameters, and tests are conducted to verify the effectiveness of the proposed method. The results reveal that the maximum root-mean-square error and the maximum absolute value error of the EIS reconstruction are lower than 0.25 m Omega and 0.7 m Omega, respectively, while the maximum relative error of the SOH estimation is lower than 2 %. Moreover, the residual value evaluation time for each RB is 3 min, which is at least 10 times shorter than that of the standard capacity test. This study has tremendous practical value for quickly evaluating and clustering residual values for large-scale RBs.