Rapid estimation of residual capacity for retired LiFePO4 batteries using voltage interval at low state of charge

Abstract

Although the second-use application of retired electric-vehicle batteries has considerable economic benefits, the bottlenecks of large-scale applications are the time-consuming detection and inaccurate estimation of the retired-battery performance. In this study, a hybrid model was proposed based on the mechanism and data-driven methods using only the voltage interval at a low state of charge (SOC) to estimate the residual capacity rapidly and accurately. First, health indicators characterising the capacity-loss mechanism were established by discussing the ageing law of the prognostic and mechanistic model parameters. Then, the conditions for identifying the proposed health indicators were determined by analysing the observability of the health indicators and phase-transition characteristics of the negative electrode during constant-current charging. Furthermore, 920 retired lithium ferro-phosphate (LFP) battery datasets were used to establish the mechanism and data-driven model using the health indicators as model inputs, and the voltage interval (0–20% SOC) was developed to detect the residual capacity rapidly by analysing the identified health indicators using partial voltages (0–20 and 0–60%). Finally, the residual capacities estimated using different training samples, voltage intervals, charging rates, and LFP batteries were analysed and discussed. The root-mean-square error of the proposed method was within 2.67% using only 10% of the data and the 0–20% SOC voltage interval. The results indicated that the proposed method shortened the detection time of large-scale retired batteries from 30 to 50 min, as required for full voltage, to only 6–10 min.