Rapid High-Precision Diagnosis of the Capacity and Internal Resistance of Lithium-Ion Batteries Using Impedance Measurements

Abstract

Lithium-ion batteries (LIBs), which are increasingly employed for energy storage, should be reused whenever possible to mitigate resource depletion and environmental problems. This requires a nondestructive diagnostic method to assess their potential for reuse in terms of capacity reduction and power drop, which occurs with increased internal resistance. The diagnostic method should be fast and simplified with no need to adjust the temperature or state of charge (SOC) when classifying numerous LIBs. To develop such a method, we compiled 4,220 impedance measurements taken at temperatures ranging from -20 degrees C to 50 degrees C and SOCs from 0% to 100%. 18650-type cylindrical LIBs were used to construct a prediction model of the capacity or internal resistance via machine learning using the impedance, temperature, and open-circuit voltage, instead of the SOC. Through a strict selection of frequencies at which the impedances of used LIBs were measured, it was possible to simultaneously predict LIB capacity and internal resistance with high precision at any temperature or SOC after less than 1 min of impedance measurement. To enhance the generalization performance, three types of degraded LIBs were employed in the prediction model. Finally, this study demonstrated improved data prediction in the extrapolation area.