Aging characteristics-based health diagnosis and remaining useful life prognostics for lithium-ion batteries

Abstract This paper developed methods for improving the practicability of battery health diagnosis and remaining useful life prognostics. Battery state of health was estimated using a feature extraction-based method based on the charging voltage curve. Battery remaining useful life was predicted by identifying recognizable aging stages. Acceleration aging test data for 9 cells at constant current rates including 0.5C, 1C, 1.5C, and 2C, and dynamic current rates were used to validate the developed methods. The capacity estimates were accurate with estimation errors less than 1% at most cycles. The remaining useful life was predicted within 0.3 s at dynamic current rates, with the prediction errors at most cycles less than 10 after 300 cycles and the 95% confidence intervals covering about 20 cycles for each prediction.