High precision estimation of remaining useful life of lithium-ion batteries based on strongly correlated aging feature factors and AdaBoost framework

Abstract

In response to the current issue of low accuracy and robustness in the remaining useful life (RUL) model of lithium-ion batteries. In the framework of AdaBoost, a lithium-ion battery life prediction model based on an improved whale optimization algorithm to optimize the Kernel Extreme Learning Machine (IWOA-KELM) is proposed. The IWOA-KELM model is used as a weak predictor. A weighted voting mechanism is used to set a weight coefficient for each weak predictor and then combine the strong predictor of battery RUL. Constant current charge time, constant voltage charge time, internal resistance, and incremental capacity curves peak were extracted from the Cycle data set as health features to accurately describe battery degradation. Pearson correlation coefficient and Savitzky-Golay filter preprocessed health features. Tent chaotic mapping is used to initialize whale populations and maintain their diversity. The iterative updating strategy of the hunting speed control factor is introduced to reduce the probability of the local optimal case of the whale optimization algorithm. The kernel function parameters and regularization parameters of KELM are optimized by IWOA to improve the model prediction ability. After verification, the RUL error of the method proposed in this article can be as accurate as 4 cycles.