Improved Chaotic Sparrow Search-Optimized Deep Multiple Kernel ELM for Lithium-Ion Battery RUL Prediction

Abstract

The widespread adoption of lithium-ion batteries as energy storage components necessitates accurate prediction of their remaining useful life, which is critical to ensuring operational safety and enabling timely maintenance. To address this challenge, this study proposes a data-driven deep multiple kernel extreme learning machine (DMK-ELM) model, optimized by an improved chaotic sparrow search algorithm (ICSSA), for remaining useful life prediction of lithium-ion batteries. The model constructs a mapping from historical health features to future capacity degradation, strictly adhering to a causal prediction protocol that avoids information leakage. The proposed ICSSA enhances the quality of the initial population and bolsters the global search capability through an improved Tent chaotic sequence initialization and a Gaussian variation strategy. This optimization framework is employed to fine-tune the hyperparameters of the DMK-ELM model. To comprehensively evaluate battery degradation, experimental data from ten cells across two distinct datasets are analyzed. Health features, including constant current charging time (CCCT), constant voltage charging time (CVCT), internal resistance, and incremental capacity peak value (ICCP), are extracted and refined using a causally implemented Hampel filter combined with Pearson correlation analysis performed exclusively on training data. The results confirm the accuracy and robustness of the proposed approach, offering a practical strategy for precise remaining useful life estimation in lithium-ion batteries through the extraction of highly correlated health features.