High Precision SOH Estimation for Lithium-ion Batteries: A Synergistic Approach Using Autoencoder, Machine Learning, and BiGRU

Since lithium-ion batteries are used in a variety of industries, it is vital to correctly predict the state of health (SOH) in order to ensure safe application. At the same time, accurate SOH estimation is crucial for battery management systems (BMS). This article proposes a Crested Porcupine Optimizer (CPO) algorithm optimization deep Kernel Extreme learning machine (DKELM) model for SOH diagnosis of lithium-ion batteries. This work selects three individual batteries as the research objects. Extracting three health features from a single battery characterizes the performance degradation process of the battery. Health features are processed using the bidirectional gated recurrent unit (BiGRU) method. Regularization and kernel parameters of KELM are optimized by the CPO algorithm. The use of the Autoencoder method for deep training of the KELM model further improves the predictive ability of the model. The experimental findings demonstrate the validity and efficacy of the model presented in this study.