A hybrid kernel extreme learning machine modeling method based on improved dung beetle algorithm optimization for lithium-ion battery state of health estimation

The state of health (SOH) of lithium-ion batteries is an important indicator for evaluating the degradation of battery performance, which is crucial in battery management systems. With the development of science and technology, data-driven models used to predict SOH are widely used, while data-driven models generally suffer from a narrow estimation range, long prediction time, and poor accuracy. Therefore, this paper introduces the extreme learning machine model with a single hidden layer that incorporates hybrid kernel functions, which overcomes the mapping defect problem of single kernel extreme learning machine and extreme learning machine by the introduction of hybrid kernel functions and improves the range and speed of data-driven models for SOH prediction. In addition, to automatically find the optimal parameters of the kernel extreme learning machine to improve the accuracy, the dung beetle algorithm with strong parameter finding ability is used to optimize the model in this paper. Meanwhile, based on the algorithm, we introduced the optimal Latin hypercube idea and weight factor for initializing the population and regulating the position update, respectively, which effectively improved the convergence and parameter-seeking ability of the algorithm to improve the accuracy of the model. The experiments verified the reasonableness and effectiveness of the proposed model, in which the optimal results of root mean square error, average absolute percentage error, and coefficient of determination of the proposed model on the battery dataset were 0.177%, 0.046%, and 99.76%, respectively, which reflected the high-precision prediction ability and strong robustness.