Holistic Comparison of Different Kernel Functions for Support Vector Regression Based on State-of-health Prediction of Lithium-ion Battery

Battery health prediction is an important part of prognostic and health management. It is particularly important to accurately predict the state of health (SOH) in the management of lithium-ion batteries. Owing to its excellent regression ability, support vector regression (SVR) is the preferred data-driven method used to predict the remaining useful life (RUL) of batteries. Kernel functions, such as linear, polynomial, gaussian, and sigmoid kernels are commonly used to map samples from a lower to a higher dimensional space for accurate classification. Parameter estimation considerably affects the mapping ability of kernel functions. The size and distribution of the training set also affect the regression ability of SVR. In this study, lithium-ion battery data are divided into prophase, metaphase, and anaphase training data. Particle swarm optimization (PSO) and a genetic algorithm (GA) are applied to optimize the parameters of the four kernel functions. A comparison of the experimental results showed that GA-SVR with a radial basis function (RBF) and PSO-SVR with an RBF in the anaphase can accurately predict the RUL of lithium-ion batteries.