The state of health (SOH) of lithium-ion batteries is a decisive factor in ensuring the stability of electric vehicle systems. To solve the problem of low accuracy and robustness of lithium-ion battery SOH prediction models, this article proposes a differential evolution grey wolf optimization algorithm mixed kernel least squares support vector regression (MK-LSSVR) prediction model. Four health features were extracted from individual batteries from NASA and Cycle datasets. These features can describe the degradation properties of lithium-ion batteries. The Pearson correlation coefficient is used to detect the correlation between battery SOH and health features. Principal component analysis performs dimensionality reduction and fusion processing on the health feature dataset to reduce data redundancy. The genetic, selection, and mutation rules of the differential evolution algorithm are improved to enhance the grey wolf (DEGWO) search algorithm. The DEGWO algorithm optimizes the core parameters of the MK-LSSVR model to enhance its predictive ability. The research results indicate that the average absolute error of the prediction model is between 0.36 and 0.62%. The prediction model proposed in this article effectively improves the prediction accuracy and robustness of the battery health state.