Abstract: In order to improve the accuracy of SOC estimation results for the state of charge electric vehicle power batteries, a method for SOC estimation of electric vehicle power batteries based on improved sparrow search algorithm（ISSA） optimized kernel extreme learning machine（KELM） is proposed.Taking voltage, current, and temperature as inputs and battery SOC as output, the ISSA algorithm improved by using logistic chaos initialization, nonlinear decreasing inertia weight, and Levi flight strategies is used to optimize KELM.Secondly, an electric vehicle power battery SOC estimation model based on ISSA-KELM is established.Finally, simulation analysis is conducted using lithium iron phosphate battery charge and discharge data, and comparison is made with other battery SOC estimation methods.The results show that the maximum relative error, average relative error, and root mean square error of the ISSA-KELM model for predicting the test set are 6.232%,3.964%,and 0.0197 respectively.The diagnostic accuracy and model stability are better than other methods, verifying the effectiveness of the proposed battery SOC estimation method.