Abstract: State of health(SOH) is one of the core parameters for the safe management and operation of battery systems. Accurate and rapid estimation of SOH is of great significance to the safe utilization of batteries. To address the issue that conventional SOH estimation algorithms are difficult to involve both high running speed and high accuracy, a fast estimation method of battery SOH using an improved least squares support vector machine(ILS-SVM) is proposed. The robustness and running speed of the algorithm can be effectively improved by setting appropriate critical parameters, which discards some support vectors and weakens the influence of boundary samples on the algorithm. The ILS-SVM is then employed to fast estimate SOH of different battery data sets. By analyzing the operation data of batteries, a specific voltage data interval is used for data preprocessing to avoid the complete charging and discharging measurement of the battery, thus improving the efficiency of battery SOH estimation. The verified results show that accurate estimates can be achieved, and most of estimation errors of battery SOH are less than 1%. Compared with the original LS-SVM algorithm, the running speed of the ILS-SVM can be improved by up to 20%.