Abstract:  Rapid and accurate estimation of battery state of health (SOH) is the basis for ensuring the safety of power battery systems. During the operation of pure electric vehicles, traditional estimation methods are difficult to use the limited computing resources of the vehicle to construct accurate SOH estimation models online. To solve this problem, an online feature extraction method using short-term monitoring data to construct battery health indicators (HI) is proposed. In this method, the accumulated electricity in different voltage ranges is regarded as frequency items, and the Lossy Counting algorithm is used to construct a summary data structure to perform statistic on the distribution of frequency items, and the battery health status is characterized based on the change of the distribution law of the frequency items. Simulation and experimental results show that the proposed method can use on-board computing resources to extract battery health status indicators with small time and space complexity.