The state of charge (SOC) estimation method for power batteries in engine waste heat recovery system is studied in this article. The nonmodel-based Coulomb counting (CC) method is among the most popular techniques. However, challenges imposed by inaccurate estimation of initial SOC, current measurement error as well as uncertain system dynamics caused by temperature, and thermal effects will produce an accumulated error. To overcome this error and improve the CC method's disturbance rejection capability, a linear second-order extended state observer (ESO) is designed to address this problem. A comparative simulation study is carried out to study the estimation performance of the proposed ESO-assisted CC method and the conventional CC method. Study results of the proposed approach exhibit better estimation performance and adaptability for inaccurate estimation of initial SOC, current measurement error, and uncertain system dynamics. There is only one parameter omega(o) to tune. The adaptability of estimation performance can be maintained without parameters resetting, which proves the effectiveness and disturbance rejection capability of the proposed ESO-assisted CC method for SOC estimation.