To improve the accuracy of thermal model, optimize the design of heat dissipation system and evaluate online the thermal management system, an online estimation method of key thermal parameters is developed from carefully designed experiments, rather than being taken from the literature or the empirical value. The accurate prediction of heat generation, which is based on a reduced wide-temperature-range electro-thermal coupled model, is presented under easily obtainable alternating current excitation at different temperatures. To circumvent some inherent errors, a combined experimental/computational approach to simultaneously estimate the specific heat capacity and thermal resistance is proposed using quasi step power, with which the identification time is significantly reduced. The identified values of specific heat capacity and thermal resistance are validated with high accuracy. The adaptability validation is carried out under different temperatures and cooling conditions, as well as using different battery chemistries, indicating that the proposed method is generic. The in-situ methodology, thanks to good robustness on the colored noise, is capable of providing a promising candidate for accurate thermal modeling, on-board evaluation of battery thermal safety, and advanced design of thermal management system for electric vehicles.