Abstract: In lithium-ion battery power management systems, dynamic modeling accurately is the key techniques that maintain the battery energy storage system operating safely and reliably. For the issue about test data with uncertain noise signals, data saturation and biased identification parameters, the bias compensation recursive least-squares（BCRLS） method was used to identify model parameters online based on the second-order RC equivalent model. The proposed method were validated in four different tests and compared with the conventional RLS method. According to the three commonly used open circuit voltage（OCV） test methods, six typical OCV test profiles were designed and then the piecewise cubic Hermite interpolation was proposed for improvement of the identified OCV curves. The experimental results show that BCRLS algorithm has better convergence than RLS and the obtained model has higher precision and better performance, hence the feasibility and effectiveness of the proposed method was verified. Through comparative analysis of the modeling performance of the identified OCV curves, the OCV test method, constant current charging and discharging with short-time rest test profile, is considered as the preferred choice due to its less duration time and equivalent modeling performance.