Abstract: State of charge（SOC） estimation technique is one of the most important functions of battery management system. The main purpose of this paper is to accurately estimate SOC of each cell in the series connected Li FePO₄ battery pack. Firstly a comprehensive battery model was established based on Thevenin model and Ah counting model; and then a square root sigma point Kalman filter（SRSPKF） was adopted for SOC estimation, besides, a recursive least square（RLS） algorithm was also used to identify model parameters, in this way, model parameter identification and SOC estimation can be realized simultaneously by combined SPKF-RLS method. Theoretically speaking, by using SRSPKF, the system states were propagated in the form of square root of its variance, and thus the computation complexity of conventional SPKF can be significantly reduced. Experimental results show that, compared with SPKF, SRSPKF possesses a stronger error suppression capability of state estimation, and more accurate SOC estimation results can be obtained by SRSPKF.