The state-of-charge (SOC) estimation for LiFePO4 batteries is one of the most important issues in battery management system (BMS) on electric vehicles (EVs). Significant temperature changes and drift current noises are inevitable in EVs and cause strong interference in SOC estimation, therefore a SOC-Particle filter (PF) estimator is proposed for SOC estimation. This paper tries to make three contributions: (1) a temperature composed battery model is established based on commercial LiFePO4 cells which can be used for SOC estimation at dynamic temperatures. (2) A capacity retention ratio (CRR) aging model is established based on the real history statistical analysis of the running mileage of the battery on an urban bus. (3) The proposed models are combined with an electrochemical model and the PF method is employed for SOC estimation to eliminate the drift noise effects. Experiments under dynamic current and temperature conditions are designed and performed to verify the accuracy and robustness of the proposed method. The numeral results of the validation experiments have verified that accurate and robust SOC estimation results can be obtained by the proposed method. (C) 2015 Elsevier B.V. All rights reserved.