A model-based method of parameters estimation and fault diagnosis for lithium-ion battery pack was proposed considering the internal short circuit fault which was easy to occur in lithium-ion battery pack under variable low temperature environment. The relationship between capacity and temperature of lithium-ion battery was established through the characteristic test. The unscented Kalman filter was utilized to estimate state and parameters of standard cell according to the standard-deviation model. The real-time estimation results of battery pack were obtained by double Kalman filter, and the internal short-circuit fault was quantitatively diagnosed depending on the depletion of capacity. The actual driving condition test was conducted on a series-connected battery pack under the time-varying temperature of 5~25 °C, and internal short circuit fault was simulated by the parallel resistance method. The battery state estimation and internal short circuit fault diagnosis results were compared with real value. The experimental results show that the proposed method can precisely track the state of battery pack and accurately diagnose the internal short-circuit fault.