The internal resistance of lithium-ion batteries changes with temperatures and state of charge(SOC),the differences are especially larger at temperatures below zero.Thus, the battery model with changeless parameters is not accurate.In order to solve the problem that temperature changes have an impact on SOC estimation and enhance the accuracy of the model,temperature correction factors are pulled in.A variable temperature model with two RC networks is established.A maximum likelihood estimation(MLE) algorithm is proposed to obtain parameters in the process of model parameter identification.The MLE is simple,it can solve the data saturation problem caused by the data increase in the least squares algorithm.The convergence property of MLE becomes better as the number of samples increase. Loadcycle experiments are conducted to validate the correctness and accuracy of the proposed parameters identification algorithm.Adaptive Kalman filtering (AEKF) method is adopted to calculate the battery SOC based on the variable temperature model in a temperature-changeable environment.The result is compared with that based on single value model using AEKF and unscented Kalman filter(UKF).It can be seen from the comparison that the accuracy of SOC estimation obtained by AEKF is high,the error never exceeds 2.2%.