Temperature is an important factor affecting the life and safety performance of power batteries. Under complex driving conditions, temperature prediction control of vehicle power batteries is very important. Aiming at the problem of the accuracy measurement of the temperature measurement of the pure electric vehicle power battery, the temperature data of the vehicle under the typical driving conditions of FTP75 and UDDS are collected respectively. Based on the thermodynamic model of the power battery, the parameter model based generalization is adopted. The predictive control algorithm establishes the vehicle model of pure electric vehicle and the accurate prediction model of power battery temperature, and builds a test bench for power battery temperature measurement. The test results show that the model predicts the temperature in two typical driving conditions of vehicle FTP75 and UDDS. The error with the actual temperature is very small, and the generalized predictive control algorithm model satisfies the target demand for accurate prediction for the temperature of vehicle power battery.