In this study, a prediction model based on artificial neural network is constructed for surface temperature simulation of nickel-metal hydride battery. The model is developed from a back-propagation network which is trained by Levenberg-Marquardt algorithm. Under each ambient temperature of 10 degrees C, 20 degrees C, 30 degrees C and 40 degrees C. an 8 Ah cylindrical Ni-MH battery is charged in the rate of 1C, 3C and SC to its SOC of 110% in order to provide data for the model training. Linear regression method is adopted to check the quality of the model training, as well as mean square error and absolute error. It is shown that the constructed model is of excellent training quality for the guarantee of prediction accuracy. The surface temperature of battery during charging is predicted under various ambient temperatures of 50 degrees C, 60 degrees C, 70 degrees C by the model. The results are validated in good agreement with experimental data. The value of battery surface temperature is calculated to exceed 90 degrees C under the ambient temperature of 60 degrees C if it is overcharged in SC, which might cause battery safety issues. (C) 2012 Elsevier B.V. All rights reserved.