Now the lithium ion batteries are widely used in electric vehicles (EV). The cycle life is among the most important characteristics of the power battery in EV. In this report, the battery cycle life experiment is designed according to the actual working condition in EV. Five different commercial lithium ion cells are cycled alternatively under 45 degrees C and 5 degrees C and the test results are compared. Based on the cycle life experiment results and the identified battery aging mechanism, the battery cycle life models are built and fitted by the genetic algorithm. The capacity loss follows a power law relation with the cycle times and an Arrhenius law relation with the temperature. For automotive application, to save the cost and the testing time, a battery SOH (state of health) estimation method combined the on-line model based capacity estimation and regular calibration is proposed. (C) 2014 Elsevier B.V. All rights reserved.