The reasonable operation reliability evaluation of transformer is the basis of condition based maintenance. Firstly, the influence of gas content, gas production rate and fault time distribution parameters on transformer operation reliability is analyzed, and the dynamic failure rate model is set up. Then, aiming at the inherent randomness of oil-immersed transformer fault time, the Markov Chain Monte Carlo method is used to simulate the transformer failure cycle, and the statistical method is used to solve the historical failure rate of the transformer as the training data, so as to solve the related parameters of the model and effectively avoid the difficulty of direct solution by analytical method. Finally, the short-term and long-term failure rate of the transformer is predicted and compared with the traditional model. The results show that considering the influence of failure time distribution parameters on transformer operation reliability, the prediction accuracy of the model can be effectively improved. This work is supported by National Key Research and Development Program of China (No. 2016YFB0900500).