Abstract This paper extracts a comprehensive reliability model for a transformer with oil natural air forced (ONAF) cooling system, which is the most common type of power transformers. The transformer is first classified into three subsystems and a reliability model is thereafter developed corresponding to each subsystem. Markov process and frequency/duration approaches are applied to analyze the extracted models and achieve their equalized representations. Merging subsystem models leads to the transformer entire reliability model consisting 17 states. The model is then reduced to an 11-state one which is more computationally manageable. Numerical analyses and sensitivity studies are conducted to demonstrate the performance of the proposed model and the results are thoroughly discussed.