This paper studies computational algorithms for the thermal state of turbine generators active parts. There are well-known algorithms, based on ventilation substitution schemes, which do not allow considering the nature of the flow of the cooling medium in detail, but form averaged initial data for the formulation of the heat transfer problem. Resource-intensive CFD calculations are also used to simulate a full ventilation path. Here, an applied hybrid model is proposed, combining advantages of the described approaches and taking into account the return coupling between equivalent resistance change and mass flow of the ventilation path elements in multivariate simulations. This model allows us to increase the calculation speed and the results accuracy. The proposed approach is verified by comparing the simulated results with the experimental data.