Prediction of Cooling Air Flow in Electric Generators

Doctoral thesis, 2013

The cooling air flow in hydro power electric generators is investigated experimentally and numerically. A fully predictive numerical approach is presented and validated, in which the inlet and outlet boundaries are eliminated from the computational domain. Instead, a part of the space outside the machine is included in the computational domain, allowing for recirculation of the cooling air. The predicted flow is therefore driven solely by the rotation of the rotating parts of the generator. In this way, the predicted flow field is independent of any experimental data at the inlet, and is determined completely by the solution. Using the fully predictive approach, a number of parametric numerical studies are performed on the rotor and stator geometries. The effect of adding geometrical details to the rotor and stator are investigated, and stator baffles and rotor fan blades are concluded to increase the volume flow rate through the machine. The volume flow rate through the machine is found to vary linearly with the rotor rotational speed, while the required rotor axial power increases cubically with the rotor rotational speed. The numerical results are validated against experimental measurements in a real electric generator. Flow visualizations, and 5-hole probe and total pressure measurements are performed. A comparison of the numerical results and the experimental data reveals a good qualitative prediction of the flow by the fully predictive numerical approach. The sensitivity of the numerical results to different choices of inlet boundary conditions is also investigated. The level of detail in the boundary conditions proves to play an important role in predicting correct flow features. A half-scale laboratory model, based on the above studied electric generator, is specifically designed and manufactured for experimental studies of the cooling air flow. The measurement accuracy in the half-scale model is significantly improved compared to that in the real generator. The model is provided with static pressure holes and optical access for flow measurements using Particle Image Velocimetry (PIV). The fully predictive numerical approach is shown to yield quantitatively similar results as the experimental flow measurements. The numerical simulations are also performed with inlet and outlet boundary conditions, by specifying the inlet volume flow rates from the experimental measurements. The results of the fully predictive numerical approach are shown to agree better with the experimental data, than those of the simulations with inlet and outlet boundary conditions.