Rotor endturn temperature predictor in large air-cooled turbo-generators

Abstract

The increasing demand for more electricity have brought about the need for improved rotor ventilation performances of the large industrial turbo-generators. Overheating in the rotor endturn region under the expected operating conditions is one of the major ventilation concern for the air-cooled machines, but a fast and accurate temperature monitoring method is not yet available. The rotor endturn coil temperature evaluations are currently heavily relied on the time-consuming computational fluid dynamics simulations, which posses difficulties in the rotor ventilation design. This presented work develops a fast-solving analytical predictor, which can estimate the rotor end-arc coil temperatures at the expected operating conditions. The predictor is developed using the lumped-circuit method, which has been implemented by many researchers to successfully apply for many electric machines. To do so, the fluid properties in the nested cavities adjacent to the heated coils are first numerically captured with detailed CFD simulations. The cavity flow phenomenon is inspected and a virtual pipe method is proposed to describe and quantify the observed internal bulk flow characteristics. The virtual pipe hydraulic diameters are quasi-analytically determined, by mathematically matching the CFD extracted velocity profile with the expected theoretical profile for the internal turbulent pipe flow. Generic correlations of the defined local virtual pipe segments are developed in terms of the identified design parameters, as a tool to estimate the Dh values for any design scenarios within the applicable ranges of the design parameters. A set of CFD parametric studies are conducted to provide a reasonable sampling set to use for the correlation development. Finally, a thermal-fluidic network is developed by implementing the virtual pipe description of the cavity internal bulk flow, which accounts for convective and conductive heat transfers at the coils to calculate the coil temperatures under different heat loads. The predictor is verified with both the reported factory sensor readings and CHT simulation, and is concluded to be able to achieve the desired ±5 degrees of Celsius error.