Modelling of methane and hydrogen enriched methane flames in industrial gas turbine burners

The majority of the power production in the world today is based on combustion of coal, oil and natural gas. Most countries have agreed that the amount of CO2 emitted to the atmosphere will have to be decreased to keep the global warming at sustainable levels. One way of reducing the CO2 emitted to the atmosphere is by introducing hydrogen co-firing. The hydrogen may be produced from renewable sources and can be used when the solar and wind power are not producing electricity. Another way of reducing the emitted CO2 is to increase the gas turbine efficiency. The two best ways of increasing the efficiency is to increase the turbine inlet temperature and to reduce the cooling air usage. There are also increasing demands on decreasing emissions toxic to humans, such as nitric oxides NOx and carbon monoxide (CO). The NOx emissions are increasing exponentially with the firing temperature. To reduce the NOx emissions lean premixed combustion is used in most modern industrial gas turbines. The gas turbine is operated close to the lean blow out limit, sometimes with stability issues as a result. To keep the NOx and CO emissions to a minimum and at the same time increase the hydrogen addition and the firing temperature while reducing the cooling air usage, accurate predictive tools are required to give good estimates of the flame shape and position, the wall heat load close to the flame, the turbine inlet temperature profile and if possible the combustion dynamics and emission performance.

This thesis aims to explore the usage of scale resolving turbulence models combined with flamelet based combustion models using both methane and hydrogen enriched methane. The main focus areas are predictions of mean flame shape and position as well as flame dynamics. The burner studied is the Siemens 3rd generation DLE burner. Both scale adaptive simulations (SAS) and large eddy simulations (LES) are applied where stationary flamelets combined with a fractal combustion model is used in the SAS case and flamelet generated manifolds integrated across presumed PDFs are used in the LES case. The simulation results are compared against measurement data including OH-PLIF, dynamic pressure and static pressure drop. The usage of non-scale resolving methods show that the flame location is unaffected by adjusting the combustion model constant, which directly affects the mean reaction rate, making them unsuitable for predictions of flame movements due to different fuels. Both scale resolved methods applied here does a good job in predicting the change in flame shape and location when introducing hydrogen enrichment. The SAS-fractal reaction rate model constants had to be adjusted to get the flame position similar to the measurement data whereas no adjustments were made in the LES-FGM case. The shape of the PDF for the reaction progress variable is investigated in the LES-FGM case with only minor differences in flame shape and position as a result.The combustion dynamics is fairly well predicted using the SAS-fractal model but excellent agreement with measurement data is only achieved using the LES-FGM model.