Siemens SGT-800 gas turbine is the largest industrial gas turbine within Siemens medium gas turbine size range. The power rating is 53MW at 39% electrical efficiency in open cycle (ISO) and, for its power range, world class combined-cycle performance of >56%. The SGT-800 convectively cooled annular combustor with 30 Dry Low Emissions (DLE) burners has proven, for 50–100% load range, NOx emissions below 15/25ppm for gas/liquids fuels and CO emissions below 5ppm for all fuels, as well as extensive gas fuel flexible DLE capability.
In this work the focus is on the combustion modelling of one burner sector of the SGT-800 annular combustor, which includes several challenges since various different physical phenomena interacts in the process. One of the most important aspects of the combustion in a gas turbine combustor is the turbulence chemistry interaction, which is dependent on both the turbulence model and the combustion model. Some turbulence-combustion model combinations that have shown reasonable results for academic generic cases and/or industrial applications at low pressure, might fail when applied to complex geometries at industrial gas turbine conditions since the combustion regime may be different.
Therefore is here evaluated the performance of Reynolds Averaged Navier-Stokes (RANS) and Scale Adaptive Simulation (SAS) turbulence models combined with different combustion models, which includes the Eddy Dissipation Model (EDM) combined with Finite Rate Chemistry (FRC) using an optimized reduced 4-step scheme and two flamelet based models; Zimont’s Burning Velocity model and Lindstedt & Vaos Fractal model. The results are compared to obtained engine data and field experience, which includes for example flame position in order to evaluate the advantages and drawbacks of each model. All models could predict the flame shape and position in reasonable agreement with available data; however, for the flamelet based methods adjusted calibration constants were required to avoid a flame too far upstream or non-sufficient burn out which is not in agreement with engine data. In addition both the flamelet based models suffer from spurious results when fresh air is mixed into fully reacted gases and BVM also from spurious results inside the fuel system. The combined EDM-FRC with a properly optimized reduced chemical kinetic scheme seems to minimize these issues without the need of any calibration, with only a slight increase in computational cost.