Investigation of a prototype industrial gas turbine combustor using alternative gaseous fuels

Sammanfattning: In this thesis, the effect of alternative gaseous fuels, with high hydrogen content and lower calorific value, on gas turbine combustion was investigated experimentally. The aim of the investigation was to find operational limitations for an experimental burner and to supply data for validation of computational fluid dynamics (CFD). Before examination of the actual burner, the laminar flame speed was measured for a range of gases. The measurement technique was based on Schlieren imaging which is a measure of the density gradient through a flame surface. A Bunsen type burner was used to measure the angle of a conical flame from which the laminar flame speed was calculated. In order to improve the comparability of these measurements with other measurement methods the laminar flame speed was corrected for the influence of stretch. The effect of stretch will increase or decrease the flame speed depending on the curvature of the flame and the physical properties of the gases involved in the combustion, e.g. the Lewis number and preferential diffusivity. The gas turbine burner examined was a downscaled version of the burner that is now found in the commercial gas turbine, SGT-750. The burner consists of three concentric sections. The central part is a precombustor called rich-pilot-lean (RPL). The purpose of the RPL is to supply heat and radicals to the other sections to stabilize combustion. The next section is the Pilot, which serves as an intermediate burner in which the equivalence ratio can be optimized to stabilize combustion and minimize NOX emissions. The outermost section is the Main. For the experimental burner approximately 79% of the mass flow passes through this section. All sections have their own swirlers that create recirculation zones for flame stabilization. The experimental work in this thesis includes measurements of the lean stability limit, emission optimization (primarily NOX), flame diagnostic through OH-Laser induced fluorescence (LIF) and particle image velocimetry (PIV). Tests were conducted at both atmospheric conditions with preheated air (650 K) and at elevated pressure up to 9 bar. Results from the experimental investigations were also used to validate CFD computations using reduced chemical kinetic schemes, and to validate reactor network calculations based on perfectly stirred reactors (PSR) and plug flow reactors (PFR). Lean stability limit experiments showed how the RPL equivalence ratio could be optimized to lower the lean blowout limit. Increasing the RPL equivalence ratio was shown to extend the lean blowout limit, up to a limit after which the RPL flame was quenched. Reactor network modelling showed that the stabilizing effect of the RPL was a combination of thermal energy and reactive radicals supplied to the flame zone. The important radicals were shown to be H, O and OH. The emission optimization measurements showed that lowering the equivalence ratio in both the RPL and the pilot minimized the NOX emissions. CFD simulation showed that the degree of mixing of both the RPL and the Pilot at point of ignition was not perfect. Imperfect mixing causes pockets of stoichiometric mixtures to react, which in turn create hot spots where thermal NOX can be formed. At rich RPL equivalence ratios, a flame could be visualized with OH-LIF after the RPL exit. This flame probably to some extent combusts closer to stoichiometry, which increases thermal NOX. These theories of how NOX is formed were confirmed by reactor network calculations.