Gas turbines integrated with biomass gasification in a combined cycle power plant (Bio-IGCC) provide a path to power production with very high efficiency. Over 60% fuel-to-power efficiency has been demonstrated with natural gas. The fast ramp and relatively low cost make Bio-IGCC via gas turbines the ideal complement to intermittent power from wind turbines and PV cells. With stricter pollutant regulations and in order to promote the use of renewable fuels there is a great interest in improving fuel flexibility. An important feature of biomass gasification is that its properties vary depending on the feedstock and gasification principle and that the combustion characteristics are significantly different from conventional fuels. This makes it interesting to develop CFD models that can be used to simulate the combustion of syngas in existing gas turbines and for design optimization of new gas turbines. 

The TECFLAM swirl burner geometry, which is designed to be representative of common gas turbine burners, was selected for an assessment of the differences between a typical hydrocarbon fuel and syngas. A two-stage approach was employed with development and validation of an advanced CFD model. The validated model was used to compare the flame shape and other characteristics of the flow between methane, 40% hydrogen enriched methane and four typical syngas compositions. The syngas compositions used are representative of practical gasification processes and biomass feedstocks. It was found that the syngas fuels experience lower swirl intensity due to high axial velocities that weaken the inner recirculation zone. A strong correlation was found between the laminar flame speed and the flame shape. 

The simulation of a typical combustion geometry with syngas is quite demanding and requires a long computational time. In order to speed up the parametric analysis and to make it possible to test more configurations a Two-Step, One Way coupled method was assessed. This is a common approximation in CFD that is used to solve complex problems with limited computational resources. The test case used for the assessment was the CeCOST burner that uses strong swirl for flame stabilization. Only isothermal flow was investigated to eliminate the influence from flow – chemistry interactions. This method effectively divides the domain in two parts, one downstream and one upstream. The assumption behind this method is that the downstream part should not have a big influence on the upstream part and hence it could be solved separately. From the comparison it was found that the full solution and the approximations were in good qualitative agreement. However, there were some minor quantitative discrepancies, and it was proposed that the explanation for the differences could be the slightly different solution approaches that were used for the full simulation (URANS) and the two approximate solutions (RANS). The speed-up from using the approximate method was close to one order of magnitude. 

However, because an artificial steady inlet cannot reproduce all the dynamic phenomena created by a swirler, for the continuation a full CeCOST domain was used. LES modelling was also employed to be able to identify smaller structures that would affect flame stability. Using LES and the Artificially Thickened Flame model, a syngas composition that relates to Black Liquor gasification was modelled. The flame front position using the CH2O mole fraction was estimated and it correlated well with the position estimated by the progress variable. The flame front position found by using the OH mole fraction was different to the two previous ones, predicting the hot part of the flame.

The loss of biodiversity, due to pollution and global warming, is more important and relevant now than it has ever been. Man made climate warming has its roots in the huge amount of CH4 and CO2 emissions, which have increased by 47% and 156% respectively since 1750. To minimise the effects of the already heavily damaged climate, the share of renewables, as well as their efficiency, need to be increased. Besides solar and wind, a promising alternative for countries with a large forest industry is energy through biomass based fuels. Gasification of biomass, already a mature technology, produces a mixture of CO, CO2 and H2 called synthesis gas (syngas). Syngas can be refined into high added value biofuels or used directly in a gas turbine. Syngas combustion in a gas turbine will not replace the aforementioned renewable energy sources but can be used in conjunction with them to account for the electrical grid peak load demand and intermittency issues. However, syngas combustion in a gas turbine gives rise to quite a few challenges as these machines have been optimised for hydrocarbons, and more specifically, methane. Another challenge is that the composition of syngas varies depending on the biomass and gasification method used for its production. It is then necessary to use both advanced experimental methods and Computational Fluid Dynamics (CFD) simulations to assess the combustion behaviour of syngas in a gas turbine environment.

In this work, Reynolds Averaged Navier Stokes (RANS) and Large Eddy Simulation (LES) approaches are used to study the behaviour of syngas combustion from different biomass and gasification principles. Unteady RANS (URANS) was also used to assess good practice guidelines when dealing with complex geometry and flow fields in a burner setup. The numerical modelling is verified and validated against advanced experiments for both burner geometries used in this thesis. The first burner that was investigated (Paper I) was the TECFLAM swirl burner. This work was aimed at comparing the flame shape and other characteristics of methane, 40% hydrogen enriched methane and four typical syngas compositions from practical application (e.g., black liquor gasification). It was found that the syngas fuels experience lower swirl intensity, due to the higher flame speed of the syngas fuels that weaken the inner recirculation zone. A strong correlation was found between the laminar flame speed and the flame shape.

Due to the computational demand of the CFD simulations, a detailed parametric analysis on fuels can become very expensive. Thus, in Paper II, in order to speed up the parametric analysis a Two-Step, One Way (TSOW) coupled method was assessed. The domain under question was the CeCOST swirl burner (similarly for Papers III-V). Using the TSOW method, the solution is divided into an upstream and downstream part, where the inlet conditions from the downstream part (TC) are derived from the upstream solution. The full domain solution (FS) was then compared to the TC solution and they were found to be in good agreement. In addition, for the TC domain a RANS approach could be used for further simplification (an average based method could not be used on the FS due to the flow created by the swirler). However, this introduced further discrepancies between the numerical solutions. The TC numerical solutions using a RANS approach instead of URANS showed a speedup of almost one order of magnitude.

Combustion in the CeCOST burner was investigated experimentally in Paper III for three CO2 dilutions of a black liquor based syngas with a composition of 53% H2, 44% CO and 3% CH4 on a volume basis and a Reynolds number (Re) of 10 000. It was found that the safe operability range is shifted to richer equivalence ratios with increased dilution. In addition, for the higher dilutions, combustion occurred in isolated pockets rather than producing a continuous flame. In Paper IV a LES model, with 53 species and 325 chemical reactions using artificial thickening of the flame, is verified using a quality measure from the LES_IQ method and validated against the experiments on the CeCOST burner. The overall velocity comparison against experiments was significantly good, which enabled the use of the LES model for flame predictions at higher Re. Higher Re cases (20 000 and 25 000) predicted an arrower Inner Recirculation Zone and higher velocity magnitude at the Outer Recirculation Zone, which induced the edges of the flame to move upstream. Unfortunately, the model was not able to capture the flame pocket structure of the experiments. This was attributed to the preferential diffusion mechanism in low Lewis number fuels (hydrogen rich) and lack thereof in the version of the Flamelet Generated Manifold sub model that was used for reactive flow modelling. The flammability limits of these dilutions were also investigated and compared to experimental data (Paper V). The lean blowout limit could be resolved for all cases, but the attempts at flashback simulations, being more complex than blowout, led to severe numerical instabilities. Instead, an assumed flashback limit is introduced prior to where the solution becomes unstable. From these, a numerical stable flame region map is shown, which covers approximately 72% of the experimental range. Similarly to experiments, the numerical results predict the same trend of the stable flame region shifting to richer conditions with increased CO2 dilution. In addition, it was found that going from lean blowout to flashback the flame shape changed from a compact V-shape to a M-shape.