| dc.description.abstract | The major challenge in designing oxy-fuel combustors in supercritical power generation cycles is the lack of fundamental understanding of the combustion process at high pressures exceeding the critical point of the fuel and oxygen (~200 bar). This combustion process becomes unique as it uses supercritical CO2 (sCO2) streams with high dilution rates to achieve lower combustor exit temperature (~1150ºC). In the absence of reliable combustion diagnostics at such high pressure and temperatures, a high-fidelity yet numerically efficient modeling framework is needed to enable the design of sCO2 oxy-combustors. Reactive turbulent combustion simulations are computationally very expensive which can be prohibitive when the combustion regime of interest is supercritical due to the challenges in implementing real-gas equation of state and considering multi-step species and reactions in combustion modeling. To overcome these challenges, two different computational frameworks have been developed in this study: (1) A Direct Numerical Simulation (DNS) integrated with a one-step chemistry mechanism and multispecies real-fluid properties to gain a fundamental understanding of supercritical mixing and combustion in the sCO2 oxy-combustor designed by Southwest Research Institute (SWRI). (2) A high-fidelity turbulent reacting flow simulation Large Eddy Simulation (LES) real gas solver at supercritical combustion using a detailed chemistry mechanism comprised of several species and reactions for methane combustion is developed. To enable computational feasibility a new high accuracy fast chemistry reduction tool using machine-learning based tools is developed and validated a-priori and a-posteriori. The developed tool is integrated with an open-source real-gas combustion solver and 3D Large Eddy Simulations (LES) are conducted for SWRI combustor.
The results indicate noticeable distinctions between mixing and combustion behaviors predicted by ideal- and the real-gas Equation of State (EoS), particularly in the near-inlet region. The reacting flow predicted by the real-gas EoS showed a wider density gradient barrier which tends to impede mixing, a longer wavelength at the flame edge, and a smaller flame edge thickness. Differences were also noted between the consumption/production rates of key species between real and ideal cases where the ideal-gas case predicted higher rates due to higher heat release rate caused by differences in density and specific heat. Combustion behavior at different %CO2 dilution showed that dilution has a major impact on key combustion metrics such as heat release rate, temperature, and flame edge thickness. Zero-D and DNS results showed a peak in heat release rate for a given air-fuel ratio and the lowest CO production for 75%-80% CO2 dilution with a maximum flame temperature of 2000 K. The results will provide crucial insight for designing sCO2 oxy-combustors. A new reduction tool called the Supervised Learning Global Pathway Selection (SL-GPS) is developed and validated with a standard deflagration flame (Sandia Flame D) experiment. The new tool is integrated with the real gas effects for ILES of the SWRI combustor. The SL-GPS is able to capture the flame characteristics within satisfactory error margin as compared to the detailed case. The ILES of SWRI combustor shows unique ignition and flame kernel expansion behavior. The ignition occurs in the radially outward region of the injection at a high turbulence zone. The ignition kernel is composed mainly of the CO mass fraction among the minor species which indicates incomplete combustion. This CO mass fraction reduces as the flame propagates and secondary oxidation takes over converting CO to CO2. The flame expands radially and axially and interacts with the wall where major swirl motion takes over and the flame collapses onto itself leading to formation of a thin stable flame centered on the axis. This stable flame is closer to the injection point as compared to the original ignition location. | |
