Validation of a Numerical Approach to Model Hydrogen Combustion in Gas Turbine Engines

Abstract

Abstract Hydrogen as a gas turbine fuel presents significant challenges for combustor design. Hydrogen has a much higher diffusivity and flame speed than traditional gas turbine fuels, and new combustor designs are required to run with high hydrogen blends or full hydrogen. Computational fluid dynamics (CFD) codes are important tools to aid in the design of combustors, so accurate prediction of hydrogen flame behavior is necessary. Accurate numerical models with fast runtime are needed for these applications. While CFD approaches are well established for gaseous methane combustion in air, hydrogen’s non-unity Lewis number, high diffusivity, and high flame speeds pose a challenge to flame modeling for gas turbine-relevant combustion regimes. Emissions and flame length behavior are difficult to capture correctly. In this study, a commercial CFD code (CONVERGE CFD) was used to evaluate combustion and flow modeling approaches for gas turbine hydrogen fuel injectors. The approaches were judged in the context a future broad design study, where a single approach must be able to simulate both premixed and non-premixed flame behavior. Both partially premixed and fully premixed experimental datasets were considered to account for a wide range of possible combustion problems. The partially premixed cases were based on jet in crossflow (JICF) fuel injection into cylindrical tubes and were part of NASA studies in the early 2000s (Schefer et al. [1] and Marek et al. [2]). The fully premixed case was from more recent work by Jin & Kim [3]. The partially premixed cases attempted micromixing, while the fully premixed case performed the premixing before injecting the mixture into swirlers. These two cases capture a range of fluid physics and chemical kinetics that are relevant to the simulating flow in hydrogen fuel injectors for gas turbines. For capturing the flow, Reynolds-averaged Navier-Stokes (RANS) turbulence models have historically been popular with industrial CFD applications. Typical hydrogen fuel injectors like micromixers have complex flow fields with significant unsteadiness. RANS simulations may not be able to fully capture the fuel-air mixing within micromixers. Current fuel injector design requires higher-fidelity spatial and temporal discretization for accurate flow simulations. In this case, and investment in the costlier large eddy simulation (LES) can improve simulation fidelity significantly. For the flame, FGM and detailed chemistry modeling approaches (SAGE) are evaluated against experimental studies to assess their suitability for use in hydrogen combustion. For flow simulations, steady RANS and unsteady LES approaches are compared. Experimental flame length, NOx production, and other metrics are used to evaluate the performance of the various combinations of combustion and flow models against experiments. The results suggest the use of detailed chemistry with LES to be the most appropriate for general gas turbine hydrogen combustion modeling across a range of possible operating conditions, even considering the increased cost of this approach.