Flashback prevention in a hydrogen-fueled original micro Gas Turbine without any combustor redesign: Large-Eddy Simulation verification of a strategy based on 0D/1D predictions

Premixed burners, commonly used in micro Gas Turbines (mGT), are prone to flame instabilities when fueled with hydrogen or hydrogen blends. Humidification and Exhaust Gas Recirculation (EGR) provide solutions to prevent flashback in an mGT burner without requiring any combustor redesign. Currently, these solutions are rarely considered, and no prediction methodology exists. Therefore, a low-complexity method using a hybrid 0D/1D model was proposed in previous work to predict the dilution levels necessary to prevent flashback, based on an easily accessible parameter, the laminar flame speed. This work aims to verify the potential of humidification and EGR, as well as the predicted dilution levels generated by the 0D/1D model, through high-fidelity simulations. To achieve this, Large-Eddy Simulations (LES) of an original mGT combustor layout are performed using a complex chemistry approach with an Adaptive Mesh Refinement methodology. These simulations compare pure methane combustion, considered as a reference case, to a case with a CH 4 /H 2 blend of 50/50% v o l , highlighting flashback, and to cases that implement solutions to limit this flashback – such as humidification and EGR – for different hydrogen fractions, 50% and 100%. The numerical results show that flashback is prevented in all solution cases. The 50% H 2 case with humidification exhibits behavior most similar to the reference case. Whereas the EGR case results in a longer and colder flame, the case with 100 %H 2 plus water shows a shorter, more reactive flame with a high turbulent flame speed near the flame foot, indicating conditions near the limit of flashback disruption. • Flashback limitation without any redesign of a micro Gas Turbine combustor. • Large-Eddy Simulations of hydrogen combustion with Adaptive Mesh Refinement. • Performing humidification and EGR as a solution to avoid flashback. • Verification of 0D/1D predictions for diluted conditions to limit flashback.