Large Eddy Simulations for the Prediction of Fuel-Bound NOx Emissions: Application to NH3 and NH3-CH4 Blends at Different Operating Conditions

Ammonia and its blend with natural gas and/or hydrogen will play a significant role in decarbonizing the energy sector in the next decades. Moreover, green fuels can help in reducing the fluctuations introduced by intermittent renewable sources providing an interesting alternative to the storage by the adoption of P2x2P schemes. The challenge GT manufactures are asked to solve is mainly related to the minimization of NOx emissions and the discovery of the optimal operating windows of the engines taking into consideration the potential fuel composition fluctuations associated with the availability of ammonia and other carbon-free fuels. In this context, the definition of dedicated numerical tools able to predict the combustor performance becomes crucial for the re-design of burner architecture allowing detailed analysis of technical solutions prior to experimental tests. The availability of detailed measurements, even at lab scale, represents a milestone for the improvement of these tools, especially for the Computational Fluid Dynamics codes. In the present paper, a Computational Fluid Dynamics (CFD) methodology for the calculation of NOx emissions is developed and validated against unique data collected at Cardiff University’s Gas Turbine Research Center for blends involving methane and ammonia, considering also different pressure levels. The novelty of this methodology, based on the switch between the “in-flame” – “post-flame” contributions, lies in its extension and application for the first time to the fuel-bound NOx formation path. Thus, the main goal of this investigation is the validation of the proposed approach in the framework of this NOx mechanism, hence allowing to overcome the limits of the pre-tabulated combustion models in the prediction of this pollutant. Further, the method is implemented along a computationally cheap model like the FGM, providing an additional benefit in terms of delivery time especially in the context of applied research. Regarding the results, the findings show good agreement with the data: for most of the considered cases, not only the trend is captured, but the numerical results offer also a good quantitative prediction of the tests.