Computational modelling of emissions from gas turbines: Ensuring experimental reliability and quantifying emissions uncertainty

Global environmental concerns and the increasing amount of power generated using gas turbines have led to the development of dry low emission (DLE) combustion systems to minimize the NOx and CO pollutants. Improvements in current emissions modelling approaches are essential to precisely calculate emissions and develop probabilistic emissions prediction tools for uncertainty quantification (UQ) studies. An accurate chemical kinetic mechanism for natural gas and NO chemistry is the basis of emission modelling. Rapid compression machines (RCMs) as laboratory instruments are widely used to develop chemical kinetic mechanisms in a same and much more controllable environment than actual gas turbine combustion system. However, a critical attribute of RCMs is temperature inhomogeneity which needs to be addressed. Because of the exponential dependence of rate constants on temperature, even minor departures from homogeneity can significantly complicate the interpretation of experimental results to develop accurate chemical kinetic mechanisms. Therefore, the first goal of this study is to predict temperature inhomogeneity and propose a framework for RCM testing campaigns to develop chemical kinetic mechanisms accurately. A simple and general framework is implemented to develop a correlation for temperature inhomogeneity in RCMs. A set of CFD simulations using three-dimensional large eddy simulation (LES) and two-dimensional laminar model are then implemented to validate the developed correlation and propose the RCM testing campaigns. In addition, most of the available detailed deterministic models for emissions modelling are suitable for a single analysis and may be inappropriate for UQ studies since the computational costs can quickly become prohibitive. The computational cost is also a significant concern in traditional UQ methods due to many evaluations of the emissions prediction model. Therefore, the second goal of this study is to develop a novel and computationally-efficient framework for UQ study of emissions in gas turbine combustion systems. Non-intrusive polynomial chaos expansion (NIPCE) based on point collocation method (PCM), Sobol sensitivity indices and first-order reliability method (FORM) are coupled with chemical reactor network (CRN) modelling approach in Python to develop the UQ-enabled emissions prediction tool. The UQ-enabled tool is then implemented on a high-pressure premixed burner and a swirl-stabilized premixed burner. The results show the capability of the developed framework as a computationally-efficient approach for UQ study.

Funder

College of Engineering and Informatics, National University of Ireland, Galway

Publisher

NUI Galway

Rights

Attribution-NonCommercial-NoDerivs 3.0 Ireland

cbn

Collections

University of Galway Theses (PhD Theses)