OCR Text |
Show given furnace. However traditional turbulent reacting flow models face well-known difficulties in achieving such predictions. These are due to the modeling required in all time-averaged representations of the turbulent transport and mixing-chemistry coupling processes. All traditional codes are based on such time-averaging of the governing equations, which leads to turbulent transport terms that must be modeled with diffusion-like representations of the turbulent mixing. In this key respect, the various codes available differ only in the comparatively minor issue of how they choose to represent the resulting fictitious turbulent (or eddy) diffusivity in their diffusion-like approximation of turbulent mixing. Turbulence research over the past two decades has clearly shown, however, that it is the underlying diffusion-like representation in all traditional models that is fundamentally incorrect. Such a representation of turbulence ignores the dominant role of large-scale structures in the turbulent mixing process, which cannot be modeled by a diffusion-like representation. For this reason, turbulence modeling is moving away from the traditional time-averaged approach, and toward a new class of models based on "large eddy simulations". Large eddy simulations allow time-dependent computations of the large-scale structures that are key to the turbulent mixing processes, and thereby avoid the traditional diffusion-like representation of these scales. The Local Integral Moment (LIM) model [ 1-7] is based on these large-eddy simulation concepts, and has been applied to numerically simulate the turbulent mixing and chemical reaction processes involved in gas reburn applications [4]. Unlike the traditional approach to turbulence modeling, LIM simulations involve no time-averaging, and instead are based on a Lagrangian front-tracking method in which the time-evolution of flow and mixing are computed directly. A detailed summary of LIM modeling was presented at the 1996 AFRC International Symposium [61 and elsewhere [1-3. 5. 7] and its application to FLGR modeling in utility boilers was presented at the 1997 AFRC International Symposium [4]. Here, LIM-based modeling is applied to assist in development of an FLGR system for the Riverbend Unit 7 boiler. 4. FLGR SYSTEM DESIGN FOR FUVEFtBEND UNIT 7 Design of an FLGR system for any particular boiler involves determining the elevation(s) at which gas is to be injected, the number of individual injectors to be used at each elevation, the injector layout at each elevation, the sizes and orientations of individual injectors, the gas flow rate(s) to be delivered though each injector, and any recommended adjustments to the excess oxygen level at which the boiler is to be operated. The system is typically designed to meet specified emissions targets over a range of boiler loads. For Riverbend Unit 7, these targets were to reduce NOx emissions to 0.25 lb/MMBtu at full load operation (140 M W ) and at lower load conditions (90 M W and 65 MW) using no more than 7 % gas, while keeping CO emissions no higher than 200 ppm above the baseline level and avoiding adverse impacts on steam temperature, slagging, LOI, and stack opacity. Energy Systems Associates (ESA) and NGB Technologies worked together to develop a suitable design for the FLGR system at Riverbend Unit 7. Since this was the first implementation of FLGR technology in a tangentially-fired boiler, it required |