470530 Validation of Agglomeration Modeling Methodology and Proposed Mechanism of Ash Agglomerate Growth in Fluidized Bed Combustors

Tuesday, November 15, 2016: 4:12 PM
Peninsula (Hotel Nikko San Francisco)
Aditi Khadilkar, Energy and Mineral Engineering Depaartmentand EMS Energy Institute, The Pennsylvania State University, University Park, PA, Sarma Pisupati, Energy and Mineral Engineering Department and EMS Energy Institute, The Pennsylvania State University, University Park, PA and Peter Rozelle, US Department of Energy, Germantown, DC

Increase in particle size occurs during several processes that use fluidized beds. While it is desirable in processes such as granulation and pellet-making its occurrence can cause undesirable operational issues in combustion and gasification processes. Models that incorporate interdependencies and variations in both bed-hydrodynamics and binder chemistry are required for accurate predictions of agglomerate growth kinetics. Penn State has developed a unique two-particle collision-based model that combines these two aspects. The model is tested with poly-disperse fluidized bed combustion and gasification systems.

In fluidized bed gasification and combustion systems, agglomeration occurs by sticking of fuel ash particles that are wetted by slag-liquid. The non-uniform temperature, gaseous atmosphere and heterogeneity in ash chemical composition need to be accounted for in the prediction of the slag-liquid binder formation and chemistry. Additionally, the poly-dispersity of bed ash, particle velocities and collision frequencies are the important physics-based factors that affect particle growth rate. In the Penn State model, thermodynamic equilibrium calculations are used to estimate the amount of slag-liquid in the system, and the changes in particle collision frequencies are accounted for by continuously tracking the number density of the various particle sizes. The particle number density is also affected by the ash chemistry as solid particles melt to form slag. Computational fluid dynamics modeling is used in conjunction to obtain the initial granular physics inputs to the model. Thus, the heterogeneities in ash chemistry and granular physics are integrated in this model. The model development was presented previously and the present study focuses on validation of the modeling results and the methodology developed. For this purpose, ash agglomerates were generated in a laboratory-scale fluidized bed combustor at Penn State. Ash agglomerates were produced by operating the reactor at a superficial gas velocity close to the minimum fluidization velocity of the particles, using rejects from a bituminous coal with about 82 % ash content, under oxidizing conditions. Agglomerate samples were also obtained from another fluidized bed reactor in Canada. Polished cross-sections of these agglomerates were studied using scanning electron microscopy with energy disperse x-ray scattering (SEM-EDX) to relate the particle-level slag formation and sticking to the chemical composition. FactSage simulations of the slag-forming components were used to estimate the agglomeration temperatures.

The results of this study indicated that agglomerate growth in fluidized bed combustors (FBC) is initiated at the particle-level by low-melting components rich in iron- and calcium-based minerals. Although the bulk ash chemical composition does not indicate potential for agglomeration, study of particle-level heterogeneities revealed that agglomeration can begin at lower temperatures than the FBC operating temperatures of 850 °C. After initiation at the particle-level, more slag is observed to form from alumino-silicate components at slightly higher temperatures caused from changes in the system, and agglomerate growth propagates in the bed. A post-mortem study of ash agglomerates using SEM-EDX helped to identify stages of agglomerate growth.

Additionally, the Penn State ash agglomeration model developed was used to simulate agglomerate growth in a laboratory-scale fluidized bed combustor firing palm shells (biomass), reported in the literature. A comparison of the defluidization time obtained by simulations to the experimental values reported in the case-study was made for the different operating conditions studied.

Extended Abstract: File Uploaded