A dynamic reduced order model for simulating entrained flow gasifiers. Part II: Model validation and sensitivity analysis

Abstract Part I of this series describes a dynamic reduced order model (ROM) that has been developed in Aspen Custom Modeler (ACM) for a range of entrained flow gasifiers (EFGs) [1] . The ROM incorporates submodels for multiple feedstocks, mixing and recirculation, particle properties, drying and devolatilization, chemical kinetics, fluid dynamics, heat transfer, pollutant formation, slag behavior and syngas cooling. This paper describes ROM validation for steady-state simulation of four entrained flow gasifiers for which experimental data is available, and sensitivity analysis for the GE gasifier design. The throughputs of these gasifiers range from 0.1 to 1000 metric tonnes per day (tpd) (3 kW th –240 MW th ). Gasifier designs vary widely and simulations encompass the following configurations: dry and slurry feed, oxygen and air blowing, up and down flow, one and two stages, membrane and refractory lining, and quench and radiant cooling. Available experimental data consists of axial profiles for temperature, gas composition and carbon conversion, as well as exit values for temperature, composition, carbon conversion, char flow rate, syngas heating value and cold gas efficiency. Results show satisfactory ROM accuracy for all four gasifier designs simulated, which increases with knowledge of design, operating conditions and experimental data. In cases, where more detailed models that incorporate computational fluid dynamics (CFD) have been used for gasifier simulation, the ROM exhibits comparable accuracy. In sensitivity analysis, important input parameters are identified and varied 10% around their base case (validation) values. The resulting changes in ROM-predicted gasifier performance revealed that the most important parameters are those that determine reactor network model (RNM) geometry (reactor sizes and mass flow rates), particle physical (porosity, surface area, density, etc.) and kinetic properties, and slagging. In addition, the ROM takes about 1 min to run on a desktop PC, while CFD-based models can take multiple days on multiple processors.