| Both a dynamic computer simulation model and an internal model control system have been developed and experimentally verified for a pressurized fluidized bed coal gasification reactor. A Texas lignite and a New Mexico subbituminous coal were the reactor feeds.; The lumped parameter phenomenological model uses kinetic and equilibrium models to account for the important reactions and incorporates mass transport and energy transfer rates. Where equations from the published literature were not applicable, reactor experiments were conducted to develop the required relations (devolatilization yields, water-gas shift reaction extent, and particle size dynamics are prime examples). Some of the hydrodynamic and kinetic expressions needed to describe the reactor performance involve system-dependent adjustable parameters, three of which are defined by feedback of measured steady-state on-line reactor data.; Desired control of the reactor is hindered by a two hour dead time, non-measureability of one variable, multivariable coupling, and disturbances; but control was achieved through a supervisory strategy that uses the adaptive model and is termed internal adaptive-model control (IAMC). To surmount the dead time and unmeasured variable problems, the controller uses the model to estimate the present reactor state and controls the reactor inferentially. Model adaptation accounts for disturbances, keeps the model true and useful for control extrapolations. Four operational objectives (methanol-equivalent production rate, H(,2)/CO ratio, carbon utilization, fluidized bed height) are controlled through four manipulated variables (coal, steam, and oxygen feed rates, spent ash removal rate). The controller has been tested in a servo-control mode.; Complete experimental data from fifty-nine gasifier runs are included.; Conclusions include: (1) The water-gas shift reaction is at equilibrium within the bed. (2) The in-bed particle size distribution does not change as the reactor changes operating state. (3) The in-situ coal devolatilization yields can be "backed-out" of the normal reactor operating data. (4) The lumped parameter model is an excellent dynamic simulator. (5) In limited testing in the servo-control mode, the IAMC strategy works. |
