| A model of coal combustion dynamics in a fluidized bed is developed in this study. The model is carried out on the premise that the total combustion response can be decomposed into two distinct processes: volatile combustion and char combustion. For an impulse in the feed rate of coal, the transient response of volatile combustion is modeled as an exponential decay. In contrast with volatile combustion, the transient response of char combustion cannot be approximated by an exponential decay. Instead, the transient response of char combustion is based on a closed-form solution to a population balance equation. This solution is used in conjunction with primary fragmentation distributions to explain the wide differences observed between the transient responses of small and large coal particles. To verify this model, theoretical responses are compared with experimental data obtained from batch tests of coal with known initial particle distributions. Experimental results suggest that particle distributions--even narrow distributions--significantly affect the transient response of coal combustion in fluidized beds.; Besides developing a nonlinear transient combustion model, system identification algorithms are used in this study to develop linear transfer functions that characterize coal combustion dynamics. However, as with many chemical processes, data signals from fluidized beds invariably contain measurement noise. Consequently, this investigation examines the applicability of various system identification algorithms to processes with significant measurement noise. Included in this investigation are studies of discrete-time domain, frequency domain, and continuous-time domain algorithms.; Apart from individual methodologies and algorithms, this study highlights some of the difficulties inherent with identification of processes corrupted by measurement noise. While each group of identification algorithms has shortcomings, performance tests suggest that continuous-time systems are best modeled through continuous-time parameter identification techniques, such as the Poisson moment functional approach. |
