Search

Creation of an autothermal system by coupling endothermic and exothermic reactions demands matching the thermal requirements of two reactions. Autothermal operation is applied here to a solid oxide fuel cell with internal methane steam reforming (IR-SOFC), for which the excess heat generated in the cell provides the heat required by the endothermic reforming reactions. However, such coupling is not easy to achieve because of the mismatch between the thermal load associated with the reforming rate at typical SOFC temperatures and the local amount of heat available from the fuel cell reactions. This paper compares the performance of a SOFC with two different internal reformers and investigates the dynamic response of a coated-wall indirect IR-SOFC to changes in current density. It is shown that a coated-wall indirect IR-SOFC design could be effective in reducing the IR reaction rate, preventing steep temperature gradients, and increasing efficiency. Dynamic simulations show that, after a positive current density step-change, the intermediate period between the disturbance imposed and the new steady-state is characterized by cell potential undershoot and hydrogen conversion and temperature increase. The SOFC unit studied has a total fuel residence time of 0.22 s, for which the response is essentially complete in 320 s.

This work examines the relative advantages and limitations of carrying out both simulation studies and experimental tests in order to evaluate the effectiveness of advanced algorithms for process control. Currently (and for good reason), the most common method of carrying out such evaluations is by simulation alone. Experimental testing is difficult and often leads to ambiguous conclusions. Nevertheless, there are some practical aspects of estimation and control algorithms which are revealed in experiments, but which never present themselves in simulation studies. Some of these are crucial for successful industrial implementation. The results reported here come from both simulation studies and experiments on three pilot plants available at Imperial College: a large-scale CO 2 absorption/desorption plant; a single-tube full-scale reactor for gas-phase catalytic partial oxidation of hydrocarbons; a partially-simulated stirred reactor for general liquid-phase exothermic reactions. Examples which are illustrated include case studies on both linear and nonlinear algorithms of identification, estimation, and control. Some of the specific control algorithms which have been studied were: nonlinear geometric control, linear and nonlinear model predictive control, controllers based upon neural networks, and inferential controllers based upon on-line estimation of unmeasurable states. Comparison of simulation and experiment revealed that, in some cases, the experiments showed qualitatively different behaviour from even the most detailed simulation studies. Reasons are put forward as to why such large differences occasionally exist between the two types of tests. Among the most important reasons for these differences are the uncertain nature of plant-model mismatch in nonlinear systems, the crucial role of process and controller constraints (especially in experimental studies), and the sensitivity of some algorithms to the exact implementation of their calculated control action—something which is virtually impossible in practice. The case studies confirm the complementary nature of the important roles played by both simulation studies and detailed experimental tests in an industrially realistic environment.

Recently, the use of control strategies based upon inverse process models for non-linear systems has been found promising. The requirement of a true analytical inverse can be avoided when neural network models are used; they have the ability to approximate both the forward and the inverse system dynamics. Although many simulation studies have illustrated the use of neural network inverse models for control, very few on-line applications have been reported. This paper describes a novel implementation of a neural network inverse-model based control method on a experimental system—a partially simulated reactor, designed to test the use of such non-linear algorithms. The implementation involved the control of the reactor temperature in the face of set point changes and load disturbances despite the existence of significant plant/model mismatch. Comparison was also made with conventional PID cascade control in several cases. The results obtained show the capability of these neural-network-based controllers and, incidentally, point out the differences between simulation studies and on-line experimental tests. Since the system in this study was only mildly non-linear, in some cases, the performance was comparable to that achieved by classical controllers while in other cases an improved control was achieved.