(373f) Estimation of Faults in Water Gas Shift (WGS) Reactor Using Extended Kalman Filter

Mobed, P., Texas Tech University
Turton, R., West Virginia University
Bhattacharyya, D., West Virginia University
Rengaswamy, R., Texas Tech University

Efficient operation of the water gas shift reactor (WGSR) is very important for satisfying the CO2 capture target in integrated gasification combined cycle (IGCC) plants with pre-combustion CO2 capture.  The operation of the WGSRs should not only satisfy the desired H2/CO ratio at the inlet of the acid gas removal (AGR) plant, but also convert most of the COS content in the syngas.  For pre-combustion CO2 capture, a sulfur-tolerant catalyst is used in the WGSR for simultaneous hydrolysis of CO and COS.  In these WGSRs, a number faults can occur. For example, the residual fly ash in the syngas can deposit on the catalyst reducing the number of active sites for reaction and decreasing the porosity of the catalyst bed. As the WGS reactions are exothermic, temperatures that are higher than that allowed by the catalyst manufacturer can cause micro-structural changes to the catalyst. Our current work focuses on estimation of these faults using an Extended Kalman Filter (EKF) framework.

The first-principles dynamic WGSR model used in this work is a nonlinear differential algebraic equations (DAE) system that poses unique challenge in estimation. Some of the algebraic equations have a process noise term while others have to be satisfied exactly. Further, some of the algebraic states are also a part of the measurement system. A recently proposed EKF approach for state estimation in DAE systems is modified to solve the WGSR state estimation problem. The fault variables are considered as additional states and estimated jointly with the other states in the proposed approach.  The results show that a very accurate estimation of the faults in a WGSR can be obtained by using our proposed enhanced EKF. The presentation will include a discussion on the effect of number of measurements and their sampling rates on the accuracy of prediction of the time of occurrence of the faults.

See more of this Session: Process Monitoring and Fault Detection II

See more of this Group/Topical: Computing and Systems Technology Division