(703d) Development of a Novel Ekf-GA Approach for Distributed Sensor Placement – Application to WGSR in an IGCC Plant

The efficiency of a process is highly affected by disturbances and faults, which in extreme cases may lead to unsafe conditions or to process shutdown and hence need to be detected early for preventive actions or maintenance planning. However, disturbances and faults may not be available for measurement directly and therefore need to be estimated using available measurements and the underlying model of the process. The number, type and location of measurements (termed as measurement model) will influence the accuracy of the estimate and hence careful consideration needs to be done for selection of measurements. In this work, we will describe an approach for synthesizing an optimal measurement model and its application to the water gas shift reactor (WGSR) as part of an integrated gasification combined cycle (IGCC) plant. The WGSR is one of the important units in an IGCC plant for satisfying power production and carbon capture goals. A first-principles model of the WGSR has been developed in our previous work where the reactor is described by a system of non-linear differential and algebraic equations (DAE).  The design of the measurement model is done by using the Genetic algorithm (GA) and extended Kalman filter (EKF).

An enhancement to the existing EKF that not only handles the uncertainties in the differential equations, but also considers the presence of uncertainties in algebraic equations while satisfying other state constraints will be described. The GA is used to search the large number of possible combinations of sensors while the EKF is used to estimate the values of the desired variables for each set of sensors that is considered by the GA. However, this approach is cumbersome due to the vast search space and computational burden for nonlinear state estimation. To address the latter issue, reduction in computational burden is studied by developing simplified WGSR models using scaling analysis. For reducing the computational burden further, in-situ adaptive tabulation (ISAT), a storage and retrieval technique, is used. The reduction in computational burden and the accuracy of the resulting measurement model are compared against the use of the original model.