Model-Order Reduction Technique for Temperature Prediction and Sensor Placement in Cylindrical Steam Reformer for HT–PEMFC

Abstract A model-order reduction technique employing proper orthogonal decomposition (POD) with a robust point interpolation method (RPIM) is proposed and validated herein; a proposed method has been used for temperature prediction and sensor placement for a cylindrical steam reformer. The uncertainty associated with a set of obtained observations typically leads to an infeasible prediction for designing and operating systems. To address this problem, a novel interpolation method is introduced for reducing the uncertainty via optimal sensor placement. The objective function of the proposed method is the norm of the linear combination of the POD basis function extracted from a set of error-free observations and a mask matrix determined based on the sensor placement. The error bound of the proposed POD–RPIM method is analyzed and compared with that of well-known conventional interpolation methods. It is demonstrated that the POD–RPIM is a promising method that can address the shortcomings of the conventional methods, with regard to the uncertainty involved.