Linear Parameter Varying (LPV) Model Identification on Primary Reformer Section of Ammonia Plant for Operator Training Simulator

Abstract

System or process identification is the field of mathematical modeling of systems (processes) from test or experimental data. Most of the existing works on process modeling were based on LTI (Linier Time Invariant) model. This paper presents the Linear Parameter Varying (LPV) model identification for primary reforming process in ammonia plant to cover changes in process operating conditions, such as start-up, normal operation and shut-down. Recursive Least Square (RLS) based algorithm with parameter function p =sin(0.0001*k) is employed in the identification process. Data needed for identification are taken from DCS historian data of the primary reformer process. The identification result is simulated and validated with the measured data. The process models obtained by this technique show reasonably fit with the mean value of 90.42 %. The process models are implemented on a simulation engine of Operator Training Simulator (OTS) using Scilab/Scicos software tools. Simulation result shows that the indicators of process variable mostly accepted. The mean value of deviations between process variables of simulated models and historian data are 10%, 20%, and 5%, for operation rate 40%, 65%, and 100%.