Development of Inferential Models for Fractionation Reformate Unit

Industrial facilities show an increasing need for continuous measurements and monitoring a large number of process variables and properties due to strict product quality requirements, environmental laws and advanced process control. On-line analyzers often suffer from long measurement delay which is not desirable for closed-loop implementation. Laboratory assays are irregular and not suitable for process control. Inevitable alternative are soft sensors and inferential control. Since production standards in refineries become more stringent, there is a need for continuous monitoring and control of benzene content in gasoline. Development of soft sensor models for the estimation of light reformate benzene content is carried out. Linear dynamical autoregressive model with external inputs (ARX), autoregressive moving average model with exogenous inputs (ARMAX) and Box-Jenkins (BJ) models are developed. For the problem of regression vector optimization usually performed by trial and error, Genetic Algorithm (GA) and Simulated Annealing (SA) methods have been applied. The results indicate that the GA and SA as global optimization methods are both suitable for the regressor order estimation of linear dynamical models with multiple inputs. This makes the development of soft sensors easier and systematic. Based on developed soft sensors, it is possible to apply advanced process control for better product quality.