Original scientific paper

MODELS FOR CONTIONUOS ESTIMATION OF BENZENE IN REFORMATE

Željka Ujević Andrijić orcid.org/0000-0003-2290-9664 ; University of Zagreb, Faculty of Chemical Engineering and Technology,
Romano Karlović ; University of Zagreb, Faculty of Chemical Engineering and Technology,
Nenad Bolf ; University of Zagreb, Faculty of Chemical Engineering and Technology,
Ivana Šarlija ; INA Industrija nafte d.d., Sektor Rafinerija nafte Rijeka

Abstract

Due to environmental regulations and production requirement the benzene content in fuels need to be limited. Therefore, it is necessary to continuously monitor the benzene content in light and heavy reformate. As the process analyzers that measure the benzene content in reformate, are often out of service, models of soft sensor are developed for the continuous estimation of benzene content. Soft sensors are developed using linear identification methods and global optimization methods. The development of Finite Impulse Response (FIR)) model and Output Error (OE) model are presented. To overcome the problem of selecting the best model order for multiple input models, by trial and error, genetic algorithms (GA) was used which makes the development of the soft sensors more systematic. Developed models show a satisfactory match with analyzer data on a validation data set. Models are implemented on the fractionation plant for the estimation of benzene content in light reformate.

Keywords

soft sensor; linear dynamic model; identification

Hrčak ID:

106829

URI

https://hrcak.srce.hr/106829

Publication date:

28.6.2013.

Article data in other languages: croatian

Visits: 1.725 *