Development of a Set of Models for Reactors of a Catalytic Reforming Unit Using Information of a Different Nature

Abstract

Annotation. Mathematical models of reforming reactors of a catalytic reforming unit of the LG-35-11 / 300–95 type of the Atyrau oil refinery have been built. Since this technological unit is characterized by a deficit of quantitative information and the vagueness of some of the available information, the ideas of the system analysis methodology were used in the work, which allows the systematic use of information of various nature in a complex. Mathematical models of reforming reactors are developed on the basis of the systematic use of statistical information based on experimental statistical methods and fuzzy information based on expert assessment methods. As a result, a system of hybrid-type models was built, i.e. the models are built on the basis of experimental-statistical and fuzzy information. In this case, the models describing the dependence of the volume of production, i.e. produced gasoline and technical hydrogen from the input, operating parameters are obtained in the form of statistical models in the form of regression equations of multiple regression. And fuzzy models that assess the quality indicators of gasoline are built in the form of fuzzy multiple regression equations with fuzzy parameters. The processes of obtaining, formalizing, processing and using fuzzy information in developing a model are based on the methods of expert assessments and theories of fuzzy sets. To determine the structures of the developed models, the idea of the method of sequential inclusion of regressors was used. And for parametric identification, i.e. to determine the values of the regression coefficients, a modified least squares method was applied. In order to identify fuzzy parameters, first on the basis of level sets α=0.5; 0.75; 1, the fuzzy regression equation is transformed into an equivalent system of conventional regression equations for various α. Then the regression coefficients of the obtained regression equations for each level are identified by the usual method of parametric identification. Then, on the basis of the corresponding formulas of the fuzzy set theories, all the coefficients of the level α are combined, which makes it possible to switch to computer modeling.