Study On Modeling Of Reactor&Regenerator System Of FCCU

Catalytic cracking is one of the most important processes to light heavy oil in petroleumprocessing industry, occupying an extremely important position in gasoline and dieselproduction. Therefore, modelling of the FCC reactor®enerator system is necessary.Thispaper discussed the modelling methods of FCCU reactor®enerator system. The statisticalmodel, principle mechanism model and hybrid model were established. This work can bedivided into three parts:Firstly, this work established a simulation and optimization model based on neuralnetworks and use the genetic algorithm to do operation optimization of the model. This paperpresented four rules for the selection of neural network input variables. Besides，this paper setintervention rule when use genetic algorithm to search the optimal result. Therefor, operationparameters which contradict the reality can be excluded.Secondly, this work also established a mechanism model of a MIP FCCU riser. Based onthe theory of the FCC lumping theory, the reaction system was divided into ten lumps.Standard molar enthalpy of combustion was used to calculate the chemical reaction heat,which modified the previous assumption that reaction heat is a constant. This work leaded tohigher accuracy of the temperature forecasts. The model was used to text a catalytic crackingunit processing capacity of2.8Mt/year. The calculated values of reaction riser outlettemperature agreed well with the actual value, and the relative error was0.31%.Finally, this paper also proposed combining the mechanistic model and neural networkmodel to form a hybrid model. By using neural network model to correct the mechanismmodel, the hybrid model had a better forecasts accuracy and the average absolute error of theproduct distribution changed from5.61%to1.59%which proved that the accuracy of hybridmodel was better than the mechanistic model.