Dynamic Simulation And Parameter Estimation For Industrial FCC Reactor And Regenerator System Using GPROMS Platform

With expanding economic scale the demand for energy from all trades is increasing.Most of the industries like the manufacturing industry,the construction industry,the textile industry and the transportation industry need energetic support to develop vigorously.Therefore,the energy industry has played an important role in economic development.Although various new energy technologies start to appear,the traditional petroleum industry still accounts for half of the energy industry and is the most important source of energy supply in the world.In order to optimize the product yields and quality,China’s oil refining industry is committed to building an intelligent refinery,hoping to improve the economical benefit by using automation technology,for example,production automation and device operation systematization.In order to establish an automated optimization platform,the equipment operating variables need to be optimal controlled.However,the safety demands and complexity of the refinery process determine that the optimal control can not be verified directly on an actual plant.Therefore,an accurate model must be applied to simulate the actual plant operating status and to perform the dynamic simulation.The catalytic cracking reaction and regeneration system is the most important secondary processing unit in the refining process.In this paper,the modeling and simulation of this unit are carried out.First of all,this thesis establish the mechanism model of the actual industrial catalytic cracking reaction regeneration system.By analyzing the operating status and structure of the actual industrial plant,the mechanism models of regenerator and reactor are respectively built.The high-efficiency regenerator is considered as plug flow reactor model connected with a CSTR(continuous stirred tank)model.In addition,the hydrogen combustion reaction is considered in the reaction network.The MIP-CGP reactor of the industrial plant is also considered as a cascade model.It is to be considered that a complex model is not conducive to the design of optimal control algorithms and too many reactions require a large number of experiments to measure the chemical reaction rate coefficients in the reaction network.In this thesis,we choose a five-lumped reaction network and establish a two-stage series plug flow reactor model.The reaction between products is regarded as a quasi-steady-state process,and the temperature change is regarded as a slow time-varying process.Secondly,this thesis identifies the unknown parameters in the established mechanism model.For different units and catalyst types,the reaction rate constants in the reaction network are also different.Therefore,the actual plant data of the plant studied in this paper needs to be collected for parameter identification.Since the mechanism model is composed of a series of partial differential equations,the optimization problem of identifying unknown parameters is a non-linear non-convex optimization problem.In this thesis,an improved LM algorithm based on stochastic damping factor is proposed on the basis of traditional LM algorithm.By introducing a random factor in the LM algorithm,the LM algorithm can jump out of the current local optimal solution.In addition,aiming at the identification problem with more parameters,a pre-identification process with dimension-reduction model is proposed,and then the stochastic damping factor based LM algorithm is used to re-identify the optimization problem.Finally,the parameters of regenerator and reactor were obtained respectively,and the steady-state model is verified and simulated with the measured data.In the end,the model of FCC reactor and regenerator is set up on the process simulation platform gPROMS.According to the requirements of the actual operating variables of the refinery,the main model inputs are changed and the simulation is carried out.The trend of the dynamic curve is analyzed according to the actual reaction situation.By analyzing the changing trend of the outputs the model is validated,which can be used as a simulation platform for optimal control.