The Middle Scale Research On Catalytic Combustion Of Waste Gas Containing Acrylonitrile With Flow Reversal Technology And Research Of Multivariable System Identification

Environmental pollution is one of the most important problems overthe world and it badly affect human life ,health and society sustaindevelopment .Volatile Organic Compounds (VOC) are one of the majorsources of air contaminations. Generally, it contains the compounds which cancause the cancer,firedamp which can lead to the Greenhouse Effect and thetoxic carbonic oxide(CO) gas, etc. So the let of VOC have drawninternational attention. The absorber treatment of acrylonitrile plant in the Daqing petrochemical company discharges an average of about 40000Nm/h(50t/h) waste gas. The concentration of propane, propylene and carbonmonoxide were very high in the acrylonitrile containing waste gas emittedfrom adsorption tower in acrylonitrile plant, which also contains smallconcentration of acrylonitrile and HCN, et al. ACN is one of the VOC,which pollutes the environment propylene, Propylene, propane like firedampcan lead to the Greenhouse Effect. In the elimination of VOCs, the reverse flow catalytic combustion technique is a unique solution. The reversed flow catalytic combustion reactor highly integrates the catalytic combustion and the recovery of the reaction heat, which can cut down the operation cost and save the investment.The lowest concentration of Reactant Autothermal Reforming in the reversed flow catalytic combustion is about decuple less than that in traditional catalytic combustion. Based on the superiority showed in the pilot scale investigation on acrylonitrile containing waste gas in the reverse flow reactor, researchers in the Daqing Petrochemical Company longed for the middle scale investigation on catalytic combustion of waste gas containing acrylonitrile in the reverse flow reactor.According to the requirement of the Daqing Petrochemical Company, we improved on the pilot scale reverse flow reactor for catalytic combustion of VOCS built in the laboratory. A scale reverse flow reactor and the monitoring system of it was designed and built. The former data-collections, control mode and alarm were rebuilt. In order to apply advanced process control algorithm and optimal operation to the reversed-flow catalytic combustion process, the quasi-steady model of temperature profile for a reverse flow reactor with catalytic combustion developed by RBF(Radial Basis Function) neural networks of the reactor considered is improved by LS-SVM(Least Squares Support Vector Machine) In the process of carrying out, the epigynous part of the monitoring system was mainly composed of industrial controlling computer, cards and the software written with MCGS configuration and the hypogynous part of it was mainly composed of sensors, such as thermocouples, flowmeters and so on. The system has been used in experimental environment, and the application of the system shows that the monitoring system is more reliable, real-time and intelligent. The quasi-steady model of temperature profile for a reverse flow reactor with catalytic combustion developed by RBF neural networks of the reactor considered is improved by LS-SVM.The train data used in the LS-SVM is much less than in the RBF neural networks. So the cost of time and money in the experience is decreased. The application of the equipment in the plant will become quicker.Besides, we carried through basic research on multivariable system identification methods in the paper, mainly introduced and studied the vital method of multivariable system identification: subspace state space system identification method.First, we studied deeply the theory and methods of multivariable system identification. Based on the study of basic idea and algorithms of subspace state space system identification method, we emphasized on the study of the subspace identification based on orthogonal projection and PCA (Principal Component Analysis). We checked the validity and veracity through the simulation experiments in the matlab. Second, it is impossible that acquiring the system model order exactly used basic idea and algorithms of subspace state space system identification method apply to the subspace identification based on orthogonal projection and PCA. The system order is decided by non-zero singular values of the observation matrix in the basic subspace state space system identification method. But we can not find the observation matrix needed in the subspace identification based on orthogonal projection and PCA, especially in the closed-loop plant, which will lead to a bad deflection, sometimes we can't identify the system model. The paper introduces the AIC criterion in order to acquiring the system model order, which can get the system order exactly, improve the identification effect and reduce the error.Last, we develop identification software based on the one of subspace state space identification methods: N4SID, and get a good identification result through some simulation samples. The software will help to the further development of the identification methods applying to the fieldwork of plant.