Study On Software Instrument And Modeling Of Nonlinear Multivariable System On The Crude Oil Atmospheric Distillable Tower

Aiming at the requirement on the atmospheric tower in KaLamay petrochemical factory. Two problems are mainly discussed: establishing software instruments of flash point and viscosity using neural networks; modeling on side line temperature multivariable nonlinear system of the atmospheric tower based on generic algorithm and neural networks.Based on the comprehension of process mechanism, process flow and principle of crude oil distillation, various factors which affect flash point and viscosity are analyzed. Sequentially, software instruments of flash point about No.2&3 side line and viscosity about No.3 side line of the atmospheric tower are established. Firstly the generic algorithm is offline used to search the initial parameters of RBF software instruments model, then the grads descend method is online used to adjust software instruments model, and then the centers C, widths B, weights W are ascertained. This method overcomes the problem that parameters of RBF are hard to design and generic algorithm's disadvantage in real time aspect, and makes full use of the speediness of grads descend. From the training and testing curves, software instruments obtain high accuracy and satisfy the requirement of project.KaLamay petrochemical factory sees high level transformer oil and lubricating oil as characteristic production. Transformer oil and lubricating oil which can satisfy the spaceflight requirement is mostly produced here, but the side line temperature multivariable nonlinear system of the atmospheric tower coupled badly and is often operated by open-looped control , thus the yield of special type oil is very little. In order to increase the yield of special type oil, the modeling and control accuracy of side line temperature system is eager to be heightened. Data which is needed to model and control is collected in open-looped condition of the atmospheric tower, the multivariable nonlinear dynamic model about the relationship among the temperature and the flux of No.2&3 side line is established. Through the comparison of the dynamic model and the authoritative mechanism model, the superiority of the dynamic model is indicated. Based on the analysis of DRNN neural network and the design method of adaptive control, the PID self-adjusting decoupling control method is brought forward, which establishs the foundation for latter control study of the side line temperature multivariable nonlinear system of the atmospheric tower.