Research On Model Predictive Control Technology And Application For Ethylene Distillation Column

Ethylene distillation column is the core unit in ethylene production process,and its production process is complex and process variables are strongly coupled.Stable operation and reducing the fluctuation of key variables become the main goals of ethylene distillation column process control.The above objectives cannot achieve satisfactory results only by relying on conventional control.Predictive control has become the preferred technology for ethylene distillation column to meet the above objectives.In recent years,predictive control technology has made great progress in chemical production,especially in ethylene production,which is mainly due to the extensive application of online analytical instruments and digital control systems,so that the key process variables such as product quality indicators can be measured online.However,this kind of on-line analysis instrument has the problem of time delay caused by long detection cycle,resulting in the product quality index can not be real-time control,and because of the ethylene distillation column process is complex,restricted by upstream and downstream interference,interval constraints,resulting in serious coupling of production process variables.The conventional control system mainly controls the production device from single input and single output,which is not easy to solve the above problems.In order to improve the process control quality of ethylene distillation column and ensure the stability of equipment and product quality,advanced control technology is needed.Therefore,it is of universal significance to choose predictive control,which has an ideal control effect on stabilizing the operation of ethylene distillation column and realizing the real-time control of product quality at the top of the column.In this paper,an ethylene distillation column is taken as the process object,the shortcomings of the original conventional control are analyzed,the predictive control scheme is obtained through research,and the model identification,controller simulation and system integration are carried out.The results show that the predictive control effect is good,and the above problems of time delay and strong coupling are solved.Firstly,based on the process flow of ethylene distillation column,the characteristics of conventional control strategies and the current control problems were analyzed.The conventional control strategy and its problems were analyzed,and the predictive control scheme of ethylene distillation column was proposed.The relationship between controlled variables and operating variables in each sub-scheme was analyzed.On this basis,the ethylene distillation column predictive control model identification work was carried out,the core algorithm of ethylene distillation column model identification was studied,the model identification method with smoothing factor was proposed,and the actual device was step tested according to the process operation specification,and the DMCplus-Model identification tool was used to process data and identify model which is based on the finite impulse response form.Finally,taking the sensitive plate temperature as an example,the Gauss Markov assumption was used to test the preliminary identification results and evaluate the identification results of each model.Posteriorly,starting from the core algorithm of predictive control,it analyzes the prediction technology of step response model and state space equation model,and the closed-loop optimization method.It analyzes three model compensation methods: parameter correction method,intermediate state variable correction method and neural network correction method.It is explained that the state space method is difficult to realize because the state equation of the actual device is difficult to obtain and the intermediate state variables are difficult to observe.At the same time,due to the change of working conditions,the training speed of neural network can not quickly adapt to the new working conditions.And,it is clear that the DMCplus predictive control algorithm used in this project is more suitable for ethylene distillation unit.Finally,the controller simulation experiment is carried out and the simulation results are analyzed by taking the sensitive plate temperature as an example.Then,the engineering implementation is carried out by using the system integration approach.The implementation includes system software and hardware configuration.Based on Yokogawa CS3000 series DCS,the design of communication interruption detection,loop switching,undisturbed logic interface design and DCS man-machine interaction interface design are completed.The data exchange between DMCplus host computer and DCS system is realized.Finally,the predictive control system of ethylene distillation column was used to obtain the calibration operation results,and the conventional control and predictive control of the device sensitive plate temperature and the column kettle liquid level variables were compared.The results show that the designed predictive control system can significantly reduce the fluctuation of key process variables,make the plant run more smoothly,and the tower top product quality is controlled in real time.Predictive control technology is a good means to solve the ethylene distillation column variable lag and strong coupling.Therefore,it reduces the labor intensity of operators,and has positive significance for improving the operation level of the entire ethylene plant,product quality and safety production.