The real time product quality intelligent forecasting and analysis system

摘要

Catalytic cracking fractional colurnn is the most important production device for refining enterprises in China. Its main products are car gasoline and diesel fuel. The yield and quality of these two kinds of products decide directly the economie efficiency of enterprises. In order to increase the economic efficiency of enterprises, it is needed to better adjust and control the quality of car gasoline and diesel fuel. Because fluidized catalytic cracking unit (FCCU) is in closed state, it is impossible to observe actual production process manually. But if people cannot timely master product quality condition, it is impossible to adjust effectively the technological parameters in order to control product quality. But at present, it takes four hours to obtain quality level of products if using the method of manual sampling testing. If it is as this, production process cannot, based on the analyzed results, be timely adjusted. Therefore, developing the real-time product quality intellect forecasting and analysis system of fractional column and realizing forecasting and analysis on-line have important theoretic meaning and value in engineering application. This system can real-timely forecast product quality of fractional colurnn, and can also real-timely analyze the factors affecting the products. So, the adjustment oftechnological parameters is more targeted, and shortens adjustment time, and increases efficiency. It is no doubt that the economic efficiency will increase. The thesis, taking fractional colurnn of fluidized catalytic cracking unit (FCCU) as research target, with the aim of forecasting product quality level of fractional column, establishes quality forecasting model through the method of neural network, and speculates the critical technological parameters that are hard to measure or impossible to measure at all through the technological parameters that are easy to measure. The system first finishes interactive interface between control system and operator with the functions of dynamic display and real-time data acquisition through configuration software DCS (Distributed Control System), which can supervise, control, activate and manage the whole system. Then it will realize product quality forecasting of fractional colurnn through the method of combining utility function based on average level and neural network. Finally it will realize the analysis of factors affecting product quality through the method of combining fuzzy technology and neural network. The thesis, through system configuration and using neural network technology to forecast product quality of fractional colurnn and analyze the factors affecting product quality, combines fuzzy technology and neural network which play their respective advantages to finish the display and control of operation state of fractionation system and realize real-time forecasting and analysis. The online forecasting system of product quality of catalytic cracking fractional colurnn based on the method mentioned above is developed for many small and medium petrochemical enterprises. The aim is to transform the equipments under the present condition of small and medium petrochemical enterprises with no change in the hardware of the original DCS (Distributed Control System) of refining enterprises. Therefore, this system has many advantages such as small investment, short transformation time and easy realization, etc. Currently, this system has been tried on the fluidized catalytic cracking unit (FCCU) in Tianjin First Petrochemical Plant in China. The operating result shows that the value and laboratory value of dry point of car gasoline and solidifying point of diesel fuel forecasted real-timely in this model have better goodness of fit, satisfying the requirements of product quality index. The test result shows that the technical path and method using neural network technology to forecast product quality put forward in the thesis is feasible.ud______________________________________________________________________________ udMOTS-CLÉS DE L’AUTEUR : Catalytic cracking, Fractional column, Neural network