Research On Temperature Control Methods Of Cstr

Petrochemical industry is the main industry of the national economic development. CSTR is the major equipment of chemical reaction in petrochemical industry, so the study on its automatic control is significant. As the key variables of the chemical reaction in CSTR, the temperature in CSTR affects the amount and the quality of products, even the security of the production process. So it is very important to make accurate and valid control on the reaction temperature in CSTR. However, the automatic temperature control of CSTR during the whole process is still hard to solve, because the controlling objects have serious non-linearity, time-varying, uncertainty and large delay.Predictive control has low requirement of model accuracy, preferable stability and robustness while using the non-minimize descriptive model, and excellent dynamic features. It also introduces rolling optimization instead of global one-time optimization, which is useful to ease the uncertainty caused by model mismatch, aberration and disturb. So it has important theoretical meaning and practical value to use predictive control in the temperature control of CSTR.Based on the massive reference about CSTR, this paper briefly introduces the related concepts about the production flow in CSTR, and deeply analyzes the production flow and the monitoring and controlling of the important parameters in CSTR, then completes the realization of program design in lower computer and the monitoring system in upper computer. It also established the communication between monitoring software WinCC and MATLAB, which makes it possible to realize the complicated controlling algorithms.Besides, focusing on the complexity of CSTR temperature controll objects, this paper studies the use of neural network predictive control theories in the temperature control in CSTR by analyzing the thermal features of CSTR. And it also proposes improved optimization algorithm based on LM-QuasiNewton. The improved optimization algorithm will not only avoid the low speed of convergence of Gradient Descent Algorithm, but also overcome the problems of LM Algorithm, such as the low speed of convergence and the poor stability. Then, the simulation of the temperature model in CSTR proves that the improved control system is satisfactory.