Research On Steady-state Optimization Control Based On A Class Of Complicated Industry Process

With the development of chemical industry, optimal control is more and more important to enterprise. And with the mutual development of economy, the competition is fierce between enterprises, each enterprise has not only satisfied the safe operation, but also the high yield, high quality and low costs are required. Technology of steady-state optimization can find out or keep the equipment parameters of handicraft variable which make the industry process to work in the best condition according to some target of control system. In the past, technology of steady-state optimization is based on accurate system models, but the complicated industry process is nonlinear, time-varying, etc. It is difficulty to build the system of mathematical model. With the development of intelligent algorithm, complicated process modeling and optimal control problem was solved.This paper was first introduced the influence of industrial process development steady-state optimizing. From the enterprise resource conservation, from reducing energy consumption, increase the efficiency of the importance of industrial process control. According to the complicated industrial process is difficulty to establish an accurate model, and I had researched work on the RBF neural network, then improved RBF neural network build complex industry process model.In order to overcome the shortcoming that MMAS was easy to be in local optimum and its convergence was slow, MMAS was improved. The speed of looking for optimizing parameters was increased and overall was improved by immune algorithm. Effectiveness of the algorithm was verified by simulation on the route of 150 cities.In the thesis we took reactive distillation for synthesizing ethyl acetate as the object. We have determined the controllable parameters and the target of optimization specifically by carrying out large amount of scene data collection and analyzing via the acquaintance of the productive technology. We used improved RBF and improved MMAS to model and optimize, and have verified the validity of the algorithm by experiments.