Research And Application Of Intelligent Predictive Control Strategy

Model predictive control is based on advanced control technology which was produced in the late 70s of 20th century. It can overcome the modeling error of controlled object and external disturbance uncertainties, and efficiently make up some shortings that modern control theory can not avoid for complex objects. As a result, once produced, predictive control has become a hot topic in the theory control and industrial control, and has been widely applied to industrial process control.In recent years, nonlinear control system is gaining more and more attention. This is mainly because most of the actual systems are essentially nonlinear systems. Modern science and technology require that control system can process more stringent control targets and can deal with the nonlinear characteristics of the controlled system more accurately.Although the predictive control has been successfully applied to industrial process control, it is still mostly based on linear model. In fact, most real production process control systems are nonlinear control systems. Because of the complexity of nonlinear system, it is difficult to search for a general and unified predictive method for nonlinear systems.Since 1990s of the 20 century, the research results of intelligent control have been emerging. Some intelligent algorithms have been proposed and applied to practical process, which provide new methods and ideas for the control of complex systems.The dissertation focuses on three basic principles of predictive control:the prediction model, the feedback correction, rolling optimization. Nonlinear predictive control method is further researched and discussed. And its research has been applied to the control of Potential of Hydrogen(PH) value, and given the corresponding results.The following aspects are studied by this dissertation.The Back Propagation(BP) neural network model is studied. As for its slow convergence and being easy to fall into local minimum point and so on, this dissertation presents a hybrid particle swarm algorithm and BP neural network. The test verifies that it is more suitable for nonlinear prediction model predictive control.For the cumulative error problem of recursive multi-step neural network prediction, we propose a correction which is based on neural network compensation method to improve the quality of feedback correction.For the strategy problem of nonlinear optimization, the dissertation puts forward the particle swarm algorithm combinig with simulated annealing method. The method can not only solve the nonlinear optimization problem but also overcome the shortcomings of standard particle swarm algorithm. The algorithm has achieved good control effect. Hybrid intelligent predictive control algorithm is applied to the control of PH value, providing a new and more effective control method for the control of PH value.