The Predictive Control On Complex Systems And Its Applications

The predictive control is a kind of algorithm developed from practice. For the complexity of control systems, the predictive control should be studied deeply. In this paper, based on the characteristics of complex systems, several kinds of novel predictive control algorithms are proposed. The main contents are as follows:In industry application, the performance indexes have some fuzzy characteristic, and the optimization of the goal function trends to intelligence and satisfaction. Using the fuzzy membership function to describe the fuzzy characteristic, and the Yager T Aggregator to aggregate fuzzy decision goal function, the predictive control algorithm based on fuzzy decision goal function is studied in this paper. The principles of the parameter defining are given. The relationship between the parameters of membership function with control performance is studied. In order to solve the oscillatory problem caused by the discrete control variables, an improved fuzzy center method is used to aggregate the control output.The select of aggregators is a key factor in fuzzy decision goal function. Through the analysis of the aggregator's characteristic, the way to select aggregators to aggregate fuzzy decision goal function is given. The results of using three kinds of aggregator with parameter - Dombi aggregator, Yager aggregator and Mean aggregator are compared. The reason of "decision transgression" is analyzed when using Yager aggregator, and the method to solve this problem is clarified. Compared with Yager and Mean aggregators, the result using Dombi aggregator has better performance.Because of the random of the membership function, which aggregator to be selected and the non-linearity of system, the fuzzy decision goal function is non-convex commonly. To make the control variable change continuously, overcome the system oscillation, and avoid the exponential increase of computational load due to variable increasing, the evolution programming is applied to optimize fuzzy decision goal function. Further more, this method can be applied to optimize any non-convex goal function.In industry application, the input variables are not equal to output ones for many complex control systems. For the coordinated control problem of the non-symmetrical system, the coordinated control algorithm based on fuzzy decision goal function is proposed. In this method, the membership is used to describe the fuzzy characteristic of the complex system, and the Dombi T aggregator is chose to compose the goal function with the membership. The result indicates that this algorithm can solve the coordinated control problem of non-symmetrical system with constrains.According to the feature of the cement kiln, a practical fuzzy-predictive coordinative control algorithm combined the fuzzy control and predictive control is proposed. The tendency of the error is predicted through the approximate model, and the output is calculated by the fuzzy control using the predictive error. Further more, the multi-logic controller is designed for coordinating the two output goals of the system.It is difficult to model the industry objects with delay time and big inertia In order to adapt the feature of the objects, the predictive-fuzzy control algorithm is proposed, in which the compensation is calculated using the predictive error, and is used to amend the output of the fuzzy controller. This method has been applied in decomposing furnace and raw system successfully.