Delay Systems Intelligent Dynamic Matrix Control

As controlled system is very complex in industry process and precise mathematical model is very difficulty to build, the use of synthesis method in modem control theory based on strict (rigid) model have been greatly limited. Dynamic Matrix Control (DMC) is a new strategy of Predictive Control based on non-parameter model, and it is a synthesis method based on real step response of controlled system and suitable for steady asymptotically linear system. It has higher control precision without the necessary of precise mathematical model. This control method is very suitable of the time-delay system in this paper because of strong robustness to the controlled system.In this paper the elementary theory and algorithm structure of DMC are discussed and the principle of parameter choosing, stability and robustness are qualitatively analyzed. Because of the strong self-learning and self-tuning abilities of neural network, especially radial basis function network is a kind of neural network that has simple structure and high convergence speed, and it has the globe approximation performance, at the same time, it doesn't has local convergence. Based on the above merits of RBF network, using the neural network's approximation ability to build predictive model to improve the structure of DMC, a modified algorithm that is more suitable for real industry application in the DMC based on neural network model is proposed. In the end, this control method is been used in the temperature control of industry boiler, namely, simulation is done for typical time-delay system. The simulating results demonstrate that the presented control method has higher precision, rapid respondent velocity and better robustness in the case that the control parameters are designed properly. The modified algorithm method proposed has higher control effect comparing with the classical PID control and Smith predictive control.