Complex Thermal Systems Research And Development Of Generalized Predictive Control And Application Software Packages

Predictive control is regarded as the most promising optimization control algorithm. For complex control problems, model-based predictive control may give better control performance than the common PID control. Many big companies have embedded it into control system software and applied it successfully in some large-scale process control systems. However, predictive control was scarcely applied in many medium-sized and small control systems. The generalized predictive control (GPC) is one of the most efficiency algorithms of predictive control. Parameters of GPC are usually obtained by simulations or real time tests before the commissioning stage. But these parameters usually can not guarantee achieving good performance. For GPC controller, because of lacking in-depth study for parameters tuning strategy, the proper parameters of GPC are difficult to choose, which severely limits the GPC algorithm used in the production. Complex thermal system is the typical complex industry process which has large dead-time, large inertia and time-varying characters. The applications of classic PID controller in the complex process are not successful. As a result of the receding horizon and feedback correction strategy, GPC has more advantages than other control strategy. Low demand of model, better control performance, strong robustness and effective in handling constraints are all it's strongpoint. Combining parameters tuning strategy studies, GPC algorithms will greatly enhance control performance and application efficiency. Therefore, the study in GPC parameter tuning method of complex thermal system is of great theoretical significance and value of application.In the process control, if the accurate model is known, we can design a perfect model based controller. However, in practical application, it is very difficult to obtain accurate model process, even impossible. System identification has been an active field of automatic control for a few decades and it is closely linked to other areas of engineering including advanced control strategy, optimization and signal processing. A lot of model identification algorithms have been proposed by early researchers. Since step response identification is simple, easy to control, etc., it is used widely in industrial process control. In this paper the author considers the multivariable system identification based on step response by time-domain methods.This dissertation is consisted by five chapters. The main work is summarized as follows:(i)A survey on the development and status of the predictive control is discussed. Meanwhile, this paper reviews the application of generalized predictive control in thermal process, and presents the main task of this thesis.(ii)A new identification method in the time domain from decentralized step-test is proposed for TITO processes with significant interactions. In terms of parameter identification, the coupled closed-loop TITO system has been decoupled into four individual single open-loop systems with the same input signal. Simulation examples are given to show both effectiveness and practicality of the identification method for a wide range of multivariable processes. Furthermore, the method is robust in the presence of high measurement noise. The usefulness of the identified method in multivariable process modeling and controller design has been demonstrated.(iii)This paper reviews the formulation of GPC and develops an equivalent transfer function form of the control law. A close-loop analysis, used to determine the role of the various design and tuning parameters, also reveals that GPC includes many well-known control algorithms as special cases. Three strategies for assign proper default values to the design parameters during the commissioning stage are devised. Each of these strategies allows the user to vary the close-loop speed of response over a very wide range by adjusting a single parameters. Simulation examples confirm the theoretical analysis and demonstrate the ability of GPC to yield consistent closed-loop behavior for large process variations.(iv) Based on the C + + Builder, integrating the above model identification methods and GPC parameter tuning method above, GPC application package are developed for small and medium-sized devices.