Model Predictive Control Design for f Nonlinear Four-Tank System

In recent years, MPC has become a prominent advanced control technique, especially in large industrial processes. However, for enormous complexity, non-convexity and computational reasons, MPC practice and applications have been restricted to linear plants. During the last decade, many formulations have been developed for MPC formulation of linear and nonlinear, stable and unstable plants but still there remain some unsolved issues which depend on plant specifications. Instability, unfeasibility, non-convexity and lack of robustness are examples of unsolved issues. In this thesis a control system has been designed for a highly nonlinear and non minimum phase Four-Tank system. Then constrained optimization is employed in the MPC formulation to repair violation on boundaries. It also leads the system to work with the best performance. Additionally, the influence of most effective tuning parameters in MPC strategy has been investigated. In particular main part of the thesis has focused on performance criteria based on good reference tracking in model predictive control domain. Regarding to investigate the performance of this algorithm and due to application of “nonlinear Four-Tank system” in control theory and industry, this system is considered as a plant to be examined under this method. The most attractive aspect of this system is; the time-varying movement of a right half plane transmission zeros across the imaginary axis. This system’s configuration makes the process difficult to control under the previous controllers. This problem appears to be one of the most important and practical designs of nonlinear system in process control. In this thesis, besides good performance, the algorithm enjoys from relative simplicity and faster response in compare with the algorithms developed in other previous works. The problems of complexity of algorithm, non-convexity of the optimization, especially when working with nonlinear plants are the most common problems in the control design criteria. Since linear model predictive control is used instead of nonlinear model predictive control; these problems are avoided to be appeared in this work. All the results in this study show fast performance in controlling the Four-Tank system. Both of the weighting matrices are considered so that a system is fast enough smooth control signals and they are tuned till the desired performance is achieved.. Low value of prediction horizon and weighting matrices are more preferable to reduce number of free variable and avoid complexity of analysis