Control of nonlinear industrial systems : approximate model based approaches

Abstract

Nowadays, industrial systems in manufacturing, energy, chemical and others are becoming more complex due to strong nonlinearities, multivariable coupling inter- actions, actuator saturations, internal/external disturbances, quick load variations and uncertainties. These salient features are bringing more challenges on the opera- tion and optimization of such industrial systems. Furthermore, intense competition among industrial operators increases the demand of highly effcient and dynamically- managed operating units. Modern control strategy provides the possibility to optimize the operation of industrial plants, especially under the constraints imposed by physical restrictions and safety limits. In fact, advanced control approaches are generally dependent on the exact nonlinear plant model. Unfortunately, the comprehensive first-principles models of practical industrial systems are hard to obtain due to the complexity of physical, chemical or other hybrid behaviors. Even if the nonlinear dynamic model is accessible, the nonlinear optimizing controllers are generally very difficult, if not impossible, to obtain and/or to be implemented in practice. In other words, optimal control and operation of these complex industrial plants is still a challenge. This thesis is to investigate and develop approaches to address this problem. In partic- ular, several novel approximate model-based control strategies are developed with guaranteed stability, optimality, feasibility, robustness, and tracking property of the closed loop control systems. Firstly, a number of state feedback stabilizing predictive control approaches are proposed based on fuzzy dynamic models and piecewise quadratic Lyapunov func- tions. The asymptotic stability of these closed-loop fuzzy predictive control systems can be established for industrial plants with input/state constraints, via the feasibility of convex optimization problems subject to a set of linear matrix inequalities. However, for practical industrial systems, output feedback control is more desir- able since not all plant states are available for control implementation. Based on a fuzzy model and piecewise affine model, dynamic and static output feedback robust H1 control approaches are developed for constrained nonlinear processes with distur- bances. It is shown that the output feedback controllers can be obtained by solving a semi-definite programming problem subject to some linear matrix inequalities, and the resulting closed loop control systems are asymptotically stable. In addition to stabilization, the output tracking and regulation of industrial sys- tems is of more interest and importance in practice and therefore is further explored in this thesis. A novel output regulation approach is developed based on piecewise discrete time linear models and error feedback scheme, and then applied to chaos synchronization. Furthermore, an efficient offset-free output feedback predictive con- trol system is developed for nonlinear processes based on their approximate fuzzy models. The zero offset output tracking property and input-to-state stability of the closed loop control system are guaranteed. Finally, this thesis presents a case study on fuzzy model predictive control of solid oxide fuel cell. A data-driven fuzzy identification method is applied to the dynamic modeling of an integrated solid oxide fuel cell stack and capacitor system. An input- to-state stable fuzzy predictive tracking controller is then developed based on the identified fuzzy model. It is shown that both the rapid power load following and safe fuel cell operation requirements can be achieved for the resulting closed-loop control system.