Study On Model Predictive Control Of Complex Industrial Process

Supported by the National Fundamental Research Program of China (973), "Real-time intelligent control theory and algorithm research for complex industrial processes" (No.2002CB312201), model predictive control methods for three classes of complex systems, namely, uncertain time-delay systems with input constraints, parameter-unknown nonlinear systems and parameter-unknonwn multivariable nonlinear systems are studied in this dissertation. Stability analysis of the proposed predictive control methods is given, while the validity of the proposed control methods is verified via simulation experiments. And also, taken forced-circulation evaporation system as a typical example of a class of complex industrial processes which has multivariable, strong couplings between loops, nonlinearities, multiple models adaptive decoupling predictive control approach is applied to forced-circulation evaporation system, while the simulation experiments are developed. The main contents are outlined as follows:(1) A delay-dependent robust model predictive control strategy is designed for time-delay systems with polytopic-type uncertainties and input constraints. By appropriately choosing a quadratic function, the original min-max optimization problem that is difficult to solve, is transformed into an approachable convex optimization problem based on linear matrix inequalities. For those systems mentioned above, a robust one-step model predictive control scheme is developed. The strategy is to introduce a finite horizon cost function that includes multi-terminal weighting terms. The proposed robust one-step model predictive control scheme allows the first moveu(k|k) to be separated from the rest of the control moves which are governed by astate feedback law. A linear matrix inequality approach is applied to both delay-independent and delay-dependent robust one-step MPC controller synthesis. The feasibility of the proposed robust MPC algorithms and the robust stability of the closed-loop system are also proved. The effectiveness of the proposed methods is demonstrated by the simulations results.(2) Combining adaptive predictive control algorithm with multiple models method, a multiple models based adaptive predictive control technique is presented for a class of parameter-unknown nonlinear systems. The control approach is composed of a linear robust adaptive generalized predictive controller, a neural-network based nonlinear adaptive generalized predictive controller and a switching mechanism. The linear robust adaptive generalized predictive controller ensures the boundedness of the input and output signal in the closed-loop system and the neural network based nonlinear adaptive generalized predictive controller improves the performance of the system. Designing advisable switching scheme between the two controllers, it is demonstrated that the stability and the improved system performance can be achieved simultaneously. Stability and convergence analysis are also given. The simulation is given to illustrate the effectiveness of the proposed algorithm.(3) A nonlinear decoupling controller that is composed of feedback controller, decoupling compensators and neural network nonlinear compensations is designed for a class of parameter-unknown multivariable nonlinear systems. On the basis of the nonlinear decoupling controller structure, a multiple models based adaptive decoupling predictive control technique, which consists of a linear robust adaptive decoupling predictive controller, a neural network based nonlinear adaptive decoupling predictive controller and a suitable switching scheme between the above two controllers, is presented. The stability and convergence analysis of the closed-loop switching system are given. The simulation results show that the satisfactory decoupling effect is achieved by using the proposed method.(4) A dynamic model of the forced-circulation evaporation process is established, meanwhile its integrated complexities, such uncertainty, multivariable, strong couplings between loops and high nonlinearities are analysed. The multiple models based adaptive decoupling predictive control approach is applied to forced-circulation evaporation process. The dynamic model of forced-circulation evaporation process is chosed as the controlled plant. The simulation experiments for the forced-circulation evaporation process are illustrated to verify the effectiveness and practicality of the proposed control scheme.