Industrial Robust PID Tuning And Multivariable Model Predictive Control Research

In this thesis, the parameter optimal tuning method for PID controller and robust PID controller parameter tuning method are studied thoroughly based on the problems in industry control field. In order to obtain the dynamic model of the process to be used by robust PID, closed-loop model identification algorithm in frequency domain is researched. Multivariable model predictive control application is studied at a practical industry processes.The main contributions and innovations are as follows:1. A new performance indexes that meet the necessity of industry application is presented, and the control move constraints of PID controller are considered in the performance index in the elementary PID control system. Particle Swarm Optimization (PSO) is employed to solve the program whose objective is based on the introduced performance index, and optimal PID controller parameters are achieved.2. Uncertainty exists in real processes because parameters of plants are varied, so it is difficult to get good control performance by conventional PID controllers. Robust PID controller parameters are tuned by PSO based on MiniMax rules, the worst operation point is found in certain bounded range that the parameters of the process may varies, PID parameters are tuned on the basis of the operation point, and good control performances and robustness can be attained on the rest operation points.3. A closed-loop model identification method based on nonlinear relay feedback is introduced in this dissertation, the process is approximated as a first order plus dead time (FOPDT) model, the generalized process model in model predictive control that contains PID controller and process in closed loop control system can be identified by the method as well. Limit cycles are produced in the process by the biased relay method, and then a FOPDT model can be acquired by the limit cycles.4. An advanced process control and optimization application to acetaldehyde distillation units is introduced in this paper. The acetaldehyde distillation unitsare composed of two distillation columns; one is used to remove the light component and one end-product column. Coupling problems in the columns are encountered, which cause the operation of the distillation units difficult. The control tasks of the two columns are considered together by a multi-variable model predictive controller to overcome the mentioned problems. The variant range of the columns temperature is reduced. Optimization strategies ensure that the operating points of distillation units move towards its optimal steady-operating points.