Study Of Multivariable Process Control Method Based On Model Predictive Control

Most of the actual industrial production process is multivariable system, owing to the fact that there are coupling effects among different controlled variables, multivariable system is difficult to control. In order to improve the control quality of industrial process, the research of multivariable system control strategy is very important. There are two main control methods for multivariable system, one is the decoupling control method, the other is direct control method. Decoupling control method is complicated, and it is difficult to implement, so the direct control method is widely used in practice. Predictive control is a typical kind of direct control methods, it combines optimization, constraint and compensation as a whole, which is very suitable for multivariable control system. Therefore, the control methods of multivariable system based on predictive control algorithm are studied in this paper. The main contents of this paper are as follows:(1) A predictive control algorithm based on single neuron is studied. In order to improve the shortage of predict information processing in traditional predictive control algorithm, single neuron PID control is introduced into the predictive control algorithm, and a single neuron predictive control algorithm is studied to achieve compensation for the prediction error. The basic idea of the algorithm is that using the prediction model of dynamic matrix control to predict the future steps of the output system, the feedback correction, take the predicted value after correction as the feedback signal, it is compared with the set value, the resulting error is taken as the enter state variables of the single neuron PID controller. Finally, single neuron predictive control, single neuron PID control algorithm is respectively used to control a cold and hot water system, the simulation results show that the single neuron predictive control algorithm has advantages such as short adjusting time, good robustness and so on.(2) The orthogonal particle swarm optimization algorithm is used for tuning the parameters of dynamic matrix control. Dynamic matrix control has several parameters, and there are some associations among parameters, and different combinations of parameters will affect the control effect. It is cumbersome and time-consuming to adjust the parameters of dynamic matrix control by trial and error method, and traditional particle swarm optimization is easy to fall into local optimal solution. In order to overcome theses Shortcomings, the improved orthogonal particle swarm optimization algorithm is used in this paper for tuning dynamic matrix control parameters. Finally, taking the moisture and quantification system in papermaking process as a simulation object, this algorithm and standard particle swarm optimization algorithm are used for tuning the parameters of dynamic matrix control, the simulation results show that using this algorithm to tune dynamic matrix control parameters, the closed loop system has good robustness.(3) A predictive control algorithm based on extended state observer is studied. In order to solve the status information of the controlled object unpredictable and uncertain factors, which led to the control effect of predictive control algorithm turn worse, and even make the system unstable, the extended state observer and dynamic matrix control algorithm are combined, and a predictive control algorithm based on expansion state observer is studied in this paper. The basic idea of the algorithm is that the extended state observer is utilized to observe the parameter perturbation of the controlled plant and external disturbances, then the dynamic matrix control algorithm is adjusted based on the equivalent effects to obtain the control input. Finally, this algorithm and dynamic matrix control algorithm is respectively used to control a cold and hot water system, the simulation results show that the algorithm has advantages such as good robustness and good anti-jamming.(4) The prediction control algorithm based on disturbance observer is studied. In order to suppress external disturbances and internal disturbances caused by model mismatch and the coupling of variables, the disturbance observer and dynamic matrix control are combined to control system, improving system the closed-loop control accuracy and anti-jamming capability, accelerating system response speed. At last, this algorithm, dynamic matrix control algorithm and the conventional Smith-PID control algorithm is used to control the crude distillation column system, the simulation results show that the algorithm has good robustness.For multivariable system which is difficult to control, five algorithms are studied, i.e., the single neuron predictive control algorithm, the fuzzy PID predictive control algorithm, the predictive control algorithm based on orthogonal particle swarm optimization, the predictive control algorithm based on extended state observer, and the predictive control algorithm based on disturbance observer. The simulation results showed the effectivencess of these algorithms in this paper to multivariable system.