Research And Application Of Iterative Learning Model Predictive Control Algorithm

In modern industrial production, model predictive control originated in the real process control area has been widely applied.Model predictive control has become a recognized standards in dealing with complex constrained multivariable control problems.Its applications in many complex optimization of industrial process control achieved huge economic benefits.In the actual complex production process control, predictive control with fine control effect, robustness, etc., can effectively overcome the uncertainty,nonlinear,and contingency of the process,and easy.It also can handle constraints in the process control variables.But predictive control do not have good learning capacity, it is difficult to achieve error-free output of the controlled system to fully track the desired trajectory.While the iterative learning control is a powerful tool in disposing process with nonlinear, uncertainty, and high-precision trajectory control requirements.Considering advantages and disadvantages of both, a algorithm is studied. Model predictive control is integrated into the iterative learning control.The algorithm is applied to the reactor system which proved the algorithm can quickly and accurately track the desired trajectory, and it also has the real-time anti-interference ability.Add feed-forward control into iterative learning model predictive control algorithm, through simulation, we can see the introduction of feed-forward can reduce the number of iterations to speed up the tracking speed and improve efficiency. And the algorithm is applied to the boiler superheated steam temperature system, and compared with the effect of the PID algorithm to verify the superiority of it.Iterative learning model predictive control algorithm can not be directly applied in nonlinear operation, so T-S model is introduced.The fuzzy modeling techniques, predictive control and iterative learning control are combined.The design of T-S fuzzy model based on the iterative learning model predictive control algorithms used in repetitive motion of the robot shows that this algorithm can still achieve the desired trajectory of error-free in the nonlinear system.