Fast Predictive Control With Disturbance Rejection Ability And It's Application In Hot Steel Rolling Process

Temperature is a typical industrial control plant, which have the big inertia, time delay and time-variable parameter characteristics. Because of time delay, disturbances cannot be detected in time and control affects takes a long time to affect the control objects, which have remarkably detrimental influence on system stability. It is a problem in control theory.Predictive control is based on predictive model, online receding horizon optimization and feedback control. These characteristics can be a better compensation of time delay. Aiming at improving the stability and the quick-response of systems, a new algorithm was proposed. First, a Toeplitz prediction equation is introduced to avoid the iterative calculation resulting from solving Diophantine equation in the traditional GPC algorithm. Second, an output increment penalty term is added into objective function for inhibiting output fluctuations and the overshoot. Third, the predictive control law is softened to restrain input fluctuations and improve control performance. The simulation results show that the algorithm is fast, strong anti-disturbance capacity and smooth output.Metallurgical industry is very important in China's economic development. Finishing rolling is very important process for the many kinds of steel products. The final rolling temperature has a direct impact on the quality of strip. Hot rolling is a complex process, but only has two effective measuring points. The traditional control methods use PID algorithm to adjust the water spray between racks, which has a serious lag in finishing temperature control and cannot achieve satisfactory control performance. Because the finishing rolling temperature is a complex control object, this paper integrates massive field data and technology background to build an off-line dynamic forecasting platform, and divides the whole strip into several observation points to achieve real-time forecasting. It can analysis the reason of temperature difference between the strip head and tail, and finds out the key factor in temperature changes through a lot of simulation.According to the simulation results of the off line predictive platform, the water spray quantity was used as the dependent variable. The online nonlinear model was identified by LV-SVM methods and linearization the model if it's available. Base on the actual field situation, the paper designs a fast predictive control algorithm with disturbance inhibition. The actual application demonstrates the effectiveness of the algorithm.