Research On Predictive Control Algorithm Based On Two-time Scale Systems

In the field of modern industry,more and more advanced control algorithms are applied to the control of industrial processes.One of the most widely studied advanced control algorithms is the Model Predictive Control algorithm(MPC).The MPC algorithm has many advantages in dealing with multivariable complex control system.In general,the difference between the dynamic response speeds of each controlled variable in the system is often neglected,when the MPC algorithm is used to control the multivariable system in the industrial process,and the controlled variables are observed and controlled at the same time scale.This kind of control method is reasonable and effective for the system in which the dynamic response speed of each controlled variable is not too big.However,for the system which the dynamic response speed of each controlled variable is very different,the above control method can no longer meet the demand of each variable for the control variable at the same time.The system of the above-mentioned which has large difference between the dynamic responses speed of the controlled variable is called the multi-time scale system.Among them,the system in which the dynamic response speed of each variable is expressed as fast variable and slow variable is called two-time scale system.Therefore,it is a very important research direction that how to meet the demand of each variable for the control variable at the same time in two-time scale system when design the multivariable predictive controllers,so that the MPC can solve the practical problems and applied to complex industrial control more better.In this paper,the corresponding algorithm is proposed for the MPC algorithm of two-time scale systems.Firstly,using the singular perturbation method to decompose the two-time scale system model,so that we can obtain the fast and slow subsystem models.Then,according to the different dynamic response characteristics of the fast and slow subsystem model,the sampling time of the system output is divided into the sampling coincidence time and the sampling nocoincidence time.And then,two kinds of solutions are put forward for the calculation of the system control volume at the time of sampling coincidence and sampling nocoincidence.One way is to calculate the control variable on the basis of the fast subsystem model at the sampling nocoincidence time.At sampling coincidence time,the control variables of the two types of systems are calculated on the basis of fast and slow subsystem models respectively,and then the value obtained by weighting the two kinds of control variables is taken as the control variable of the system of such time;In view of the shortcomings of the above algorithm that the MPC controller can not calculate the system control variable using the whole outputs values because of the lack of the output value information of the slow subsystem model at sampling nocoincidence time,so another way is to considering the coupling between the control and the output of fast and slow subsystems to dealing with the.In order to achieve the control of the entire system,the output information of the slow subsystem is sampled by an estimation method during the sampling nocoincidence time,so that the predictive value information of the two subsystem models can be integrated into the same predictive control optimization problem regardless of whether the sampling coincidence time or the sampling nocoincidence time.