Research On Subspace Identification Algorithm And Predictive Control

As modern industrial processes have more complex features,such as multivariate,strong coupling,and nonlinear time-varying,the corresponding control requirements are constantly increasing.Most control methods are based on process mathematical models,Because of the complex nature of the process,it is difficult to obtain mathematical models through mechanism modeling.However,the emergence of subspace identification algorithms effectively solves the problem of difficult to obtain accurate mathematical models through mechanism modeling.The subspace identification algorithm can directly obtain the mathematical model of the state space form according to the collected input and output data,and the identified intermediate result subspace predictor equation can be used as the prediction model.Based on the basic algorithms and characteristics of subspace identification,combined with the needs of predictive control strategies,this paper studies the design and optimization of controllers.The main results of this paper are mainly the following three aspects:1.Unconstrained predictive control based on subspace identification and constrained predictive control based on subspace identification are proposed respectively for the existence of constraints.The state space model obtained by subspace identification is used as the system model.The intermediate result of the identification is the predictor model as the prediction model.The predictor model is also the focus of the combination of subspace identification and model prediction.When the amount of input and output data is small,the nuclear norm optimization technique can be used to obtain the optimized variables,and then used for subspace identification,which can improve the identification accuracy.In the case of unconstrained and constrained conditions,the optimal control quantities are solved using derivation and quadratic programming,respectively.2.In the case of a constraint on the system,if the parameter setting of the controller is too large,the computational complexity will increase.In order to reduce the variables encountered in the calculation of quadratic programming and reduce the computational complexity,the design of the parameterization of the Laguerre function is introduced.By parameterizing the input incremental signal,the computation time of the algorithm can be effectively reduced without sacrificing the control effect of the algorithm.Finally,through experimental simulation,the controller parameter setting gradually becomes larger during the experiment.According to the experimental results,it can be found that the parameterized design method with Laguerre function can reduce the computational complexity and reduce the calculation time.