The Research And Software Design Of Hybrid Model Predictive Control On Multivariable System

The problem for controling the complex multivariable system is ome of the mostdifficult problems among the area of control owing to the characteristics of the complexmulti-variable system which takes on strong property of non-linear and has strongcoupling between each variable, which is hard to control by using traditional controlmethod. The predictive cotnrol of the multivariable system is made by using the method ofhydrid model predictive control in the paper which doesn’t have to deal with the complexrelationship of mechanism between variables but just analyze the input and output data ofthe system. Then discriminate the BP steady-state model and ARMAX dynamic modelrespectively according to the steady-state and dynamic data of the system, combine thesetwo models by adopting the variable gain predictive control method to form the hybridmodel which can represent properties of the system. Then deduce the multi-step predictedoutput based on the hydrid model, rectify predicted outputs according to the error betweenpredicted outputs and pre-set outputs, and employ the sequence quadratic programmingmethod for solving the optimization problem of the objective function. Specific researchworks are as follows:First of all, study the identification of the hydrid model in the multivariable system,which is mainly focused on the identification methods for the multivariable dynamicARMAX model due to the existing unmeasurable noise terms. Based on a great deal ofliterature reading, a comparative accurate identification effect can be achieved byrecursive least square. Employ BP algorithm to identify the parameters for the stableneural network model. To gain the BP-ARMAX hydrid model, use the gain of thesteady-state model to adjust the dynamic model parameter which is normalized.Then, based on the BP-ARMAX hydrid model, deduce and design the predictedoutputs of multivariable system model. Rectify predicted outputs according to the errorbetween predicted outputs, the sequential quadratic programming algorithm is used tosolve the objective function, then make the obtained optimal control gain work on thecurrent input for real-time rolling optimization to achieve the goal of stable control. Finally, design hydrid model predicted control software by combining the theory ofhydrid model predicted control of multivariable system and the practical situation ofDistributed Control System in industry scene. And ultimately realize simulate applicationof model predictive control software by means of employing advanced C#language toprogramme and rational designing of category, function and interface.