New Framework Of Model Predictive Control Considering Margin Loss Function

In the process of conventional control,controllers only take the precision of the controlled variables tracking the expectations as the standard of control effect,never consider the situation that control consume margin.However,in the field of control,control effect can not be said absolutely good or absolutely bad,the controlled variables tracking the expectations accurately is a kind good of control effect in traditional meaning.But if we pursue the control effect on one-side and cause control variables swing or going into the limit of process margin,even cause boundary effect,will cause the potential safety hazard to the production facility.To solve the above problems,adding a margin loss function which indicates the situation of control variables consuming margin into the model predictive control,build a new framework of model predictive control considering margin loss function,in order to consider the control effect and the cost of control comprehensively in the dynamic process of control.Considering margin comprehensively in the dynamic short-cycle process of control is an improvement and thinking of conventional model predictive control.Therefore,from the perspective of philosophy,interpretate the formation,development and existing framework of model predictive control philosophically.And through the marxist dialectical materialism principle propose the correctness and improvement scheme of model predictive control considering margin loss function,proposing a new framework of model predictive control.In order to analyze margin saved model predictive control considering margin loss function deeply,define margin loss function,build the framework of model predictive control considering margin loss function,derive the process of optimal solution,configure coefficients and analyze the conditions of convex optimization.At the same time,through the steady optimal value and the steady state analyze the effects on saving the space of the operating variables and the effects on the controlled variables tracking the expectations.In the long period of the process of chemical industry,the upper constraints of control variables are getting smaller with the consumption of process margin.Therefore,analyze the situations of saving margin when the upper constraints change in model predictive control considering margin loss function.In additon,propose another mathematical description and solution of model predictive control considering margin loss function.Finally,In order to consider the problem of periodicity in control and margin optimization,base on multi-layer hierarchical structure in industry,propose a margin saved online optimization implementation on model predictive control,adding a optimized implementation layer between global optimization layer and dynamic optimization layer,giving the optimal expectations calculated by technologists in global optimization layer into optimized implementation layer,and proceed steady-state optimization.Design the objective function of steady-state optimization based on robust multivariable predictive control technology in optimized implementation layer,and consider not only the optimal expectations but also margin loss function.Finally,can get optimal expectations which save margin,and turin the expectations into model predictive control in dynamic optimization layer,and we can also get the effect on saving margin.