Structure Analysis And Controller Design For Double-Layered Model Predictive Control Systems

The double-layered model predictive controller is usually used for the complex plant-wide system optimization control.The steady state optimization in the upper layer gives the optimal setting values to the lower layer,but the dynamic controller may not be able to precisely track these setting values within a finite control horizon because the real-time system with close loops is subjected to actual physical constraints such as the production conditions and the change of the raw material.This paper mainly studies the multi-variable constrained control problem in the double-layered model predictive controller.Considering new requirements of the practical industrial process and the stability requirements of complex industrial systems,two systematic solutions are proposed.First,this paper introduces the structure of the double-layered model predictive controller.And it is the control feasibility problem arising in such hierarchical control structure that seizes our attention,which is mostly caused by the real-time system states beyond the constraints set ordered by the SSTC layer.This paper designs an innovative control strategy,including the detailed design of each part of the whole controller and introduces the design of stabilization domain,consisting of multiple convex hulls.The strategy in this paper designs the stabilization domain off line and dynamically choose stabilization ranges online as the state constraints according to the real-time system states,as well as the control horizon of the DC layer.To solve the feasibility and stability problem of the double-layered model predictive control algorithm,this paper also designes the inter-layer interaction mechanism for the double-layered model predictive controller.The upper level outputs the steady target values of the optimized variables,and transfers these targets with the system constraint and the priority index table to the lower layer;when the actual state of the lower layer does not satisfy the current system constraints,we dynamically adjust constraints and the priority index table,and then feedback the updated information to the upper level through the information interaction mechanism.Based on these feedback information,the upper layer of the double-layered model predictive controller will re-compute proper steady state targets for tracking.Such an interactive mechanism can solve the problem caused by the inter-layered information mismatch,and enable the dynamic adjustment of the system constraints and control objects when the initial control states of the system do not satisfy the system constraints.Therefore,the control resources can be fully utilized to complete the tracking of the steady state target.e propose an evaluation algorithm,by which we can analyze the initial controllable domain of this control strategy with different control horizon.Finally,this paper analyzes the stability of the above algorithm and proves that by using the algorithm in this pape,the closed-loop system stability and recursive feasibility can be guaranteed over the control process.This paper checks the effectiveness of the algorithm with stable and unstable research objects.