Performance Analysis And Research Of Networked Distributed Model Predictive Control

As the production process is gradually extended to large-scale and complex,extensive application of distributed systems,field bus and other technologies,and establishment of information networks,systems in the fields of chemical and energy production consist of multiple local subsystems.In addition to optimizing control of individual subsystem,it is also necessary to pursue global optimization of the entire system.As a kind of regression time domain control that can deal with system constraints in the optimization control,model predictive control can effectively solve the optimization control problem of distributed systems.Therefore,in the modes of the network information,it is of great significance to study the distributed model predictive control and optimization of complex industrial processes.The main research contents of this article are as follows:This thesis summarizes the research significance and research status of distributed model predictive control,and its basic principles and basic problems are introduced in detail.And the influence on the control system is analyzed.This makes the research more purposeful,and provides a theoretical basis for the following research and analysis.In the network environment,centralized model predictive control may be transformed into the distributed model predictive control method,which may cause a variety of network communication problems because of the excessive amount of communication data.It replaces a single overall target with multiple targets of multiple controllers.However,due to the existence of correlations among various subsystems,it is adopted Nash optimization strategy when solving multi-objective optimization,and each subsystem through the communication network to handle each other's interaction information,thus improve the whole system's control performance.Simulation results show that distributed model predictive control based on Nash optimization can effectively solve the multivariable strong coupling systems.For industrial systems,there are subsystems with different dynamic processes and multiple random factors,relying only on conventional distributed model predictive control can't effectively capture their own dynamic behavior or even lead to ill-conditioning.In order to solve this problem,a multi-timescale distributed model predictive control method is proposed.That is,for each subsystem's fast and slow characteristics,corresponding control strategies are adopted to achieve the local control objectives,and the optimization control of the whole system is realized through estimation or communication to improve the control effect.Applied to the chemical process that contains two continuous stirred tank reactors,simulation studies show that this method has better control performance.Time delay is the inherent feature of network control environment.For cascade distributed systems with input coupling,under the condition that there exists timevarying delay in the communications between subsystems.A cooperative distributed model predictive control algorithm based on neighbourhood optimization is proposed on the foundation of the distributed model predictive control algorithm based on Nash optimality.The algorithm has been improved by selecting a new performance index.The detailed derivations of control law are given.The analytical form of constraint quadratic programming is given.By using Lyapunov stability theory,derive the sufficient conditions for ensuring the asymptotic stability of the whole closed-loop system.Finally,through the example simulation,the proposed method is compared with other kinds of predictive control strategies,and the obtained results can obtain better overall optimization performance.