Design And Synthesis Of Distributed Reconfigurable Predictive Control Systems

Distributed systems are widely existed in various fields of modern industry,including petrochemical,manufacturing,transportation,aerospace,etc.It has the advantages of flexible system structure,lower computational load,simple installation and easy maintenance,supporting for information sharing and remote communication,and has been paid extensive attention by people both in academy and industry.As a key method to solve the optimization control problem of large-scale complex systems,the advanced distributed model predictive control has the ability to effectively deal with the multi-variable and multi-constraint optimization control problems,and thus has become the focus of research in the field of distributed control systems.In the production process of actual large-scale complex systems,the physical structure of distributed systems may change due to external environment and production conditions,such as changes in user requirements,the adjustment of production tasks,variations in the types of raw materials,and improvements in production processes,etc.For these reasons,a class of distributed reconfigurable predictive control systems emerges.Considering the control tasks and design requirements of distributed systems in different application scenarios,meanwhile,to meet various requirements of structure changes and realize a reliable distributed system function under the reconfigured architecture,it is of great significance to study how to design an effective control structure and an optimal distributed predictive controller for global systems in the case of dynamic structure changes.In consideration of both control cost and control performance,the minimum input design with guaranteed the structural controllability of global system is an important basis for the control structure design of distributed predictive control systems.However,affected by the dynamic changes of system architectures,the distributed system in a reconfigured structure may no longer meet the requirement of structural controllability or having the lowest control cost.This makes it difficult to design the control structure of distributed reconfigurable systems.At the same time,the dynamic reconstruction of the physical and communication topology of distributed systems often leads to the decline of system control performance,the infeasibility or instability of reconstruction control.These bring new challenges to the control of distributed reconfigurable systems based on the advanced model predictive control algorithms.All around these issues,this dissertation studies the design and synthesis of distributed reconfigurable predictive control systems.The main research work and contributions are as follows:(1)A minimum input design method for distributed systems with ensured global structural controllability is proposed.Considering the minimum input design problem of a(massive)complex distributed system with a huge number of nodes and associations.Since the traditional minimum input selection methods that implemented by integrated structural analysis based on the global system model maybe inefficient,a minimum input design approach suitable for distributed system structure is presented.Based on the structural information of local subsystems and that of their associated subsystems,a new graphical algorithm to determine the global non-top linked strongly connected component and minimum input is proposed to achieve the structural controllability of global system.Meanwhile,by analyzing the time complexity of the proposed algorithm and comparing it with the fastest algorithm designed based on a monolithic system model,we prove that when the distributed system structure satisfies a weak constraint,the presented algorithm has a lower complexity and it thus can be used to increase the efficiency of the minimum input control structure design of complex distributed systems.(2)The minimum input allocation approach of distributed reconfigurable systems is studied.The problem of optimization and redesign to the control structure of global systems with a reconfigurable distributed architecture is considered.For any reconfigured distributed systems with structural controllability,a graphical algorithm for removing the maximum redundant inputs from the existing control structure while keeping the structural controllability of the reconfigured system is proposed.Meanwhile,for any reconfigured distributed systems that are not structurally controllable,a graphical algorithm to determine the minimum additional inputs based on the existent control structure while achieving the structural controllability of reconfigured system is also developed.Compared with the traditional approaches,the presented polynomial time algorithm solves the minimum input allocation problem of distributed reconfigurable systems under a unified framework,which realizes the optimal control structure design for the global system under a renewed distributed architecture.(3)A reconfigurable distributed predictive control method based on the minimum input allocation is proposed.Based on the minimum input selection of the distributed reconfigurable system,a reconfigured distributed predictive control method for the local subsystems directly and indirectly effected by the structure changes is proposed.By designing a non-cooperative reconfigurable distributed model predictive controller,the effective control of reconfigured systems is achieved with using less computing and communication load.Additionally,in order to equip the reconfigured system with a quick response ability to different reconfiguration demands,the proposed reconfigurable controller design technique is combined with the powerful algorithm of alternating direction method of multipliers,and the high-efficiency iterative formulas for computing the optimal reconfigured control laws are obtained.Eventually,based on the minimum input allocation of the reconfigured system,a quick and effective control performance is achieved by the proposed reconfigurable controller design method.(4)A reconfigurable neighbor-based distributed model predictive control method with guaranteed control feasibility is presented.Aiming at the problem that the reconfigured controller design maybe initially infeasible at any arbitrary structure reconfiguration moment or in any arbitrary reconstructed system architecture,a smooth and feasible on-line transition process between the initial and the reconfigured system is established,and a dynamic reconstruction process design method with ensured feasibility is proposed.Meanwhile,for any auxiliary subsystem of directly effected subsystems in the renewed system architecture,the reconfigured distributed predictive controller design method based on the neighborhood collaborative optimization is developed.According to the minimum input allocation of the distributed reconfigurable system,the extended neighborhood optimization sub-model of auxiliary subsystems is constructed.Then by applying the neighbor-based distributed predictive control strategy within neighborhood regions,a benefit balance mechanism between associated subsystems is formed,which accelerates the cooperative convergence rate of distributed subsystems and improves the dynamic control performance of global reconfigured systems.