Distributed Model Predictive Control For Changing-targets Tracking With Non-global Communication Restriction

With the development of Industry 4.0,in control system such as modern energy,manufacturing,heavy industry,large-scale distributed control systems need to have the ability of automatically switching the outputs as production requirements change.Cooperative distributed predictive control ensures the feasibility of changed target tracking by introducing a tracking target optimizer to each subsystem controller.But in the actual production environment,The internal structure and coupling relationship of the distributed control system are complex.Due to the location of the subsystem,security constraints,subsystem failures and other issues,it is often impossible for subsystems to achieve global communication.In view of this problem,this paper studies the changed target tracking control problem of large-scale complex coupled distributed control systems without the global information of the system.Consider whether the large-scale distributed system has weak subsystem coupling,the system design flexibility requirements,and computational efficiency requirements,This paper proposes two distributed model predictive control for tracking under non-global communication restriction.The details are as follows:(1)Consider that in distributed predictive control the sub-controller uses the cooperative communication to design its own tracking control law,and cooperation between the sub-controller and some weakly coupled subsystems has little impact on the system performance,it increases the complexity of the communication network.Thus this paper proposes a strong coupling neighbor-based optimization distributed model predictive control for tracking algorithm.The algorithm proposed an appropriate strong-coupling decision-making criterion.At the beginning of the system model establishment,strongweak coupling subsystems are determined according to this criterion.The state prediction of the system is based on the strong-coupling association and the weak-coupling neighobrs are neglected with the help of robust positively invariant set and feedback control.Given that deviations brought by ignoring weak-coupling neighbors are fully considered by introducing robust control,so that the tracking performance of the system is maintained near the optimal tracking solution.The strong-weak coupling criterion fully consider the control performance while pursuing communication simplification.It combines the minimum interference upper limit caused by weak coupling ignorance and the system network connectivity which characterize the communication topology complexity.The weighted optimal solution balances control performance and communication flexibility.At the same time,coordination scope of the states coupled system controller is supplemented.By introducing the related matrix in the predicted horizon,subsystems get their correlation range which is equivalent to the global communication.Because the strength and weakness coupling decision strategy is adjustable,the control system design is flexible.At the same time,each sub-controller uses a cooperative mode to optimize its performance,thus the control efficiency is high.(2)Considering in actual production,when the subsystems of the complex coupled large-scale distributed system do not contain obvious weak coupling relationship,forcibly discarding the coupling relationship between some subsystems seriously affects the actual benefit of the control problem.For this kind of system,this paper proposes a non-cooperative distributed model predictive control for tracking algorithm based on dual decomposition.The algorithm uses the idea of double decomposition to transform the coupling of dynamic variables and steady-state tracking values between subsystems into a set of equivalent constraints.In this way,the global performance corresponding to the current subsystem is decoupled from other subsystems,and cooperative communication is no longer needed when computing subsystem optimization problems.However,the Lagrangian multipliers of each subsystem in the algorithm are coupled,and each subsystem relies on iterative communication to optimize the Lagrangian multiplier.Therefore,the overall algorithm satisfies the communication between current subsystem and their upstream neighbors.Because the algorithm has the iterative efficiency problem of dual ascending iterative solution,this paper also proposes the early termination algorithm.The early termination algorithm gives the early termination assumptions and the robust control is used to solve the inconsistency introduced by the sub-controller early termination.Under the premise of not affecting the communication range,tracking is guaranteed,which alleviates the problem of low computational efficiency of the algorithm.In comparison,the non-cooperative distributed model predictive control for tracking algorithm has no requirements on the coupling form and coupling strength requirements,thus it is more common for different control scenarios.But the system design flexibility and computational efficiency are relatively poor.