Design And Analysis Of Event-Driven Strategies For Distributed Model Predictive Control

Distributed model predictive control(DMPC)has the advantages of excellent control performance,strong multi-constraint and multi-objective processing ability,flexibility and high fault tolerance,and has been widely used in many fields such as smart power grids,urban traffic networks,and industrial control.In this kind of control system,the computing resources of nodes responsible for local optimization are often limited and the transmission of information between nodes is easily restricted by network resources.This makes the event-driven strategy that only executes the control action at a specific time become a research hotspot in distributed model predictive control.Although some research progress has been made,the low trigger strategy that takes into account the stability of the system and the feasibility of the algorithm remains to be studied.The core reason is that the Lyapunov function decreasing principle used to ensure stability and the inexact state prediction error to analyze the feasibility will inevitably bring frequent trigger problems.In response to this challenge,this dissertation studies the event-driven strategy of distributed model predictive control from the perspective of changing the stability guarantee principle and improving the accuracy of prediction error.Specifically,1.Aiming at the triggering strategy design under the conservative Lyapunov analysis method,an adaptive event-triggered DMPC strategy is proposed.The optimal control problem with decreased prediction horizon is defined and the triggering conditions is designed based on this,reducing computational complexity and triggering frequency,thereby reducing computational and communication loads.2.Aiming at the triggering strategy design under the imprecise knowledge of neighbor information,a compound event-triggered DMPC strategy is proposed,which solves the problem of frequent triggering under a single stable triggering condition.A stability triggering condition independent of the neighbor system estimation information is designed,which combines the stability condition based on the Lyapunov function to trigger in parallel,and reduces the event triggering frequency.3.Aiming at the triggering strategy design under the inaccurate prediction model,an adaptive event-triggered DMPC strategy based on disturbance prediction is proposed,which improves the accuracy of model prediction.The disturbance prediction scheme based on the central path and the adaptive trigger threshold scheme are designed to reduce the state prediction error,improve the trigger threshold,and significantly reduce the trigger frequency.4.Aiming at the triggering strategy design under dynamic interconnection of systems,a rolling self-triggered DMPC strategy is proposed,which simplifies optimal control problem design and reduces the conservatism of state prediction error estimation.A dual model optimal control problem is designed to simplify the feasibility analysis of the algorithm while optimizing control performance.A rolling self-triggering mechanism is designed to increase the trigger interval of self-triggering and reduce computational and communication loads.In summary,this dissertation systematically studies the event-driven strategy design and analysis of DMPC,innovatively proposes corresponding solutions,and promotes the further development of DMPC.