Distributed Event-triggered Stochastic Model Predictive Control For System With Uncertainty Disturbance

With the rapid development of social economy,multi-agent system has become an indispensable part of modern industrial development.Distributed Model Predictive Control(DMPC)is widely used in the theoretical research of multi-agent system because of its good control performance,efficient ability to handle various constraints and high fault tolerance.With the gradual maturity of DMPC theory,this method has also been gradually applied to practical industrial environments.However,the uncertainty disturbance in the actual industrial environment have brought great difficulties to the theoretical application of DMPC.Therefore,the Distributed Stochastic Model Predictive Control(DSMPC)method,as a model predictive control method that can handle internal model uncertainty and external disturbance uncertainty in the system,has been widely introduced into industrial control schemes in recent years.In addition,most of the current DSMPC algorithms are based on time-triggered.For some systems with small external disturbances and internal model deviation,optimizing control at every moment will cause a large waste of system computing resources.With the continuous expansion of the scale of multi-agent system,the resource consumption of time-triggered mechanism on the system is increasing.Event-triggered is different from the traditional time-triggered method.It can be judged whether it is triggered by the information of the multi-agent system itself,which can reduce the calculation of the system and save system resources.In response to the above content,this thesis aims to study a distributed event-triggered stochastic model predictive control method that can handle uncertain disturbances.Based on the previous research work of scholars,the following work has been carried out for linear discrete systems:1.A distributed stochastic model predictive control method with multiple constraints is studied.For multiple linear subsystems with uncertain disturbances,output probability constraints,coupled output probability constraints,and state probability constraints between subsystems are designed to ensure the local and overall control requirements of the multiagent system.In addition,probabilistic constraints and stability conditions are proposed.Through methods such as probability theory,iterative method,and Lyapunov principle,it is proved that the system satisfies the requirements of iterative feasibility and closed-loop stability under this algorithm.Finally,simulation verifies that the system states can converge to the vicinity of the origin and that each variable satisfies the constraint requirements under this algorithm.2.A state based distributed event-triggered stochastic model predictive control method is studied.In order to save system computing resources,an event-triggered condition based on subsystem state is designed for multiple linear subsystems with probability constraints.The optimization problem can only be solved when a certain subsystem meets the condition.In addition,constraints such as prediction time domain,disturbance,and invariant set are proposed.Through invariant set theory,comparison principle,and Lyapunov principle,it is proved that the system meets the feasibility and stability requirements at each triggering moment.Finally,simulation results show that the system can achieve stability and effectively reduce the number of controller solutions under this event-triggered strategy.3.A cost based distributed event-triggered stochastic model predictive control method is studied.In order to reduce the trigger times of the controller when the system reaches stability,an event trigger condition based on the cost equation is designed.This method combines the state information of adjacent systems,making it more comprehensive to determine whether the controller needs to be triggered;And the triggering coefficient was designed to achieve the goal of controllable triggering frequency.In addition,through the iterative method and the input-state stability(IS S)theorem,it is proved that the system meets the requirements of iterative feasibility and closed-loop stability at any time.Finally,simulation verifies that the system has less triggering times than the state based triggering strategy under the premise of satisfying stability.