Study On Approaches Of Coordinated Optimization Control For Distributed Systems

With the rapid development of new generation of information technology,the promotion of network technology,and the emergence of complex and diverse requirements for control systems,the structure of industrial control system tends to be more and more distributed.For example,industrial production processes such as steel making,power generation,and chemical industry have a distributed structure.This kind of system is composed of many heterogeneous subsystems and has the interactions of matter and energy as well as a lot of information interaction.Moreover,the development of modern industry puts higher and higher requirements on the optimization goals of the production process.In addition to the optimization control of a single subsystem,it also requires the global optimization of the entire system.Therefore,the research on cooperative optimization and predictive control of distributed systems has become a hot topic in current control area.To this end,the paper studies the cooperative optimization and predictive control methods for distributed systems under the networked environment,and discusses and analyzes the controller design of the optimization algorithm,the solving of optimization problem,and closed-loop performance of distributed systems.This paper mainly focuses on the problems of model unknown,model nonlinearity,disturbance,asynchronous communication environment and event-triggering control faced by distributed control systems,and develops the method of cooperative optimization and predictive control for distributed system under networked environment.The main objective is to improve the utilization of system calculation and communication resources under the premise of ensuring the stability of the closed-loop system and obtaining the desired dynamic characteristics,and also to provide a theoretical basis and method reference for the structural optimization of the actual distributed industrial networked system and the dynamic optimization control of the industrial process.The main work and contributions of this dissertation are as follows:1.For distributed control system with multiple linear heterogeneous subsystems,the distributed cooperative output control problem based on the network communication is studied under two scenarios: the model is known and the model is unknown.When the model information is known,a distributed collaborative control algorithm of state feedback is designed by using network transmission information.Then,with an introduction of the optimization performance index,we online optimize the gain matrix of the collaborative control algorithm by using the iteration method.When the model information is unknown,a distributed cooperative optimal control algorithm based on adaptive optimal learning is proposed.Based on the online collected data,either the system input data or tracking error data,an off-policy reinforcement learning algorithm,is developed,which can realize the online adaptive learning of the optimal control parameters for the collaborative control algorithm.The proposed learning control algorithm does not depend on the system matrix information of the system,but can also achieve distributed output synchronization and converge to the optimal control parameters.2.For distributed control system with constraints and nonlinear coupling terms,the problem of cooperative predictive control under asynchronous communication environment is studied.Since each subsystem has different dynamic characteristics,it leads to a wide range of time constants,which makes it difficult to get a unified information update and optimization.Based on the weak coupling conditions,we design an asynchronous distributed nonlinear model predictive control algorithm,which uses the asynchronous communication information to estimate the interconnected states and brings it into the design of constraints of the distributed optimal control formulations.Then,we analyze feasibility of the proposed algorithm and stability of the closed-loop system and provide necessary conditions for selecting parameters of those optimization problems.3.For linear distributed system with nonlinear coupling terms and bounded unknown disturbance,we investigate the problem of cooperative predictive control based on event-triggering communication.To make full use of the asynchronous communication information,we proposed a state estimation method in an asynchronous environment,so that the estimated states are incorporated into the prediction model at each step.Then,the distributed event-triggering mechanism is designed to determine the time instant of solving the MPC optimization problem,which theoretically eliminates the Zeno trigger phenomenon.Our method effectively improves the utilization of system computing and communication resources.Then,combined with the Lyapunov stability theory,the feasibility of the calculation method and the stability of the closed-loop system are analyzed,and the selection conditions of the design parameters are given.4.For distributed control systems with a more general form of nonlinear couplings,we propose an asynchronous distributed model predictive control algorithm based on periodic event-triggering mechanism.Based on the weak coupling conditions,an adaptive state estimation method based on latest communication information is designed to effectively utilize the asynchronous information,which also achieves decoupling between subsystems.In this way,with an adaptive update prediction model,we design a distributed model predictive control algorithm through periodic testing of the designed triggering conditions,which ensures that the error between the predicted state and the actual state of the system is bounded.The algorithm not only coordinates the asynchronous communication information,but also guarantees that the real state is within the bounded range of the predicted state.Then,by giving selection conditions of the design parameters,feasibility of the algorithm and stability of the closed-loop system are guaranteed.