Font Size: a A A

Study On Networked Cooperative Distributed Model Predictive Control

Posted on:2016-09-23Degree:DoctorType:Dissertation
Country:ChinaCandidate:L W ZhangFull Text:PDF
GTID:1108330503493766Subject:Control Science and Engineering
Abstract/Summary:PDF Full Text Request
In model predictive control technique, we usually design a single center to control a whole system. At each time instant, the controller optimizes a sequence of the control inputs based on the predicted outputs and control inputs. Only the first input in the sequence is applied to the process system and the procedure is repeated at the next time instant. With the development of modern science and communication technology, the MPC technique is applied to much more complex industrial process. Due to the requirement of on-line optimization, it can be an issue when applying the MPC technique to large-scale industrial process with fast sampling rate. At the same time, since the centralized MPC uses a single controller, the whole process will be down if a local system fails to connect the control center. The robustness and flexibility of the centralized MPC are usually not acceptable for real applications. These deficiencies have greatly retracted the development of centralized MPC, especially for the complex industrial systems. The control of complex industrial process is being transformed into the distributed control style. Compared with the centralized MPC, distributed MPC has advantages on safety, reliability, flexibility and fault tolerance. The whole system can keep on working if one of the sensors or actuators fails. The researches on the distributed MPC provide a basic theory for applying the MPC to wider areas.In recent years, distributed MPC technique has attracted wide attentions. Different kinds of algorithms are proposed and applied to large-scale systems, such as the transportation systems and chemical systems. However, with the repaid development of the economic and technology, the existing theories are far more enough to deal with the high efficient control requirements. There are still many open problems including how to reduce the conservatism of controller design, how to develop high efficient algorithms and how to further reduce the computation burden with less performance loss. In this thesis, we deal with several critical important issues of the distributed MPC, including the actuator saturation, model uncertainty, time delays and coordination schemes etc. The research contents are concluded as follows:Firstly, we give a solution to distributed MPC controller design for the LPV systems subject to actuator saturation. The global system is decomposed into several subsystems and independent distributed MPC controllers are designed in parallel. Set invariance condition for the polytopic uncertain system is given. Based on the invariant set, distributed MPC controllers are obtained by solving an LMIs optimization problem. The distributed MPC controllers reach a coronation in an iterative fashion. The proposed distributed MPC algorithm is tested via case studies.Secondly, a robust distributed MPC algorithm is proposed for the uncertain systems with time delays. In order to reduce the computation burden, the optimization problem of model predictive controller design is decomposed into solving several sub-controllers. An upper bound of the performance is obtained by solving an LMI optimization problem. For multiple state-delayed case, a robust distributed MPC algorithm is presented and the closed-loop stability is proved. Then, the results are extended to the multiple delayed states and control input delays case by augmenting the system state. Multi-step state feedback law is introduced to reduce the conservatism of distributed MPC controller design. Numerical examples are carried out to test the theory results.Thirdly, we discuss the quantized communication and packet dropout problems in distributed MPC controller design. It is assumed that each state feedback law is quantized and communicated via unreliable network. The packet loss is modelled by the Bernoulli process and log quantizer is considered for the state feedback laws. By proposing a packet dropout compensation strategy, the closed-loop control performance of the distributed MPC algorithm is improved. The control performance is tested via a numerical example.Finally, we design a distributed MPC algorithm with novel sequential update scheme for both nominal system and LPV system, respectively. It is known that centralized MPC algorithm requires optimizing all control inputs at each time instant. In our distributed MPC with sequential update scheme, the inputs are divided into several subsets and only one subset is optimized at each time interval. For the nominal system, the control inputs are obtained by solving a smaller dimensional Quadratic Programming(QP) problem. Furthermore, the multi-step scheme is extended for the LPV systems. The state feedback laws are obtained by solving an LMI optimization problem. The results are tested on a tunnel boring machine(TBM) cutterhead systems.
Keywords/Search Tags:Distributed model predictive control, Cooperative, actuator saturation, time-delayed systems, LPV systems
PDF Full Text Request
Related items