Limits of Control Performance for Networked Control Systems with Random Communication Delays

  • Author / Creator
    Yan, Guoyang
  • Networked control systems (NCSs) and distributed networked control systems (DNCSs) increasingly appear in the modern process industry due to continuous expansion of system scales, physical setups and functionalities. Control loops in a NCS are closed through information exchange between the spatially distributed controller and system components over a shared network, while local information in a DNCS is transmitted between different subsystems through a communication network to compensate for plant-wide interaction. However, the inevitable and time-varying network-induced communication delays degrade the system control performance and lead to a non-stationary behavior of the closed-loop system, which pose great challenges in the design of automatic control systems over network. On the other hand, control performance assessment is an asset-management technology aiming at optimal control performance and cost effectiveness. The key to control performance assessment is first to find the limit of control performance and then to estimate this benchmark control performance from routine operating data. This thesis extends the first step of centralized control performance assessment techniques to distributed networked control and networked control cases with random communication delays.

    Input and output communication delays between different subsystems are posed as the controller and observer structure constraints in DNCSs. In order to handle random communication delays, the limits of control performance in terms of variance for DNCSs is proposed as a bounded performance region with respect to the range of communication delays. Then, the same idea is extend to characterize the limits of linear quadratic Gaussian (LQG) control performance for DNCSs with the upper and lower LQG tradeoff curves.

    Controller-to-actuator and sensor-to-controller communication delays are both considered as random values or first order Markov chains in NCSs. A practical linear time-varying (LTV) minimum variance benchmark is proposed for NCSs by using order of the interactor matrix (OIM) and relative degree of the interactor matrix (RIM). It is shown that the obtained benchmark terms can be estimated from routine operating data. Further, an explicit solution to time-varying model predictive control (MPC) is derived for NCSs, based on which the limits of control performance for networked model predictive control systems is proposed as the time-varying MPC performance tradeoff curve. The applicability and effectiveness of the proposed approaches are illustrated via their applications to different numerical and chemical process examples.

  • Subjects / Keywords
  • Graduation date
    Spring 2021
  • Type of Item
  • Degree
    Doctor of Philosophy
  • DOI
  • License
    This thesis is made available by the University of Alberta Libraries with permission of the copyright owner solely for non-commercial purposes. This thesis, or any portion thereof, may not otherwise be copied or reproduced without the written consent of the copyright owner, except to the extent permitted by Canadian copyright law.