A decentralized control design approach to a class of large-scale systems

Koeln, Justin

Permalink

https://hdl.handle.net/2142/45287 Copy

Description

Title

A decentralized control design approach to a class of large-scale systems

Author(s)

Koeln, Justin

Issue Date

2013-08-22T16:34:48Z

Director of Research (if dissertation) or Advisor (if thesis)

Alleyne, Andrew G.

Department of Study

Mechanical Sci & Engineering

Discipline

Mechanical Engineering

Degree Granting Institution

University of Illinois at Urbana-Champaign

Degree Name

M.S.

Degree Level

Thesis

Keyword(s)

• Decentralized control
• Large-scale systems
• Block arrow structure
• Model predictive control
• Extremum seeking control
• Variable refrigerant flow

Abstract

Large-scale systems present a unique control challenge. The large number of states, actuators, and control objectives for these systems often restricts the ability to analyze and control the system as a whole. Typically, these large systems are decomposed into multiple smaller subsystems which can be analyzed and controlled separately using a decentralized control approach. However, if the interactions between subsystems significantly affect the dynamics of the system, a decentralized control approach may prove to be ineffective and even result in unstable behavior. This thesis develops a control strategy for a class of systems with a particular hierarchical structure known as a Block Arrow Structure (BAS). Many real world systems naturally exhibit this two-level hierarchical structure, where a common subsystem at the higher, global, level interacts with multiple subsystems at the lower, local, level. There is no direct interaction among the lower level subsystems. A standard decentralized control approach would control each subsystem separately, ignoring the interactions between the higher and lower level subsystems. However, the interaction between the two levels may significantly affect the system dynamics, rendering the decentralized control approach ineffective. The proposed control strategy, referred to as the BAS control strategy, retains the scalability of the decentralized control approach but is also able to directly consider the interactions between the higher and lower level subsystems. This allows the BAS control approach to perform significantly better than a decentralized approach. Model predictive control (MPC) is used to evaluate the performance of the BAS control strategy relative to both centralized and decentralized approaches for two different BAS systems. In addition to the BAS control approach, this thesis develops an extremum seeking control (ESC) strategy which is used to improve the overall efficiency of the BAS system. In addition to performance objectives such as tracking a desired value for a state of the system, many systems have an efficiency objective. This objective seeks to control the system in the most efficient way possible, while still meeting the performance objectives. Minimizing the total energy use of all the actuators in the system is a common example of such an efficiency objective. In this work, ESC is used to augment the BAS control strategy at the global level to further improve the efficiency of the overall system. The model-free nature of ESC makes this control strategy especially effective in the presence of unknown disturbances and system nonlinearity, which may not be captured by the models used for the MPC controllers of the BAS control strategy. A linear example system is used to demonstrate the concepts and ideas presented throughout this thesis. For this example system, the BAS control architecture with ESC is able to achieve a control performance very similar to that of the centralized control approach while retaining the scalability of the decentralized approach. The benefits of the BAS control approach are also demonstrated for a more realistic system: a variable refrigerant flow (VRF) air-conditioning and refrigeration system for a building. Through a gray-box modeling approach, it is shown that VRF systems naturally exhibit a BAS structure and, therefore, can benefit from a BAS control approach. VRF systems are becoming widely used to meet the air-conditioning and refrigeration needs of buildings because of their greater efficiency in removing heat versus the conventional forced air systems. For these systems, it is very important to meet both the performance objectives, such as maintaining a desired air temperature in a room, as well as the efficiency objective of minimizing the total energy consumed by the system. Through a series of simulation examples, the BAS control approach is found to be a very effective control strategy for meeting both of these objectives.

Graduation Semester

2013-08

Permalink

http://hdl.handle.net/2142/45287

Copyright and License Information

Copyright 2013 Justin Koeln

Owning Collections