Files in this item



application/pdfLI-THESIS-2018.pdf (2MB)
(no description provided)PDF


Title:Decentralized model predictive control of urban drainage systems
Author(s):Li, Yichuan
Advisor(s):Voulgaris, Petros; Stipanovic, Dusan
Department / Program:Mechanical Sci & Engineering
Discipline:Mechanical Engineering
Degree Granting Institution:University of Illinois at Urbana-Champaign
Subject(s):Decentralized Model Predictive Control. Drainage systems
Abstract:This thesis applies a modeling method proposed by previous literature to a specific example and develops control techniques based on this model. The inherent nonlinear behaviors of the drainage systems were accommodated by introducing binary variables and linear inequalities to merge different modes of operation into a single expression. The objective function is constituted by 3 cost functions considering several priorities. Except the normal objective of minimizing overflows, we present two methods of reducing operation costs by harvesting rain power and energy from rainfall collecting locales. Pressure forebay regulates water with large potential energy and generates electricity as the water is directed through hydraulic pumps. Surface aqueduct collects water with high kinetic energy and pushes water through spiral case to generate electricity. Locations with these devices installed induce lower operation cost and thus have higher priorities to be utilized. Once we formulate the water management problem as an optimization problem with specified constraints, we can apply Model Predictive Control (MPC) to compensate for modeling errors and prediction inaccuracies. As we regularly update system states and disturbances information, we achieve our goal of applying real time control to drainage systems. As the system size grows, the system is partitioned into several subsections and each one of them forms a subsystem which makes local decisions based on partial information. The performance of different partition schemes was compared against centralized MPC and open loop controllers. It was shown that even decentralized controllers may suffer performance loss, the computation time was significantly reduced compared with centralized controllers. In rain scenarios with large intensity, the performance loss of decentralized controllers is insignificant compared with the advantage gained by computation time reduced.
Issue Date:2018-07-13
Rights Information:Copyright 2018 Yichuan Li
Date Available in IDEALS:2018-09-27
Date Deposited:2018-08

This item appears in the following Collection(s)

Item Statistics