Please use this identifier to cite or link to this item:
Type: Theses
Title: Water distribution system optimization using metamodels
Author: Broad, Darren Ross
Issue Date: 2014
School/Discipline: School of Civil, Environmental and Mining Engineering
Abstract: Evolutionary Algorithms (EAs) have been shown to apply well to optimizing the design and operations of water distribution systems (WDS). Recent research in the field has focussed on improving existing EAs and developing new ones so as to obtain better solutions (closer to the global optimum) and/or find solutions more efficiently. The primary aim of this research, however, has been to broaden the scope of optimization to include a number of the many factors that planning engineers need to consider when designing or planning the operations of WDS. Those factors considered here are (1) water quality criteria, (2) real-world, complex systems, and (3) the incorporation of data uncertainty. Incorporating each of these factors independently increases computational run-time of EA-based optimization of an algorithm that is already computationally intensive compared to other (inferior) algorithms that have been used in WDS optimization. Water quality models tend to run slower than hydraulic models due to the shorter timestep that is required to ensure sufficient accuracy, and the need for extended period simulations thereby increasing the simulation duration. Real-world models run slower due to their size. Data uncertainty is typically accounted for through the use of Monte Carlo simulations, that add several orders of magnitude to the computational requirements of optimization. Considering each of these factors together compounds the computational requirements to a point where it is impossible to optimize WDS using EAs in a reasonable amount of time. In this research metamodels have been used in place of simulation models within an EA to reduce this computational burden. A metamodel is a model of a model that runs much faster than the said model, but is still a high-fidelity approximation of it. The particular type of metamodel used in this research is an Artificial Neural Network (ANN) due to its theoretical capabilities and demonstrated effectiveness in water resources applications. The use of metamodels to act as surrogates for complex simulation models is not a trivial task. Therefore, guidelines have been developed on how best to incorporate them into the WDS optimization process. The overall metamodel-empowered, EA-based optimization algorithm developed in this research was applied to several case studies. Two small case studies, both variations of the New York Tunnels problem were studied for proof-of-concept purposes. They demonstrated that near globally-optimal solutions could still be found using the metamodel-based approach, i.e. there was minimal compromise in the effectiveness of the EA-based approach. Two larger, real-world problems were also studied: Wallan (operations planning) and Pacific City (system augmentation). These last two case studies were key to demonstrating the power of using metamodels in that they enabled a computational speed-up of up to 1375 times (137,500%) compared to a non-metamodel approach. This speed-up includes factoring in the computational overheads of using metamodels, i.e. time to generate calibration data and calibrate the metamodels.
Advisor: Dandy, Graeme Clyde
Maier, Holger R.
Dissertation Note: Thesis (Ph.D.) (Research by Publication) -- University of Adelaide, School of Civil, Environmental and Mining Engineering, 2014.
Keywords: water distribution systems
genetic algorithms
neural networks
computional intelligence
Provenance: This electronic version is made publicly available by the University of Adelaide in accordance with its open access policy for student theses. Copyright in this thesis remains with the author. This thesis may incorporate third party material which has been used by the author pursuant to Fair Dealing exceptions. If you are the owner of any included third party copyright material you wish to be removed from this electronic version, please complete the take down form located at:
Appears in Collections:Research Theses

Files in This Item:
File Description SizeFormat 
01front.pdf704.28 kBAdobe PDFView/Open
02whole.pdf2.23 MBAdobe PDFView/Open
PermissionsLibrary staff access only689.91 kBAdobe PDFView/Open
RestrictedLibrary staff access only2.42 MBAdobe PDFView/Open

Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.