Data-driven approaches for residential water end-use classification and sustainable urban water management

This dissertation utilizes data analysis and machine learning techniques to analyze residential water consumption at an end-use level through non-intrusive load monitoring, focusing on urban water sustainability and supporting the implementation of smart water meters and machine learning techniques to improve water management. The three objectives rely on smart water meter data to disaggregate, classify, and analyze water consumption and extend the availability of labeled data in the residential water literature. Objective 1, classification assessment based on different data resolution, aims to understand how different levels of temporal data resolution can impact the accuracy of end-use classification and provide insight into the efficient resolution needed for accurate water consumption behavior analysis. Objective 2 builds on the existing labeled dataset collected through Objective 1 to create a larger synthetic dataset and compare different classification models and identify the most efficient model for water end-use classification based on accuracy and computation time. This objective compares the performance of various machine learning algorithms and identifies the most accurate and computationally efficient model for water end-use classification. Finally, Objective 3 investigates the potential insights that smart water systems can provide once end use-level disaggregation is achieved, with a particular focus on stagnation time as a new metric for analyzing residential water consumption. In this objective, high-resolution water-use data are leveraged to quantify stagnation time (i.e., the duration an appliance remains unused) across different fixtures, offering insights into usage consistency, behavioral flexibility, and the early detection of anomalies, such as leaks. Beyond conservation, stagnation time also has implications for premise plumbing water quality, since extended stagnation periods can contribute to microbial regrowth and chemical degradation. By integrating stagnation time analysis with traditional flow-based monitoring, this objective supports demand-side water management strategies and conservation interventions, and can be broadly applied as smart meter deployments expand and more labeled data become available. The overall findings contribute to a deeper understanding of how smart data collection can enhance sustainable water management, supporting more efficient resource allocation, proactive system maintenance, and informed policy development for urban water systems.

Name thesis:degree_name

Ph.D.

Level thesis:degree_level

Dissertation

Discipline thesis:degree_discipline

Civil Engineering

Grantor

University of Illinois Urbana-Champaign

Year dc:date

2025

Author dc:creator

• Heydari, Zahra

Contributors dc:contributor

• Stillwell, Ashlynn S
• Guest, Jeremy
• Tessum, Christopher
• Cominola, Andrea

• water sustainability
• smart water metering
• machine learning
• water demand management

Statement dc:rights

• Copyright 2025 Zahra Heydari

Language dc:language

en, eng

Handle dc:identifier

https://hdl.handle.net/2142/129451

OAI record oai:identifier

https://www.ideals.illinois.edu/oai-pmh?verb=GetRecord&metadataPrefix=etdms&identifier=oai%3Awww.ideals.illinois.edu%3A2142%2F129451