Development of an intelligent model to categorise residential
water end use events
Author
Nguyen, Khoi, Zhang, Hong, Stewart, Rodney Anthony
Published
2013
Journal Title
Journal of Hydro-Environment Research
DOI
https://doi.org/10.1016/j.jher.2013.02.004
Copyright Statement
© 2013 Elsevier. This is the author-manuscript version of this paper. Reproduced in accordance
with the copyright policy of the publisher. Please refer to the journal's website for access to the
definitive, published version.
Downloaded from
http://hdl.handle.net/10072/53485
Griffith Research Online
https://research-repository.griffith.edu.au
1
Citation: Nguyen, K.A. Zhang, H. and Stewart, R.A. (2013) Development of an
intelligent model to categorise residential water end use events, Journal
of Hydro-environment Research, Available online 15 March 2013, ISSN 1570-
6443, http://dx.doi.org/10.1016/j.jher.2013.02.004.
DEVELOPMENT OF AN INTELLIGENT MODEL TO CATEGORISE
RESIDENTIAL WATER END USE EVENTS
Abstract
The aim of this study was to disaggregate water flow data collected from high resolution
smart water meters into different water end use categories. The data was obtained from a
sample of 252 residential dwellings located within South East Queensland (SEQ), Australia.
An integrated approach was used, combining high resolution water meters, remote data
transfer loggers, household water appliance audits and a self-reported household water use
diary. Disaggregating water flow traces into a registry of end use events (e.g. shower, clothes
washer, etc.) is predominately a complex pattern matching problem, which requires a
comparison between presented patterns and those contained with a large registry of
categorised end use events. Water flow data collected directly from water meters includes
both single (e.g. shower event occurring alone) and combined events (i.e. an event which
comprises of several overlapped single events). To identify these former mentioned single
events, a hybrid combination of the Hidden Markov Model (HMM) and the Dynamic Time
Warping Algorithm (DTW) provided the most feasible and accurate approach available.
Additional end use event physical context algorithms have been developed to aid accurate
end use event categorisation. This paper firstly presents a thorough discussion on the single
water end use event analysis process developed and its internal validation with a testing set.
This is followed by the application of the developed approach on three independent
households to examine its degree of accuracy in disaggregating two weeks of residential flow
data into a repository of residential water end use events. Future stages of algorithm
development and testing is discussed in the final section.
Key words: water end use event, water micro-component, residential water flow trace
disaggregation, hidden markov model, dynamic time warping algorithm, water demand
management
2
1 Introduction
After decades of inadequate metering of water use, organisations have come to the realisation
that it is almost impossible to evaluate the effectiveness of residential water demand
management schemes without accurate and appropriate measurement of actual water
consumption at an end use or micro-component level. This desire to better monitor and
analyse water consumption has led to the conceptualisation of a Knowledge Management
System (KMS) which is able to collect real-time water consumption data through a smart
water metering system, transfer and store the data into a knowledge repository, analyse and
disaggregate data into a registry of end use events, and produce a wide range of reports which
can be accessed on-line by a broad range of users (e.g. consumers, water utilities, government
organisations, etc.) (Stewart et al., 2010).
Water end use data registries enable much deeper understanding on the determinants of
residential urban water demand. Recently, there have been a numbers of reported studies that
have utilised water end use data registries for a range of statistical modelling and decision-
making purposes. These studies explored the determinants of shower end use consumption
(Makki et al., 2011), influence of residential appliance stock efficiency on end use
consumption (Beal et al., 2011a, 2011b; Willis et al., 2011a), impact of visual display
monitors on shower end use consumption (Willis et al., 2010a; Stewart et al., 2011),
influence of water conservation attitudes on discretionary water end use consumption (Willis
et al., 2011b), and recycled water end uses in residential households (Willis et al, 2011c).
Water end use studies such as these demonstrate the benefits of having available such data,
and how it can be utilised to better inform urban water practices and policy going forward.
However, the findings reported in these research papers were made possible through resource
intensive flow trace analysis tasks in order to categorise water end use events; such an
approach is not economically viable for large scale samples (i.e. citywide end use dataset).
Before such an information system can be realised and the benefits of citywide end use data
registries yielded, improved approaches for disaggregating high resolution water
consumption data into end use events is required. Therefore, the key enabler for this KMS is
the development of pattern matching algorithms which are able to automatically categorise
collected flow trace data points received from wireless data loggers into particular water end-
use categories. To tackle similar complex problems, such as hand writing segmentation and
recognition, speech recognition, fingerprint recognition, surface water level and seabed
liquefaction predictions (Sannasiraj et al., 2004; Zhang et al., 2007), many powerful pattern
3
detection tools have been established and widely applied, including Artificial Neural Network
(ANN) (Moon et al., 2009), Dynamic Time Warping (DTW) (Nguyen et al., 2011) algorithm
and the Hidden Markov Model (HMM) (Cho et al., 1995). However, all these methods are
data driven that require training datasets in the case of ANN and HMM models or reference
datasets for the DTW method. To facilitate this study, over 7000 days worth of 5 second
interval flow trace data was collected from 252 homes within South East Queensland (SEQ),
Australia. This flow trace data was manually disaggregated into nine different water use
patterns, including shower, faucet (tap), dishwasher, clothes washer, full-flush toilet, half-
flush toilet, bathtub, irrigation and leak. Utilising this extensive training set and through
trialling many of the above mentioned analytical techniques (i.e. ANN, HMM and DTW), it
was revealed that a hybrid approach using a combination of the Hidden Markov Model
(HMM) and Dynamic Time Warping Algorithm (DTW) is the most suitable for solving this
type of pattern matching problem.
A comprehensive methodology is presented herein to illustrate the identification process for
all of the single event category’s disaggregation from the collected flow trace data. To
demonstrate the entire model design and verification process for single events, a complete
clothes washer event analysis is shown. The approaches for disaggregating combined events
(i.e. shower concurrent with toilet flush) are beyond the scope of this paper; however, they
are briefly discussed in the context on the overall objectives of the greater study herein.
2 Background
2.1 Existing water metering process
One of the deficiencies of the existing urban water management system is the current
simplistic metering process. The current water metering system does not typically provide
real-time water consumption data and in cases where it does, it does not provide a sufficient
level of data resolution to allow water end use event categorisation. Conventional water
meters count each kilolitre of water as it passes through the meter without the ability to
record when (i.e. time of day) and where the consumption takes place (e.g. washing machine,
leakage, etc.). Water consumption readings are generally recorded manually on a quarterly or
half yearly basis. Under most situations, a whole year’s worth of water consumption data is
described by only two to four data points in the water businesses billing system. No further
information is available to draw upon should there be any queries. Obviously, this
conventional water metering system produces limited, delayed water consumption
4
information. Such a metering system is unable to provide effective support to water planning
and management processes. This current metering approach is not adequate to meet the
increasing level of government scrutiny on the utilisation of water resources and does not
assist society at large to address the pressing water security issues associated with climate
change.
2.2 Advent of smart water metering and transition to more advanced systems
The concept of smart metering embraces two distinct elements; meters that use new
technology to capture water use information and communication systems that can capture and
transmit real-time water use information. Smart water meters essentially perform three
functions; they automatically and electronically capture, collect and communicate up-to-date
water usage readings on a real-time (or nearly real time) basis (Neenan, 2008). Hence, a
smart meter is a high frequency (e.g. 72 pulses per litre) sampling device (a data logger) that
allows for the time series reading of water consumption. The information is available as an
electronic signal; it can be captured, logged and processed like any other signal.
An extension to the existing architecture of the smart water metering system is an advanced
integrated water management system as conceptualised here, which is designed as a powerful
tool to support an integrated water conservation management system, in order to sustain
water savings. The primary functions of the system include, but are not limited to, collecting
real-time water consumption data through a smart water metering system, transferring and
storing the data into a knowledge repository and analysing the data, and producing a wide
range of reports which can be accessed on-line by a broad range of users (e.g. consumers,
water utilities, government organisations, etc.).
However, the realization of such an advanced integrated water management system will only
become possible when there is a robust analytical model available which can automatically
and accurately disaggregate flow trace data into individual water end use event categories.
The design and verification of such an analytical model is the ultimate aim of this study
2.3 Reported water end use studies
In recent years, a number of residential water end use studies have been completed using a
range of single or mixed methods, such as household auditing, diaries, high resolution smart
metering and pressure sensors, with a diverse range of per capita end use summaries. Jacobs