Desalination & Reuse Handbook - Flipbook - Page 85
SMART LEAKAGE REDUCTION MARKET PROFILE
Acoustic leak continued: detection primer
Trunk mains vs distribution networks
Trunk mains can be measured using a relatively small
number of fixed correlating pairs, or using fewer
correlators on a ‘lift and shift’ basis. With high volumes
of water transported at high pressure, these are critical
network sections for NRW reduction. Some companies
(Syrinix, Siemens) specialise in precise leak detection from
these network sections.
Distribution networks require more measurement points,
requiring correlation of signals between a far greater
number of ‘couples’. 2–7 sensors per km are required for
good coverage, depending on network conditions.
Mobile vs fixed network
Fixed network acoustic sensors – installed onto
underground piping or at above-ground access points such
as fire hydrants – require data transmission infrastructure
or integration into an existing AMI network. This provides
more comprehensive coverage than mobile acoustic
measurement, but requires greater up-front investment.
Mobile acoustic leak detection is provided either as a
service or on an equipment-only basis. This often requires
the purchase of specialised hardware and software, but
the falling cost of cloud computing and smartphone
technology is beginning to reduce reliance on specialist
hardware.
Active vs passive correlation
Active correlation is instigated by the end-user as needed
to corroborate leaks inferred from other sources such as
flow or pressure sensors, providing instant results at the
cost of shortening sensor battery life.
Passive correlation takes place at pre-determined
intervals, generally at night when background noise is at
its lowest. This generates data on a like-for-like basis, and
significantly increases sensor battery life.
of Business Development and CFO, Uri Gutermann, puts one of
his firm’s typical projects at between 100 and 500 sensors, with
2–7 measurement points per kilometre providing all the coverage
most clients would ever need. He highlights a client in Dietlikon,
Switzerland, where Gutermann fitted 120 measurement points
across 35 km of distribution network.
Software and data management
Responding to the same issues of interoperability and data
overload addressed by Gutermann and Visenti’s integrated
solutions, event management platforms draw data from sensors,
meters, SCADA data, leaks reported to call centres, or any other
inputs the client might have to hand. The different software
can perform a range of services, from aggregating data by DMA
to comparing results in real-time against computer-generated
distribution network models. No matter the specificities, these
programmes use as wide a range of input data as possible to
glean useful information operators can use to make networks
more productive. Some larger concessionaires and utilities
create their own proprietary software – with Suez marketing
their Aquadvanced platform as a standalone product, and Global
Omnium attempting to do the same with their own package – but
for most utilities external software licensing is required.
‘Event management’ software combines and processes
information from multiple sources, displaying comprehensive
or localised data and highlighting network areas in need of
attention. These platforms can come as a standalone package
or as one part of a software dedicated to all aspects of network
management. Israel’s TaKaDu takes the former approach. Many
of the company’s clients are smaller utilities looking to make
better use of available data, but without the budget or need to
integrate a platform to manage all aspects of their network.
TaKaDu’s official announcement of partnerships with acoustic
leak detection specialists Gutermann and Aquarius Spectrum
were both preceded by small utilities in Finland and Israel taking
on services from multiple providers. TaKaDu collates input data
and groups it by DMA to give an instant overview of alerts and
leaks. Its algorithms compare past data with present to highlight
anomalies, cutting down on time spent checking areas of the
network reporting normal results. That the software works with
any and all inputs is an essential part of its role in a utility’s
operational architecture.
Artificial intelligence and ‘machine learning’ are increasingly
applied in leakage reduction, with computer modelling based on
the accuracy of past predictions informing operational decision
making. Systems released by companies such as Servelec
and Royal HaskoningDHV calculate the uncertainty levels in
different network areas. These platforms adjust their alarm
thresholds based on system characteristics such as meter faults
or incomplete coverage in order to reduce false positives, and will
automatically revise these thresholds as data coverage improves.
Visenti’s suite of View products are another example of constant
insights gained from the network based on constant monitoring
of pressure and noise levels integrated with any other data
available from the utility to provide a continually updated picture
of ‘normal’ network conditions.
Meanwhile, Innovyze’s network management platform makes
increasing use of real-time network modelling – known as ‘live
modelling’ – to compare expected and actual demand, noting
discrepancies between predicted and actual water usage to
flag leaks. This requires an up-to-date hydraulic model of a
distribution network to function at its most effective, which is not
a guarantee in many areas. Utilities taking on systems such as
these must become comfortable with the ‘black box’ of artificial
intelligence, where the reasoning behind a decision becomes
potentially indecipherable to operators.
Embracing machine-learning and artificial intelligence by using
platforms such as these can also require patience. As with human
learning, at the outset a software platform’s rate of accuracy can
be relatively low, taking weeks or months to recognise the unique
indicators of a leak in a given network. Nevertheless, artificial
intelligence is here to stay in the water industry, with increasing
adoption across the entire spectrum of software and solutions
providers.
No single solution can ever provide an effective panacea, but
digital technology can provide information and answers where
previously balance-of-probability estimations were required. From
metering projects and rolling sensors into DMA installation, to
network-wide monitoring and automation, a utility considering
its path towards digital transformation can be sure that a set of
solutions exists to fit its unique needs and capabilities.
77
