The issue of how best to tackle water leakage might be literally as old as the hills itself, but water companies are increasingly taking a high-tech approach, by utilising a wide range of artificial intelligence (AI) and machine learning tools to reduce wastage.

The listening sticks of old have been replaced by acoustic microphones or loggers, which can be placed in pipelines and set to listen for any signs of leakage, with the latest software being used to analyse the recordings and spot warning signs.

This might all sound like science fiction, but today it is science fact and could be an important weapon in the water sector’s battle to improve its track record.

In October last year, the environment, food and rural affairs committee warned three billion litres of water are leaked every day and Ofwat’s target to reduce leakage by 15 per cent by 2025, as part of PR19, was not ambitious enough.

The cross-party group of MPs also called on the water industry to collectively aim to reduce leakage by 50 per cent by 2040, rather than 2050.

In April this year, trade body Water UK hit back with a pledge to triple leakage reduction rates by 2030, as part of a new sector-wide initiative to work in the public interest.

With all eyes on the water sector and its efforts to improve leakage rates, a number of utilities are now rolling out AI solutions to help crunch these all important data patterns.

Leakage projects

Southern Water recently launched a major initiative to tackle leakage in south Hampshire, the Isle of Wight and Southampton.

As part of the project it will be spending £250,000 on installing an extra 1,600 acoustic loggers, to add to the 3,760 already attached to its water pipes.

The loggers will use digital technology to monitor the noise water is making and trained technicians will feed better data from an entirely new leakage reporting system to the find and fix teams.

Phil Tapping, water demand manager, says the programme will help it reach a target of cutting leaks by 15 per cent by 2025. “We’ve had a good record historically on leakage and are near the top of the league table in England, but we have to do more,” he says.

The extra funding in Hampshire is part of an overall additional £4 million investment in acoustic logging by September and a £2.4 million extra investment in leak fixing in the coming year.

In October last year, United Utilities became the first water company in the UK to introduce large-scale AI into its operational systems. The Warrington-based company has signed a framework agreement with the Canadian technology firm Emagin after a successful trial earlier in the year.

Emagin’s artificial intelligence platform, called HARVI, can assess vast amounts of data on a wide range of factors such as weather, demand for water, pump performance and electricity prices.

This is used to help make decisions on the most cost-effective and efficient way to run pumps, detect burst pipes and minimise the risk of discoloured water.

During the 12-week trial, which took place across Oldham in Greater Manchester, HARVI demonstrated energy savings of 22 per cent. United Utilities now plans to deploy the artificial intelligence platform in phases across the whole North West region by the end of 2019.

Meanwhile, Yorkshire Water started trialling smart analytics in January to find and fix leaks in its network around Hebden Bridge and west Sheffield with the aim of avoiding interruptions to services for customers.

As part of the trial, it will provide data to both remote monitoring specialists Servelec Technologies and water consultants Artesia Consulting, who look for disruptions to normal patterns. Any discrepancies will then be flagged and passed back to Yorkshire Water to investigate.

Finally, after teaming up with Capgemini, Severn Trent announced in December last year that it has started to use machine learning – which is a form of AI that allows systems to automatically learn and improve – to help transform the way it approaches leaks.

“There are a plethora of solutions to detect leakages, such as smart balls, which can put be put in to the main pipes, which can detect leaks, or acoustic loggers, which can listen to the sounds of leakage like a traditional hydrophone and log them,” explains Capgemini’s head of insights and data for UK energy and utilities, Raj Malayathil.

“One area water companies should be exploring is the power of data and AI in detecting leakages,” he adds.

“Water companies hold a lot of data such as pressure and flow from the network and also information on historic leakage incidents such as the location of a leak, interventions done, weather data, soil conditions, etc. Using AI and machine-learning techniques on these data sets, water companies can easily detect and locate leaks faster than traditional methodologies.

“By taking these techniques one step further, water companies can use these AI and machine learning solutions to plan their mains replacement activities by identifying pipes and assets that have a higher propensity to fail and replace them quickly to improve the network resilience.

“These systems can also provide a higher level of situational awareness in planning incidents in the network due to extreme weather conditions, like the Beast from the East in 2018, and plan mitigations much faster and improve operational resilience.”

Predicting bursts

Mott MacDonald’s head of asset performance optimisation, Tom Joseph, says AI is increasingly being used to predict where the next burst will happen.

“You can predict where pipes will fail,” explains Joseph. “You can do that by using data on previous bursts and other parameters, like if a pipe is under a road, what the traffic is like on that road and what the temperature is on the ground.

“We are also doing a lot of AI work on the wastewater side of things,” he adds. “We’ve been working on correlation work on blockages, and rain-induced spillages, and when a sewer might spill and the effect it might have on the environment.”

In terms of other benefits, Graeme Wright, chief technology officer for manufacturing, Utilities and Services at Fujitsu UK, also sees a big role for AI in helping workers out in the field.

“Through the use of AI, utility companies will be able to harvest their knowledge and skills and transfer information onto a platform that is readily available to advise engineers on-site, changing the way they work significantly,” says Wright.

“For instance, engineers in the field will be able to receive information on their next task in real time, with a complete view of its status and maintenance history. Engineers can also gain access to further information that may help their task, such as manuals or information on how the system ought to be performing. This access will help them assess what needs to be done more expediently and enable them to progress the task in a more efficient manner.

“With the more data that the AI can collect, we will also see engineers start to use and be guided by ‘co-bots’ and ‘chat-bots’, which will provide guidance based on previous tasks and history, sharing knowledge captured from previous works and past expert interventions.

“As a result, utility companies will be able to share knowledge more easily and work more effectively as tasks will be resolved more often in the first instance, without the need to call for another engineer as they will have all the knowledge and capabilities required,” adds Wright.

Capgemini’s Malayathil adds: “In the utilities world, with the advances of AI, we could potentially see self-healing networks in water and electricity.

“As an example, in water networks with an eco-system of intelligent sensors and edge computing, they will be able to automatically detect a potential leak based on sensor data and automatically adjust the network attributes such as pressure and flow to avoid a leakage incident.”

And Kevan Mossman, transformation director, part of the Odgers Interim network, says AI will also help utilities build profiles of their customer base.

“For example, an AI system could monitor search trends around problems in a specific postcode, such as leaks or outages. The AI could then assess that problem and distribute relevant communication to customers who may be affected almost immediately, without the customer needing to contact them,” says Mossman.

“However, AI cannot work without being fed ‘big data’; something that utilities have historically been poor at making available. As a result, the sector as a whole has been a late adopter and it’s only in the past couple of years that companies have begun implementing large-scale digital transformation programmes to ‘crunch’ the data that they need.”

While the sentient AI systems of science fiction are still some way off, there is no doubt that the utilities sector, like many other sectors, is starting to explore the opportunities the technology can offer, especially as it comes under renewed pressure from politicians and regulators alike.

Interestingly, a study by PwC published this year revealed 85 per cent of international chief executives agree that AI will significantly change the way they do business in the next five years. Two-thirds even said it would be bigger than the internet, which itself has turned out to be fairly revolutionary.

Pinpointing a pipe among 600,000 drawings

Northumbrian Water has developed an innovative approach to managing records by using a cognitive service approach to build a data system with quicker response times.