Digital leak detection

How digital analytics can support the water industry by detecting and predicting leaks

Currently, the estimated daily personal water consumption rate in the UK is on average 152 l/day (as per 2021)1, equating to an estimated total daily usage of 14 billion litres per day. Worldwide, climate change will have a significant impact on the volume of useable freshwater, and by 2025 it has been estimated that two-thirds of the world’s population may face water shortages.

Action is required now to reduce demand, increase supply and apply the principles of a circular economy to meet future freshwater requirements. There could be enough water to meet the world’s growing needs, but only if we dramatically change the way water is used, managed, and shared. 

In recent years the UK Water Industry has made great strides in leakage reduction. However, with the Ofwat requirements for a 16% reduction in leakage by 2025 much more work is required. Therefore, to reduce leakage levels to those required by the water regulator, it will be important to use new technologies in leakage reduction.

“we need to dramatically change the way water is used, managed, and shared”

The availability of inexpensive computing power and measurement databases has enabled the development of powerful data analysis techniques that allow metering networks to be monitored daily. Such techniques can give operators details about meter performance and leakage, and are much more effective than the traditional water balance calculation over the distribution network. 

Water leakage

Leakage is not the only way water can be unaccounted for, as water can be lost in several ways: leakage or metering inaccuracies, unbilled consumption or even theft. It is therefore often better to use the term Non-Revenue Water (NRW) to describe the water that has been produced and is lost before it reaches the customer. Put simply, NRW = unaccounted for water + all other water that is used but not charged for.

It has been estimated that if the levels of NRW worldwide could be reduced by a third, this would provide water savings that would be sufficient to supply 800 million people. This would offer a financial benefit of around USD 13 billion per year. The global volume of NRW has been estimated to be 346 million m3 per day2, equating to 126 billion m3 per year. If valued conservatively at USD 0.31 per m3, lost water can be valued at USD 39 billion per year. In times of climate change and water scarcity, these vast volumes of NRW could go a long way to help meeting the extra demands that will be placed on freshwater supplies in coming years.

Digital Leak Detection

The reduction of NRW can provide numerous benefits: reduced operating costs, better water resource efficiency and an increased water supply at a fraction of the costs associated with the building of new water production facilities. However, it is important to recognise when water leakage is actual leakage and not caused by inaccuracies in the flow meter measurements.

For water companies to meet these leakage reduction targets, the accuracy of their meters must be determined. A better understanding of the uncertainties associated with the flow meter measurements allows them to be considered and accounted for in water balances. This results in the development of more accurate water balances and a more accurate reflection of water leakage rates. Providing more accurate flow meter readings will also allow the water companies to provide more accurate information into their models, so they can optimise their water network operations.

Flow measurement

Flow metering is essential for measuring water usage and managing water supplies. Most water meters around the world are small and are primarily used to record domestic water consumption. However, larger meters, whilst smaller in number, measure an equivalent volume of water and are key to managing both resource and demand. It is principally through the use of larger meters that we quantify how much water is being abstracted from underground aquifers, rivers and other water bodies to provide clean water supplies to our cities. Both small and large meters are therefore essential for effective, economic and sustainable water management.

The need for accurate measurement on large diameter transmission (trunk) mains is of vital importance to the global water industry, to optimise water resources, accurately estimate leakage and calculate the water balance across the water distribution system. A significant proportion of modern flow meters rely on assumptions about the flow profile in the pipe, i.e. assuming a fully formed symmetrical flow profile. Bends, valves, and other pipe components upstream of the measurement device will affect the assumed flow profile and therefore the accuracy of the meter.

Uncertainty is the degree of doubt about a measurement. Undertaking an analysis of the uncertainty involves identifying the main influences that affect the final measurement. This will result in a number which represents the “margin of error” in the measurement. Applying this across the network gives an uncertainty in the water balance; that is, a margin of error within which the mass balance should lie. Identifying the main contributors to this figure can ensure that capital expenditure is targeted to areas in the network where it will produce the most benefit.

Digital Leak Detection

“accurate measurement on large diameter transmission (trunk) mains is of vital importance to the global water industry”

However, the rigor with which uncertainty analysis is applied in the water industry varies widely. Some companies use only the manufacturers’ accuracy claims, while others devote effort examining meter history and location to identify the key influences. In contrast, in the oil and gas industry uncertainty analysis is integral to the business. This is driven principally by the high value of the product and companies simply cannot afford inaccurate flow measurement. Accounting for uncertainty in flow measurement allows them to see the ‘bigger picture’ – enabling them to calculate financial exposure on fields and make strategic decisions.

Increasing demands are being placed on water due to climate change and population growth. There is also the potential for large water transfer schemes to be developed. These factors are likely to result in the need for a more accurate knowledge of the flow as the effective ‘price’ and importance of water is increasing. Therefore, the water industry would gain real benefits from adopting the practice of the oil and gas industry, by applying rigorous uncertainty analysis at the heart of their network monitoring procedures.

With flow monitoring becoming an increasingly important part of a water company’s business, it is therefore crucial that:

  • Good measurement practice is followed at all times
  • Established procedures and processes are used and regularly updated
  • Staff training and competence is recorded and regularly verified

This helps to ensure that the data obtained from the metering network is reliable and can be used in demand forecasting and strategic planning. This data also act as inputs to a range of numerical analysis techniques, such as gross error detection, uncertainty analysis and data reconciliation. These techniques are cost-effective methods of improving the effectiveness of network monitoring and are now being frequently applied in the water industry.

Modern digital analysis techniques

Digitalisation is changing the way we work and live with new innovations that impact all facets of today’s industrial world. Increasing ubiquitous connectivity, the transformation from hardware to software, the rise of smart sensors, and big data, all call for more new and comprehensive approaches. Advances in technology have changed the ability to record, collect, analyse and share data. As a result, vast amounts of big data from industrial facilities continues to increase, often in real-time.

Developing tools to interpret these historical trends within the data to monitor the current condition and performance of devices has led to a new range of solutions available for industry. However, how do we provide confidence that systems are performing as they should? While industry reaps the benefits of digital technologies, the complexities involved imply new challenges in safety, security, and confidence.

TÜV SÜD National Engineering Laboratory has undertaken extensive research into digital analytical techniques to improve the information gathered by modern flow meters. This research is based on a huge database of testing different flowmeters under a range of different operating conditions within their own laboratories.

There are two types of computer models used for solving engineering problems such as those experienced in oil and gas production: physics-based models and data-driven models. These two classes of computer models differ from the way they represent physical processes.

Physics-based models attempt to gain knowledge and derive decisions through explicit representation of physical rules and generating hypotheses regarding the underlying physical system. These models are driven by physical processes and can normally be described by a set of mathematical (theoretical) equations. For example, Navier-Stokes (N-S) equations explain the motion of fluids and can govern Newton’s second law of motion for fluids.

On the other hand, data-driven models uncover relationships between system state variables without using explicit instructions. These models employ algorithms to perform statistical inference and pattern recognition wherein a model maximises its performance through an iterative learning process. It should be noted that such models do not contain the full complexity of the true physical phenomenon. Instead, they provide a less complex (but valuable) abstraction that approximates the real system. Because these models do not necessarily require knowledge about the physics of the processes, they are very flexible when testing different hypotheses and making predictions.

TÜV SÜD have developed a range of data-driven models for the oil and gas industry, and these are now being considered for use in the water industry, these models include:

Condition-based monitoring (CBM)

TÜV SÜD have developed CBM services for coriolis and ultrasonic flow meters for the oil and gas and oil bunkering industries. Previously testing was undertaken whereby known errors were introduced upstream of the flow meter and the flow meter diagnostic data, which all modern electronic flow meters generate, was interrogated in order to develop a meter health verification tool. 

Digital Leak Detection

This allowed software developed by TÜV SÜD to examine the data diagnostic information generated by the flowmeter to allow an assessment of the performance of the flow meter to be made. The aim is to now take this technology and apply it to electromagnetic flow meters, so that a meter verification tool can be developed that would allow the water industry the chance to verify the performance of their large flow meters and to ensure that the values of flow being recorded are a true and accurate assessment of real-world flows.

A further benefit of a CBM system is its capability to allow the user to modify when the flow meter is calibrated. Typically, most flow meters are calibrated using a time-based approach, whereby the flow meter has to be sent away regularly for calibration, sometimes at an annual interval. This is often an expensive and possibly unnecessary process. A more cost-effective system would be to only send the flow meter for calibration when it was actually required to be re-calibrated. This would save the operator both time and money. By using a CBM type system this modification from a time based to a more dynamic calibration approach is possible.

This is called Condition-based Calibration (CBC) and again uses the digital diagnostic data generated by most modern electronic flow meters to determine if the flow meter is still within its original calibration. By using digital science techniques it is possible to monitor the live diagnostic data and determine if any of the measured variables shows signs of drift from the baseline calibration settings. If a drift is observed, the software can advise the user that the meter should be sent for a calibration.

The benefit of this system is that, by continuous monitoring of the diagnostic data and the standard instrument outputs, it is possible to predict the flow meter calibration drift over time - thus allowing a reduction in costs by making use of a more dynamic operating pattern. This technique has been developed and tested for both coriolis and ultrasonic flow meters and could quite easily be transferred to electromagnetic flow meters.

Data validation and reconciliation (DVR)

In the past, water companies have used simple mass balance techniques to estimate the amount of leakage and other unmetered abstraction in their distribution and waste treatment flow networks.

DVR takes this a stage further by identifying the flow meters most likely to be responsible for imbalances which enables water companies to target maintenance to where it is most required. DVR is a calculation technique that is increasingly being used by water companies to monitor the quality and reliability of flow measurement data acquired from trunk mains. It performs a network self-check to ensure that all the measuring devices are consistent with each other. Using this technique, engineers may quickly identify which meters are reading outside their uncertainty bands and take appropriate remedial action. It can also be used to determine the level of leakage in a network.

By combining this technique with the CBM technique, where confidence in the performance of the flow meter is known, it will be possible to determine if any missing water in the network analysis is due to actual leakage or whether it is due to errors in the flow meter performance. By having confidence in the flow metering of the entire network it will be possible to determine approximate locations where the leakage may actually be occurring.

First and foremost, DVR may be used as a diagnostic tool to pinpoint exactly which meters are operating outside their uncertainty bands. This may indicate that operators have made incorrect assumptions about the uncertainty of the meter. This can be changed and the reconciliation re-run with the new value. Alternatively, it could mean that the meter has drifted out of calibration or that a fault has developed. Either way, the ability of the technique to highlight anomalies will allow operators to target maintenance at specific equipment. This allows plant operators to make the most of the data that they have – with the accompanying financial and operational benefits.

Fault prediction analysis

By making use of historical data and using machine learning it may be possible to predict where leakage is likely to occur in the water networks. This will require more information than what is available from flow meters and will require the data from all pressure measurements in the network.

The objective is to use machine learning to allow the software package to identify potential trigger signals from the instrumentation placed on the network. By looking at historical data, where known bursts or leaks have occurred, the signals from the instrumentation can be used to train the machine learning software to recognise these trigger signals. It then provides an alert when such signals are observed in live data, thus predicting when likely bursts or leaks may occur. This will give the water companies advance knowledge on when and where potential problems may occur and so they can undertake an inspection of the pipe before a burst or leak develops.

The individual software packages and techniques described above have all been developed for the oil and gas industry, but can be easily adapted to the benefit of the water industry. This will require physical testing to be undertaken to baseline the differences in performance between electromagnetic and coriolis flow meters. It will also require historical data on water networks, where known burst and leaks have occurred to train the machine learning programmes. However, the good news for the water industry is that the techniques and the methodologies for undertaking this work have all been identified and constructed previously.

Digital Leak Detection

Combining multiple data analysis techniques such as these will allow modern software techniques to be developed that will enable water companies to:

  • Verify the performance of modern electronic flow meters
  • Perform network analysis and identify leakage on their networks
  • Predict where leakage events may happen in the future

Data is the most valuable asset

Optimising data utilisation is an operational imperative, especially to water companies under environmental, regulatory and resource pressure. Failure to protect significant metering investments, by not complementing it with modern and cost-effective data analysis techniques, risks increased capital and operational expenditure through poor targeting of effort.

“optimising data utilisation is an operational imperative”

Therefore, smart metering and network analysis will have to be used together to achieve the improvements necessary to meet the challenges facing the water industry today. This will give water companies much more confidence in their data, alongside their investment decisions and operational expenditure levels. The application of these techniques, along with the recent advances in electronics and computing power, will give water companies the tools to meet the challenges facing them in the 21st century.


References

  1. Statista, "Average household water usage per person per day in England and Wales from 2016 to 2021," Statista, [Online]. Available: https://www.statista.com/statistics/1211708/liters-per-day-per-person-water-usage-united-kingdom-uk/. [Accessed 05 August 2022].
  2. R. Liemburger and A. Wyatt, "Quantifying the global non revenue water problem," Water Science & TechnologyWater Supply, no. July, 2018.