Few businesses can operate for long while experiencing annual inventory shrinkage of 20 percent or more. Yet, that’s the situation faced by water utilities around the world as their distribution systems regularly pour treated, potable water down the proverbial drain. Water loss represents a major fraction of non-revenue water (NRW), and it’s a problem that is easily addressed with the right tools.
According to the International Water Association’s best practice recommendations for water balance studies, more than 65 percent of NRW arises from unauthorized water consumption, meter inaccuracies, and leakage.
The Ontario Sewer and Watermain Construction Association in Canada reported that as much as $1 billion worth of fresh, clean drinking water disappears into the ground every year from rotting, leaky municipal water pipes. {sidebar id=1}
In the United Kingdom, where on average more than 15 percent of water is lost, leakage targets are set as a key performance measure with severe financial penalties if targets are not met. To meet the requirement a number of tasks must be undertaken:
— First, the level of water loss must be assessed.
— Second, the location of the losses must be determined.
— Third, reactive and proactive measures to improve the condition of the water mains must be adopted.
The problem is, existing leak detection processes (see sidebar, “Traditional Approaches to Detecting Water Leakage”) are time-consuming and costly. An interesting research effort at Bentley’s Haestad Methods Solution Center in Watertown, CT, is seeking to develop an innovative, cost-effective strategy for estimating the location and extent of hidden leakage. Included in the study are both distributed background leakage and high-intensity leakage “hot-spots.”
The leak detection approach proposed applies genetic algorithms (GA), a search technique based upon the principles of natural evolution and genetic reproduction. Using GA, leaks are simulated across the network and the results of each simulation are compared with the known flow metering and pressure monitoring locations. The working notion is that by making repeated runs with refinements introduced in a manner that emulates natural selection, an optimal solution can be found. This is an intriguing geospatial engineering modelling problem. Over the course of my 30 years in this industry, I have seen geospatial information systems converge tightly with the mainstream solutions for the modelling of municipal hydraulic network infrastructure.
Bentley spends a significant part of its R&D budget tightening integration among specialized industrial applications such as hydraulic systems modelling with GIS systems. At the Haestad Methods Solutions Center, we have been working for more than 25 years in modelling the hydraulics of buried water distribution mains. Our analysis engines are the perfect consumers of geospatial information systems that manage the mapping of water assets. If we could use our technologies to effectively map leakage zones and hot-spots, the industry would benefit, and we would have yet another new differentiator for our commercial solution.
Partnership Toward Innovation
Led by Dr. Zheng Wu, one of Bentley’s experts on advanced techniques for optimizing solutions to rather complex modelling problems, the study applies “fast-messy” GAs to optimally calibrate models and plan their operation. Dr. Wu had linked arms with Paul Sage, modelling development manager for United Utilities PLC in the United Kingdom, to jointly develop a technique for determining the most likely locations to concentrate exploration for underground leakage in systems.
Their goal was to discover a way to apply Bentley’s WaterGEMS technology to the problem. They had worked for several months, presented some preliminary findings at a couple of industry conferences, and were now ready to apply the technology on a large system – one for which a major leak had already been discovered. If this new technique could “find” the known leak, that would be a great proof-of-concept that might lead to wider adoption at the utility.
So, the two partners set up the simulations, letting the optimizing algorithms cook away, only to deliver a result that was a disappointing negative finding. The new approach failed to zero in on the known leak. This is the modelling equivalent of a missile test-launch failure. But how could it happen?
Our researchers examined the results and came up with a possible mechanism for the failure. The explanation lies in the fundamental working premise. Dr. Wu had formulated his strategy as a hydraulic model calibration problem. Utilities use hydraulic models to simulate operation and support decisions. This requires a calibration process in which the modeler adjusts and tunes modelling parameters, such as pipe friction roughness and estimated water demands, in order to coerce the model into matching actual observed pressures and flows.
The two investigators reasoned that leakage impacts performance in ways that should be approximated as pressure-dependent discharge locations. The calibration algorithm should then simultaneously estimate leakage locations and/or intensity while varying pipe roughness across the network.
The problem arose because the researchers challenged their new technique against a model that was in operational use at the utility. It had already been substantially calibrated! While leakage was accounted for in the course of calibration, it was distributed in a simplified way.
Upon inspection of the calibrated model, the modelers observed that some assigned pipe roughness values for pipes of similar age and material were inconsistent and varying spatially. They reasoned that these values might be artificially deflected as a side-effect of the calibration in the vicinity of major leaks.
Indeed, consideration of the hydraulics of a hypothetical leak determine that pipe values upstream of a leak can be expected to calibrate rougher than truth and those pipes downstream of a leak will calibrate smoother. The prior model calibration exercise had succeeded in obscuring an intense leak. After the researchers reset the roughness values to consistent and reasonable hydraulic handbook values and reran the optimization algorithm, the leakage calibration zeroed precisely into the known leakage hot-spot.
Success at last! Still, expectations must be managed. Coupling modelling and geospatial analysis might provide a sort of buried treasure map that operators can use as a basis for uncovering unallocated water losses. However, this technology will not deliver the analysis equivalent of X-ray vision that will show maintenance teams where to commence digging and repair. It does, though, show great promise in directing utility owners and operators to locations to focus testing and explore and test with their acoustic devices, saving time and minimizing cost.
An Exemplary Project
This has been an exemplary geospatial engineering project on two levels. First, the technology is useful and rational and builds on a solid theoretical and commercial foundation. Second, and most importantly, it demonstrates the good result of commercial software vendors cooperating with end-users to develop well-purposed applications.
This was a mutually cooperative effort in the truest sense of the word, not a commercial partnership. It was speculative and reflects the dedication of two like-minded researchers challenged by an interesting notion with a desire to innovate and develop something useful. Their work is testimony to the spirit behind efforts to unify geospatial modelling and engineering in new and exciting ways.
To learn more about this study please visit the Bentley WaterGEMS web page at www.bentley.com/WaterGEMS where you can download a white paper and a PDF poster as presented at the ASCE 8th Annual International Symposium on Water Distribution System Analysis in Cincinnati (August 2006) and the AWWA ACE07 conference in Toronto (June 2007).
Link Directly to the WHITE PAPER.
Learn more about the latest developments in WATER LEAKAGE.
By: Jack S. Cook, Jr. VP Water Solutions, Bentley Systems, Watertown, CT USA
www.bentley.com
