The Dutch water supply company Vitens asked VORtech to help them improve their software for detecting and locating leakages. We developed refinements of their Dynamic Bandwith Monitor (DBM) which Vitens has released under an open source license. For localising leakages, VORtech developed a new algorithm based on the standard approaches that are implemented in the open source toolbox OpenDA.
Data science for managing distribution networks
Leakages can be detected through deviations in the pressure and flow in distribution networks. But when is a deviation actually a leakage? To anwer this question, water supply company Vitens developed the Dynamic Bandwidth monitor (DBM), which is available under an open source license. The idea is simple: when the amount of water used in a region is significantly less than the expected amount, then water is getting lost somewhere in the region. But this rests on the assumption that the expectation is sufficiently accurate and that the observations are reliable. VORtech helped to improve these aspects.
The prediciton in the DBM is based on the trend in the past period. The relatively simple approach for determining the expected water use in the DBM proved to be surprisingly effective: advanced deep learning provided little improvement. But optimizing the settings of the prediciton algorithm did make the expectations better. VORtech implemented an optimization method to do this. In addition, we enhanced the reliability of the DBM significantly through a further two improvements. First, we filtered out situations where a deviation in water use was not limited to a single region but was part of an overall trend. Second, we filtered out situations where water that was lost in one region turned up in an adjacent region. In this case, the supply from one region to another was obviously not measured correctly and this should not lead to a leakage alarm.
Once a leakage has been detected, the next question is obviously: where is it? This question can be answered by using a computation model of the distribution network to compute backwards from the observations to the source of the deviation. VORtech has extensive experience with the techniques that are available to do that. We are one of the lead developers of the open source toolbox OpenDA that implements these techniques. In a number of projects, VORtech used OpenDA in combination with the open source package EPANET that is often used for the simulation of distribution networks. The combined OpenDA/EPANET tool was used for Vitens to develop an effective method for locating leakages. This method is now tested and improved. Over time, it will probably be released under an open source license as well.
Want to know more?
If you are interested to find out how leakages can be detected and located using open source tools, feel free to contact us. We will gladly come by to discuss the options.