Water Journal : Water Journal December 2012
refereed paper intelligent water networks water DECEMBER 2012 105 with the experiences of other utilities (TaKadu Ltd, 2012). In comparison, the average length of time of a burst before a customer call-in and response is typically three to four hours for bursts in the middle of the night. This represents a substantial time-saving opportunity if integrated with a control room that would promptly act on the notice. This initiative would result in reduced water loss as well as reduced customer impacts. Due to the limitations of the SCADA system at the time of the trial, Yarra Valley Water was not able to fully explore and evaluate this. This will be further explored in later trials. These results were encouraging, despite being limited by our legacy SCADA system and a noticeable improvement was seen. Events were classified quickly and accurately, and there were significant time savings for operators in responding to events. The improvements in the system were tangible, and the ability of the system to classify events in a clear manner was proven. Specifically, as the trial progressed the improvement was noticeable and the categorisation of leakage became clearer, enabling greater efficiencies. Accuracy of geolocation classifications One of the primary potential value-adding benefits of TaKaDu is the ability to not only identify and classify leakage, but to narrow down the area in which the leakage is located; this is called geolocation. Yarra Valley Water was keen to assess the accuracy, range and success of this locating leakage method. This was assessed by checking the accuracy of the identified leakage with real-world locations. From the six geolocations during the initial trial period, all but one was located in the area of highest probability (an example is shown in Figure 10). In that one outlier case, it was located just on the border of the high probability region. Therefore, the success rate was proven to be very reliable. The resolution of the geolocation algorithm is dependent on the number of metering points. In typical distributions zones, with flow and pressure meters located at the zone inlet points, the system was able to geolocate down to 25% of the connections and area of a zone. In one distribution zone, Doncaster South, six additional low-cost temporary pressure loggers were installed to work alongside the three existing zone inlet sites (with both flow and pressure meters at each). Using this, the algorithm was able to detect down to an area of 13% of the connections in the zone. The geolocation capabilities of the system were apparent, and continual improvements have been made to the algorithms. Currently the system is able to suggest priority on specific pipes, not simply just areas. Increasing metering density greatly increases the resolution of the system. Ability to detect meter issues and failures Although a SCADA system can alarm on meter failure, TaKaDu promoted the idea that they could identify many types of failures and slow degradation of meters much more reliably than SCADA thresholds. This was measured by observing the response times to meter issues and the ability to detect and classify problems compared to historical experience. Figure 6. Demonstrating the proportion of classifications of the 417 events seen during the initial stage of the trial (March to June 2011). Figure 7. The events alerted by the system were then responded to by the operators. This classification shows the confirmed nature of the events generated by the system for the 417 events in the initial stage (March to June 2011). Figure 8. Detailing the classification of events for the second stage of the trial from June to September 2011. Unconfirmed leakage and bursts were events identified in the system with a high likelihood of existing, that were not easily identifiable in the field or through existing records. Figure 9. Showing the information in relation to the leaks detected by the system in the second stage of the trial from June 2011 to September 2011. Note the increase in leaks found as the trial goes on, and the increased cost savings.
Water Journal February 2013
Water Journal November 2012-1