Water Journal : Water Journal December 2012
refereed paper intelligent water networks water DECEMBER 2012 103 of information sources that can be examined. A human operator is limited by the amount of information that they can trawl through, which takes time, whereas software systems can perform this far more quickly and from far more sources. The purpose of this is not to replace human operators, but to increase their efficiency by enabling them to monitor more sources and find the most important pieces in the sea of data. Such a system would scour the numerous information sources and find the significant trends and changes that a human eye could not see and either take action or present them for attention. At Yarra Valley Water, for instance, the data sources include hydraulic models, over 2,000 SCADA data points, works tracking databases, a GIS system covering 9,825km of mains, distribution zone formulae, wholesale water bill information and customer metering. Checking all this information manually or switching between the sources of data presents numerous difficulties. IBM is approaching the problem from a data unification and analysis perspective. The Strategic Water Information System (SWIM) (IBM, 2012) uses the business intelligence and analytics experience of IBM and applies it to water networks. Their SWIM platform includes cloud- based solutions, with dashboards for reporting, modelling and planning inputs. While this is useful from a broad perspective in business planning, at an operational level there is not yet a mature water-specific delivery solution. Siemens approaches this challenge at both the instrumentation and city- wide levels. They have experience in developing and deploying metering technologies that are accurate and cost effective with improved meter intelligence. They have sensor packages that are able to detect and correlate leakage and systems that integrate directly with pump or station PLC systems (Siemens, 2012). There is a software package that then analyses this, and building on top of this they have a model of a ‘City Cockpit’ (Siemens, 2010). This dashboard view is an approach of providing summarised key information about performance, changes from normal behaviour and significant events. There are layers, which also then present the details of performance and control optimisation right down to the physical asset level. Aside from several small-scale tests for leakage and optimisation, there is still development work to be carried out before this city- wide system is commercially available and wide-scale ready. TaKaDu is an Israeli start-up firm that approaches the challenge from a signal analysis perspective. Their product is a web service portal that synthesises utilities’ information from a variety of sources and presents it back to a user (as shown in Figure 4). Being formed out of a telecommunications and technology background, their basis is to analyse information by time series and build a model or system that understands what is ‘normal’ (Scolnicov and Horowitz, 2010). With this computerised intuition of what types of behaviour are normal, the software presents alerts to users identifying and locating where abnormal issues have occurred and giving it a classification. The models used are based on historical analysis, spatial analysis and constant evaluation for matches of best fit with other similar distribution zones. When a deviation between the meter information and the model is detected using time-series analysis, probabilistic tests are run to classify the event (as demonstrated in Figure 3). This then tests various possibilities to assign a classification that most likely explains the observed behaviour. These are presented in an ‘inbox style’ system summary of events, as seen in Figure 2. The system uses heuristic algorithms where the user’s input and feedback ‘trains’ and adjusts the model, providing a system tuned both mathematically and by users to the optimum state. The full details of TaKaDu’s algorithms are proprietary, but base reference information can be found in their US patent, in TaKaDu’s published white papers, and through discussion webinars with their research team. The benefits espoused by TaKaDu cover a wide range of areas – specifically direct cost reductions and savings in water volumes, early detection, faster repair time, as well as damage reduction (TaKaDu Ltd, 2011). These direct benefits are supported by the ability to prioritise and better manage works, identify and solve network issues and increase meter uptime. Overall they propose to reduce customer impacts and increase regulatory compliance. Though this product has a cost, the TaKaDu’s premise is that it directly makes a strong return on investment on a month-by-month basis. This system has been rolled out to several utilities in Europe, Asia and Latin America, including Thames Water (UK), Aguas de Cascais (Portugal) and Aguas de Antofagasta (Chile). It has also received green technology awards from the International Water Association (IWA), World Economic Forum and many more (TaKaDu Ltd, 2012). Based on this track record and maturity in development, Yarra Valley Water made the decision to trial the TaKaDu system and investigate the benefits in the Australian environment. Figure 4. Breakdown of the information flow in the TaKaDu system (TaKaDu Ltd, 2012). Figure 3. The TaKaDu interface and detailed information on a particular event. The green line shows the predicted expected normal values for this zone, and the actual data are shown in blue. On the right are details of leakage location using geolocation.
Water Journal February 2013
Water Journal November 2012-1