With many complex customers and more frequent company wide demand events, it is often difficult to track exactly where all the water in distribution networks is going, and reliably predict future demand. This makes day to day business activities increasingly challenging even for experienced practitioners including; deciding how much water to produce, managing the movement water around networks, and separating leakage from demand.
As an industry we are being drawn towards ‘smart’ network solutions, which promote artificial intelligence (AI) led decision making and automated ‘closed loop’ network operations. These systems rely on an accurate set of rules to tell them how to behave, and without a full understanding of where our water is going, how can we provide these rules to an automated system?
What's the solution?
To gain a full understanding of network demand, companies need to explore the gaps in their existing knowledge. This includes identifying customers who are poorly represented on their internal systems, either because they are not known by the water company, have an unmetered or incorrectly metered connection, or have unknown additional connections.
Paradigm is a forecasting model which provides the user with a detailed demand breakdown for hydraulic areas and the ability to predict consumptions and expected burst flows.
Demand prediction is undertaken at District Metered Area (DMA) level, with each DMA’s demand constructed using its component parts. This allows the user to immediately identify internal data sets that do not match the expected net flow profile for an area. Paradigm supports the interrogation of this information, and we have innovative techniques which can be deployed to identify likely causes of these problems, e.g. identifying individual unmetered customers.
We use data from multiple clients to improve the accuracy of Paradigm, allowing our client base to benefit from an extended model. This is particularly useful for clients with minor representation of a component within their own demand analysis, as they benefit from the insight gained from a wider industry model. For example, a client may have a small number of areas affected by a specific type of agricultural use, which would normally create a small statistical sample. Combining their areas with others from across the UK can help to identify patterns and dependencies on external data sources.
How we can help
Our close engagement with leakage teams has helped them to understand the underlying issues behind historically stubborn DMAs, ensuring they don’t keep wasting leakage detection time when the problems are customer based. This is creating a change in focus and ensuring valuable leak detection time isn’t spent chasing customer demand.
Using prediction data during demand events (e.g. Ramadan or a hot weather incident) allows our clients to separate leakage and demand, ensuring the effective co-ordination of leakage detection resource.
Over time, confidence in demand prediction data, and the careful resolution of identified issues, can create a strong foundation for automated systems.