Dynamo: Delivering improved network understanding
As part of our ongoing Dynamo projects we have identified a range of issues on networks for various clients, including bursts, demand anomalies and network configuration issues. We’ve discussed in previous blog entries how Dynamo was designed to continuously monitor flow and pressure data with the aim of providing near real-time localisations for leakage teams investigating anomalies on a network.
Recently, while analysing a new area, we encountered an interesting case study where a smart controller was placed on the inlet to a DMA. This controller was tasked with keeping a consistent pressure at a critical point located in the middle of a village 3 km away. This was where the majority of demand in this DMA would originate. It was identified that there were high head losses (approx. 10 m) along this stretch of main and that the network control had been set to account for this.
In Figure 1 below, we can see the network controller’s outlet pressure in orange, the critical point pressure in grey, and the flow profile in blue. Notice that the critical point pressure remains relatively consistent over the course of the four days shown as the network controller fulfils its function correctly.
Figure 1: Graph of normal DMA inlet flow, PRV downstream pressure, and critical point pressure
When we came to look closely at this DMA, we found that the smart controller would respond to sudden spikes in flow by increasing the downstream pressure from the PRV. This would cause large spikes in pressure throughout the network, including at the critical point.
The problem was an interesting one as the network control appeared to correctly maintain pressure on the network at all times other than where we found these intermittent spikes in flow. These were measured as being around 3 l/s above the otherwise consistent diurnal flow profile.
As can be seen in Figure 2, all three profiles occasionally show similar behaviours, rising and falling in unison. We would not expect to see the increase in flow from the inlet reflected at the critical point if the network controller was operating as intended.
Figure 2: Graph of abnormal DMA inlet flow, PRV downstream pressure, and critical point pressure
What the system highlighted for us was that there were spikes in water consumption far closer to the inlet than had been expected. While there were no logged users close to the inlet, the system gave us confidence that there were intermittent flows of relatively consistent size and duration close to the inlet that we saw no evidence of further into the DMA, and so we took to Google Earth to have a look at whether we could find any possible causes.
Satellite images showed us that the area between the inlet and the village was given over to arable farming and that there were a number of greenhouses and polytunnels close to the inlet. This led us to believe there could well be an unlogged user somewhere between the inlet and the critical point that was responsible for the spikes in flow.
As previously mentioned, Dynamo’s analysis of head loss along the 3 km of mains between the inlet and village had suggested a figure of around 10 m. This would have been significantly lower for a hypothetical unlogged user close to the inlet, leading us to our ‘lightbulb moment’.
When an unlogged user started to draw water from the system near to the inlet the network controller responded correctly, by increasing the outlet pressure of the PRV. However, it had been calibrated to adjust for the head losses experienced along the 3 km of mains pipe leading to the local village. Given the head losses between the inlet and hypothetical user would be far lower, the network controller was incorrectly attempting to overcome a 10 m head loss that was not actually present, causing the spikes in pressure seen at the critical point.
Once we had what appeared to be a solid theory supported by Dynamo’s localisation of the anomaly as well as the satellite images, we presented the case study to the client who was quick to log the customer we had identified as being the probable cause of the spikes. They identified that the customer was regularly drawing the 3 l/s that we were seeing at the inlet, confirming our theory.
The problem of maintaining a constant pressure at the critical point would appear to be a difficult one to solve on the face of it without a rezoning of the DMA. How could you program a network controller to differentiate between a sudden draw very close by from one the other side of the DMA, as both require differing responses (depending on the amount of head loss incurred between the two)?
The client was pleased with the analysis and has been looking into how best to make improvements to the configuration of the network. For us, Dynamo again demonstrated its value in determining where anomalies were occurring on a network.