Home / Software Posts / Can AMI Data Help with Modeling?

Can AMI Data Help with Modeling?

Tom Walski Profile Image

Tom Walski, Ph.D, P.E, Senior Product Manager, Water

Big data binary visualization cloud computing internet of things IOT
#image_title

Share

One of the sources of error in hydraulic modeling is the difficulty of knowing water demands with any real precision. Yet demand is what drives models. Over the years, modelers have developed a range of methods for taking limited demand data and developing well-calibrated models. For example, OpenFlows WaterGEMS’s Load Builder has 12 different methods for loading demands based primarily on the source of data.

There are two dimensions for demand loading in models—spatial and temporal. Spatial tells you where the demand is, and temporal tells you how it varies over time. The tools for dealing with each dimension have generally differed.

The coming of Automated Meter Reading (AMR) with knowledge of the exact x-y coordinate of the meter greatly improved the spatial allocation of demands in a model. However, these meters were generally read once per billing cycle and shed very little light on temporal variations other than seasonal trends. Temporal issues are usually handled at the level of resolution of the SCADA system. Demand patterns tend to be average demand patterns.

This past year, Stephen Jackson and I from Bentley teamed up with some folks from Sensus meters and the City of Walla Walla, Washington, for a study on the application of AMI data to modeling. We just had a paper published in the December issue of Journal AWWA (Jackson et al., “Using AMI Data for Water Distribution Modeling,” Vol. 113, No. 10, p. 44-56). We showed that we could more precisely locate demands but more importantly, we could develop demand patterns at a range of different resolutions such as by node, by type of user, by day of the week, by season, or by pressure zone.

In Walla Walla, the demand patterns were very different between summer and winter. In the summer, the peak demands occurred before sunrise when irrigation was the more prevalent use.

Seasonal Multipliers Graph

It also gave us an improved estimate of non-revenue water (NRW). We didn’t need to wait until monthly billing was complete but could estimate NRW in near real-time. Comparing water supply flows from SCADA with water demand at customers is shown in the figure below.

January Mass Balance

We were able to investigate the question, “If I want to know the demand pattern for Wednesday, is it better to use data from the previous day or from the previous Wednesday?” The answer is that the previous Wednesday is a better estimate, but they are both similar.

Some of the things we can do with this data are overkill. Given that for hydraulic calculations, demands are lumped at nodes, it is not important if the resident is flushing his toilet at 6 Maple St. or 8 Maple St. at 9:15 am. In addition, since velocity and hydraulic gradient are low in most residential neighborhoods, the improvements in estimating pressure at any house is not impacted much by doing a more accurate estimate of demands using AMI.

The place where AMI data can be most useful is understanding the demands at large commercial, industrial, and irrigation customers. These customers can be separated from the average residential customers to give a better analysis of their impacts. The graph below shows that for some large users, using an average demand pattern can miss individual peaks that occur at somewhat random times.

January Large Customer AMI

This AMI demand can help the modeler virtually eliminate demands as a source of error in model calibration. The finer the resolution of demands, the better the calibration can capture variations in pressure, as shown below.

Junction Pressure Validation

High-quality AMI data can improve modeling, but the use of AMI data to instantaneously adjust models in real-time is still limited. First, meters are only polled every x hours to save battery life so that there is a time lag between the reading and the data being available. Second, even if the demand at a node maybe 20% higher than expected at 1:00 pm, does that mean that the demand will be 20% higher at 7:00 pm, or will the demand return to the normal pattern by 7?

We’ve developed ways that can make it possible to leverage AMI data to improve your demand inputs into your OpenFlows WaterGEMS model. As more users adopt AMI systems, we’ll be working to better leverage that information.

If you want to look up past blogs, go to https://blog.bentley.com/category/hydraulics-and-hydrology/.  And if you want to contact me (Tom), you can email tom.walski@bentley.com.

Want to learn more from our resident water and wastewater expert? Join the Dr. Tom Walski Newsletter today!

Relevant Tags

I didn’t get much response to the question I posed in a recent blog about the best placement of isolation ...

We all know the meanings of the R words listed in the title of this blog (at least we think ...

Last year, I wrote a blog on “How many valves are enough?” I’m following up on that now with a ...