Does your utility use all available data for demand forecasting?

The data science community is in agreement – the more data the better for short and long-term forecasting.

But organizing water network data in spreadsheets is difficult due to Excel’s 1 million row limit. And it leads to only subsets of data being analyzed.

In contrast, engineers who use SensorClean can easily leverage all 20+ years of monitoring data for forecasting.

Using all data for demand forecasting can:

• reduce utility risk

• save on capital expenditure

• improve operational efficiency; and

• benefit maintenance protocols.

Does using all data really matter?

The data history shows how the system responds to changing population, behavior and climate change. It also quantifies the impact on consumption of wet and dry periods.

Analyzing subsets of historical data is like viewing the iceberg from your ship and missing the important context underneath.

For example, historical extreme demand events can be missed when forecasting for new infrastructure with significant implications.

Using all data reduces utility risk and brings confidence to design and more economic outcomes.

And short-term forecasting for operational purposes will always be better with more data.

How does SensorClean help engineers?

Engineers who use SensorClean can quickly clean and organize the data history for water balances in many DMAs (District Metered Areas).

The trends in the DMA’s reveal the story of changing water behaviors across the greater water system.

To analyze the DMA below, 23 million rows of data from 2 sensors were processed.

In 1.5 hours, the engineer produced key forecasting and leakage metrics.

Figure 1 Daily water consumption with daily max temperature from SensorClean. Note that 2012 was a wet year, leading to a drop in DMA water consumption.

Peak week demand events in droughts play an important role in forecasting.

And more frequent droughts can be expected with climate change.

Engineers can quickly access peak weeks based on the entire history with SensorClean.

Figure 2 Peak week demand events from SensorClean

What’s next?

Now’s the time to leverage historical utility network data for a range of forecasting benefits.

We’re here to help you on this path - please contact us.