DOE system detects natural and malicious water contamination

Clean drinking water is something most people in the developed world take for granted. But even in the most advanced countries, public water utilities have been contaminated  by chemical or biological substances. And after the 2001 terrorist attacks in the United States, the specter of water utilities being deliberately poisoned especially began to haunt security planners.

But while utility companies have a variety of sensors to test water quality, what was lacking was a system that could read data in real time and alert utilities within minutes or hours of contamination. That’s the added benefit of Canary, an open-source software system developed by the Energy Department’s Sandia National Laboratories in partnership with the Environmental Protection Agency.

Sandia researchers began working on the software’s algorithms in 2005, but work on the software itself did not begin until 2006, said Sean McKenna, lead research of the Sandia team, the same year Canary was launched as a pilot project for the greater Cincinnati water works.


Related coverage:

GIS app gives first responders in Georgia a moving picture


The software first ran live in Singapore and Cincinnati in 2009. Another major Canary user is the Metropolitan Water District of southern California, the largest water utility in the world, serving 19 million customers. The city of Philadelphia is also currently testing the software.

How it works

Canary is designed for utilities that have made the investment in real-time sensor systems, McKenna said. With the larger utilities, sensors are connected to supervisory control and data acquisition (SCADA) systems — the industrial pumps and sensors that make up the utilities’ infrastructure. The software collects and uses data from a utility’s SCADA database to create a water quality picture.

Canary runs on desktop computers. It can be downloaded and it is customizable for individual water utilities.

Canary operates by detecting major deviations in water quality. The software reports in two to five minute intervals over a day-and-a-half to two-day period. Each of these time segments lists a variety of sensor data readings, including chlorine and pH levels.

When new data is collected, Canary determines if it falls into a normal water quality range. If the software detects an anomaly, it will make note of it. If the anomaly continues over several minutes, the system will calculate whether the situation  constitutes a “contamination event” and if so will alert the operators, McKenna said.

Canary is not designed for forensic analysis, said David Heart, the team’s lead software developer. But Sandia is working with the EPA to use software models and tools to backtrack and locate sources of a contamination, to find ways to limit the impact of contamination and to clean it up. Researchers are also looking at ways that Canary can interact with these forensics and analysis tools, he said.

Besides measuring immediate water quality, Canary can also track trends over time, such as monthly and seasonal shifts in pH or chemical composition. But water quality data varies over time, especially as utilities change their equipment and water systems, noted McKenna. For immediate water quality monitoring, data from the previous 12 to 24 hours is the most relevant, McKenna said.

Open technology

Prior to Canary, there was no real-time water quality monitoring capability, McKenna said. Vendor firms sold sensors that could read data in real time, but utilities lacked a tool that would permit them to create a data picture from multiple sensors made by different vendors.

Many commercial firms that make industrial and utility equipment have traditionally used proprietary software, which limits the scope of monitoring. McKenna said utilities need to adopt a more flexible, open standard for the technology because it would allow them to work across systems and jurisdictions on water quality issues.

There are other commercial real-time water monitoring systems, but most of these are made by the hardware and sensor manufacturers, Heart said. Canary is not tied to any particular type of hardware system; it will collect data from any type or make of water monitoring device, he said.

One of the main challenges faced by the Sandia researchers was pulling data out of utility databases in a usable format. Although commercial databases are relatively interoperable with Java database connection drivers, Heart noted that some utilities still use homegrown databases. For a database operating on a commercial system, Canary can access and track new data within 30 seconds to provide almost real-time monitoring, he said.

The most recent upgrade of Canary is Version 4.3, which can match patterns in chemical and other contaminant types and compare them against normal changes in water quality. The new version can also create new SCADA signals on the fly, which allows operators to track pump data and integrate operational data into water quality observations.

Reader Comments

Please post your comments here. Comments are moderated, so they may not appear immediately after submitting. We will not post comments that we consider abusive or off-topic.

Please type the letters/numbers you see above