Decision support systems at water utilities are evolving with new technology and machine learning. Luis Montestruque, Vice President, Digital Solutions at Xylem, shares his insights on how digital twins and decision intelligence are helping utilities better understand their infrastructure so they can make smarter decisions.
As weather becomes more unpredictable, large storms can wreak havoc on water infrastructure systems, causing overflows and flooding. Despite controls and wet weather facilities, storms still hit cities and water utilities in ways that older hydraulic models cannot predict. Today, however, there are better ways to use data to guide how systems react to storms.
“Five years ago, artificial intelligence and machine learning started to appear as a viable technology for water utilities, and a lot has happened since then,” says Montestruque. “However, there haven’t been good frameworks to connect these new technologies so that utilities can benefit from their full value.”
Montestruque compares the evolution of decision support systems in water utilities to the development of smart technology in cars.
“Not too long ago, GPS-aided navigational systems became mainstream,” he says. “Nowadays, it’s almost impossible to buy a car without it. As a matter of fact, if you buy a mid-priced car it is very likely that it is able to steer you back into your lane if you start drifting. Many cars are now able to completely drive by themselves. This will likely become a standard feature in a few years. These systems fall in the category of decision support systems.”
Decision support systems (DSS) are computerized systems that help people make decisions to achieve a certain objective. In some cases, they are able to do so automatically. DSS are useful because they are able to account for a multitude of variables and can help people achieve objectives faster and cheaper. They do so by using a simple framework: sense – predict – act.
First, the DSS utilizes sensor data or data from other systems to determine the current state of the target system. Second, the DSS utilizes a model to predict possible outcomes under a variety of operational strategies. And third, DSS use an analytical engine to search for the optimal strategy that achieves desired objectives.
“For at least a decade now, we have implemented decision support systems in sewers to avoid overflows and flooding, in wastewater treatment plants to reduce energy consumption, and in drinking water networks to find leaks and reduce energy consumption,” Montestruque says.
“What has changed now is that new technologies have helped lower the cost of implementing decision support systems. For example, “Internet of Things” technologies enable companies to economically deploy hundreds of battery-operated sensors in difficult locations.”
“One technology, however, has truly transformed the way decision support systems are implemented,” Montestruque says. “It is called digital twins, and it is also not new. Digital twins are digital representations of the infrastructure. We have known them for a long time as models, and they represent the way we understand how our infrastructure works.”
Traditionally, digital twins have been built using what are called “first principles.” These first principles models utilize physical, chemical, or biological equations to simulate the infrastructure. Building and calibrating first principle models can be a long and expensive process. Decision support systems built based on first principle models are efficient but are unable to learn from past experiences unless someone calibrates their models.
“Recently, we have been augmenting our digital twins with machine learning, enabling them to continuously learn,” Montestruque says. “This is important because these digital twins are able to utilize past data and automatically ‘calibrate’ to better represent the infrastructure. This typically means significant reductions in model building costs, with a resulting digital twin that is highly and continuously accurate.”
Digital twins that perform best combine first principles models in areas with low uncertainty, such as pipe hydraulics, with machine learning in areas with higher uncertainty, such as system hydrology in sewers.
“It is when we combine decision support systems with digital twins that we are able to get something unique: a decision support system that is able to learn and continuously adapt,” Montestruque says.
“We call this new paradigm shift ‘decision intelligence.’ For example, with decision intelligence we are able to operate a wastewater treatment plant by synthesizing the last decade of operational history in a digital twin. And every time there is a new event, the system learns more and becomes more accurate. Now when a storm starts brewing on the horizon, a city’s decision intelligence can automatically start searching through thousands of possible strategies. It has probably seen something similar before.”
Using decision intelligence to operate their sewer systems, cities can reduce overflows and flooding at a fraction of the cost of traditional solutions, making a lasting difference to communities around the world.
Get the latest news and updates from Making Waves.
Making Waves is news from Xylem, a leading global water technology company. Xylem's solutions include products and services that move, treat, analyze and monitor water.
Install for this language not supported at the moment. Please try again later.