New capabilities are coming soon to the GasDay modeling software
MILWAUKEE—According to developers, approximately 35 natural gas companies nationwide regularly rely on GasDay software, built by researchers and engineers at Marquette University, to analyze demand and guide natural gas flow on a daily basis.
The software, explains those researchers, evaluates weather and market data to provide a “point forecast”—a single number representing how much natural gas customers will need. In 2016, Peoples Gas, one of the first companies to test the software during its creation, asked GasDay researchers if they could provide even more information. It would be helpful, the natural gas controllers noted, to see the total range of possibilities in addition to the most likely forecast.
Challenge accepted.
Enter Mohammad Saber. A doctoral student in electrical and computer engineering, Saber was inspired to integrate probabilistic forecasting into the GasDay algorithm. Probabilistic forecasting incorporates uncertainty into its results, something the human brain does many times a day. “In everyday life, we think in terms of probabilistic forecasting,” Saber said. “Maybe you’re thinking about how long it will take to get home from work. You don’t have any point forecasting in mind. If the traffic is heavy, you’ll leave earlier.”
For GasDay, probabilistic forecasting asks, “What is the likelihood that demand will fall within this range versus that one?”
This method has rarely been used in the energy industry because of its inherent complexity. But Saber realized that if GasDay could show probabilistic results, it would present a more complete view of the risks that could lead to over- or under-supply, which could help natural gas companies reduce costs and better manage resources.
What happens naturally in the brain requires serious work when it comes to programming software. When Saber chose this topic for his doctoral dissertation, he set out to develop not only new ways of generating probabilistic forecasts for the energy industry, but also new methods of evaluating and communicating those forecasts. He began implementing his research at the GasDay Laboratory early last year, and he successfully defended his dissertation in September. The next step is integrating the new forecasting functionality into the software itself.
An enhanced version of GasDay’s forecasting tools is under development to help natural gas controllers as they make cost-effective choices and plan for any uncertainties ahead.