Many of us start our morning by checking the weather forecast to prepare for the day ahead. But sometimes, the weather can change rapidly. Grid operators face similar challenges: they rely on forecasts to get a general idea of expected electricity demand, as well as how much energy supply will be available to power the grid during a given timeframe.
For solar and wind energy, most grid operators rely on deterministic forecasting to predict the power they can supply a few days or hours in advance. In 2021, Texas electricity grid operator, Electric Reliability Council of Texas or ERCOT, started experimenting with a new forecasting method called probabilistic forecasting, which includes detailed information about solar forecast uncertainty. This was made possible by the research funded by the U.S. Department of Energy (DOE) Solar Energy Technologies Office (SETO) under the Solar Forecasting 2 Funding Program. By adopting the probabilistic forecasting method, ERCOT is better positioned to integrate more renewable energy generation with confidence and improved reliability.
With a $1.7 million SETO award, the National Renewable Energy Laboratory (NREL) and Maxar developed advanced, data-driven probabilistic forecasting techniques that improve the accuracy of solar power generation forecasts through state-of-the-art predictive modeling and an in-depth understanding of power systems and markets. These forecasts predict how much solar energy will be produced during a variety of time frames, ranging from minutes, hours, days, and weeks ahead.
Maxar created these forecasts by using innovative approaches to quantify uncertainty and overcome the limitations in traditional weather models, while leveraging innovations in cloud computing. Produced on the fastest forecasting cadence Maxar offers, this probabilistic forecast information is delivered via application programming interface (API) and allows grid operators to more efficiently balance energy generation from solar, wind, and other sources while achieving economic savings and improved reliability.
To understand the real-world benefits of these new forecasting approaches, Maxar partnered with ERCOT to gain the end user’s perspective on data and grid reliability requirements. ERCOT manages the flow of electric power to 26 million Texas customers, representing about 90 percent of the state’s electric load. While ERCOT already used Maxar’s standard forecast solutions for grid operations prior to the DOE project, the success of this project’s efforts resulted in ERCOT having visibility into Maxar’s probabilistic solar power forecasts on a sub-hourly basis in their operational systems. By integrating probabilistic solar forecasts into power system operations, ERCOT can integrate more solar power into its grid, reducing system operating costs, and improving overall grid reliability.
You might wonder why probabilistic forecasting provides such beneficial outcomes compared to deterministic forecasting. A deterministic forecast might predict no clouds tomorrow at 3 p.m., whereas a probabilistic forecast will provide a more detailed prediction—there is a 30 percent chance of cloud cover tomorrow at 3 p.m. That additional level of detail on the forecast uncertainty enables grid operators to better manage the operation costs and risks when dispatching generation resources.
So why doesn’t every grid operator use probabilistic forecasting? Because it requires significant engineering resources to create, calibrate, deploy, and use. Further, while probabilistic forecasts can optimize decision making, they also require sifting through significantly more data than traditional forecasts, so many operators have been slow to apply this forecasting to their operations. A 30 percent chance of clouds might seem like a trivial detail, but using that additional context requires thoughtful coordination and advanced energy management systems. ERCOT, Maxar, and NREL, through the DOE project, have demonstrated that probabilistic forecasting provides a better tool for grid operators to optimally utilize all sources of electricity and improve the reliability of our nation’s grid.
In 2023, SETO launched the American-Made Net Load Forecasting Prize to increase adoption of the state-of-the-art in probabilistic net load forecasting. The prize builds on the Solar Forecasting Prize from 2022, which incentivized solar forecast providers to develop and potentially commercialize tools that generate probabilistic solar irradiance forecasts, and the Solar Forecasting 2 funding program, which, among other things, developed probabilistic forecasts that enable grid operators to calculate how much reserve power is necessary at any given time to maintain a healthy grid.
Learn more about SETO’s system integration research.