EERE Success Story—Solar Forecasting Gets a Boost from Watson, Accuracy Improved by 30%

October 27, 2015

You are here

IBM Improves Solar Forecasting Technology

IBM Youtube Video | Courtesy of IBM

Remember when IBM’s super computer Watson defeated Jeopardy! champions Ken Jennings and Brad Rutter? With funding from the U.S. Department of Energy SunShot Initiative, IBM researchers are using Watson-like technology to improve solar forecasting accuracy by as much as 30%.

As a part of the SunShot Initiative’s Improving Accuracy of Solar Forecasting funding program, IBM’s Watt-Sun project helps utilities, electric system operators, and solar project owners better predict when and how much solar power their systems will generate. Because solar production can be hampered by cloud coverage and bad weather, solar generation levels can be difficult to gauge on a day-to-day basis. Utilities and electric grid operators are responsible for meeting consumer electricity demands, and when solar system production varies, an accurate solar forecast is needed in order to maintain an efficient supply of energy reserves for solar customers. When utilities and electric system operators better understand generation patterns, they’re able to maximize solar resources, operate more efficiently, and improve solar energy’s economic competitiveness.

Researchers at IBM’s Thomas Watson Research Center believe that machine-learning technology and big data analytics can be used to develop more accurate solar forecasting solutions. The research team, led by Dr. Hendrik Hamann, developed a solution called the self-learning weather model and renewable forecasting technology (SMT). The innovative machine-learning platform synthesizes an incredible amount of data from various sources, including historical weather data and real-time measurements from local weather stations, sensor networks, satellites, and sky imagers. With added processing power from the Department of Energy’s computing facilities, SMT uses the data to pinpoint which weather prediction models are the most accurate and yield forecasts that can be used from minutes to months in advance. Utilities and electric system operators can then use this information to anticipate solar generation levels and better match consumer electricity demands.

The impact of the machine-learning platform is huge. IBM plans to make U.S. solar forecasts available to government agencies and other organizations in an effort to spur solar adoption and help utilities understand how to integrate more solar into their generation portfolios. IBM’s work doesn’t stop there – researchers have also been exploring ways to use this machine-learning platform to aid wind and hydropower forecasting. With applications beyond solar, IBM’s solar forecasting research will remain relevant in the renewable energy industry for years to come.  

Learn more about the Systems Integration projects funded by the SunShot Initiative. 

The Office of Energy Efficiency and Renewable Energy (EERE) success stories highlight the positive impact of its work with businesses, industry partners, universities, research labs, and other entities.