New Hampshire, USA — IBM has decloaked a system combining weather modeling, sensors, and analytics to better forecast a wind farm’s output from minutes to weeks away, helping utilities and grid operators more confidently integrate an intermittent energy source.
IBM’s “Hybrid Renewable Energy Forecasting” (HyRef) system combines weather modeling capabilities, sky-gazing cameras and imaging technology, sensors on the turbines, and sorts it all through analytics software to predict incoming weather patterns and calculate wind turbine performance, from 15-minute intervals up to 30 days in advance. The upshot is more efficiency and reliability from wind farms, and reduced intermittency of delivery into the grid.
Michael Valocchi, VP in IBM’s Global Energy and Utilities Industry business, talked us through the HyRef’s origins and applications. Essentially HyRef folds in several key components:
- Weather modeling capabilities that have been fine-tuned “to a much more granular basis,” from a smaller area perspective within a square kilometer to vertical heights where turbine hubs and rotors are located;
- Advanced imaging technology, from cloud imaging to sky-facing cameras;
- Sensors on the turbines to monitor wind speed, turbulence, temperature, and direction; and
- Analytics capabilities to collect and manage the data, both structured and “unstructured” (data that can’t be natively stored in a spreadsheet; images, possibly multimedia, etc.). SAS provides the backbone for this, on a DB2 platform.
- Here’s an infographic explaining all that in pleasing pictorial form. And here’s a video for a more interactive view of the wind intermittency problem and HyRef’s solution.
Finally, check out the Image Gallery at right to see a customer’s dashboard view of data.
HyRef had its origins in a late-1990s project dubbed “Deep Thunder” to produce weather modeling and forecasts to anticipate major weather events. Two years ago the work took a different turn when Vestas hired IBM to help optimize wind turbine placement at facilities in Denmark, analyzing “petabytes of data” — weather reports, tidal phases, geospatial and sensor data, satellite images, deforestation maps, and weather modeling — to shave the process from several weeks to less than an hour, according to Valocchi.
Fast-forward to today, and those efforts in weather forecasting and analytics are intersecting to provide “a more holistic prediction capability” for a wind farm’s operation and output, Valocchi explains. HyRef “looks at the output of the set of turbines — not just ‘is it cloudy?’ or ‘is it windy today?’ — pulling together thousands of data points to come to that level of predictability,” he said. “It’s not just a high-level weather forecast.”
Other firms provide wind forecasting services, and some are seeking to improve upon those efforts. The National Center for Atmospheric Research (NCAR) and Xcel Energy recently extended a partnership in developing real-time wind analysis to focus on improving site-specific accuracy of “ramp” events, from passing fronts to icing dangers of freezing rain and fog. (NCAR is exploring solar energy forecasting as well.) GE’s new “brilliant” wind turbines equip sensors, energy storage (100-350kW) and algorithms to predict power output in 15-30 minute stages. Even employing supercomputer muscle to wind forecasting isn’t new — Iberdrola and Spain’s National Renewable Energy Center (CENER) recently pledged to apply teraflops of computing power to more accurately forecast onshore and offshore wind production.
Those other efforts, however, are still in the initial testing stages or being developed over the next couple of years, and in the case of NCAR/Xcel aiming for “probabilistic” estimates of weather events anywhere from 20-80 percent. Meanwhile, IBM’s HyRef is running right now at a customer site: the Chinese State Grid Corp.’s 670-MW Zhangbei “demonstration” project, home to 500 MW of wind, on top of 100-MW of solar and a 70-MW battery storage system. And it’s proven to deliver 92 percent accuracy.
Here a snapshot of Zhangbei’s HyRef dashboard, provided by IBM. Again, more screenshots are in the image gallery, see above & to the right.
Zhangbei also wants to apply the same forecasting abilities to its solar output; “obviously there are different things to look at” with solar forecasting, he noted, “but absolutely it’s meant for not just wind.”
HyRef, IBM claims, is helping the Zhangbei project better manage its wind output, reduce the need for curtailment, and increase the integration of renewable power generation by 10 percent, according to IBM. “It’s better operations from a grid perspective, increasing output, better predictability,” Velocchi said.