Better Weather Analysis Could Lead to Cheaper Renewables
Predictive analytics can lower the costs associated with connecting wind and solar to the grid, says IBM.
Because renewable power is often intermittent, utilities typically need to use fossil fuel plants as backups.
Because the output from wind and solar power plants varies, they need backup—either fossil fuel plants or energy storage—to compensate for dips and spikes. But it’s rarely clear just how much the output will vary, so that backup power is often on standby even when it’s not needed.
Now IBM has developed software to address this problem. The software performs advanced data analysis that IBM hopes can improve predictions of renewables’ power output, and thus reduce the need for backup power. Using multiple data sources, including wind turbine sensors, weather forecasts, and images of clouds, the software can forecast power output as little as 15 minutes and as much as a month in advance. It’s now operating at a combined solar and wind demonstration project in Zhangbei, China.
Better data and predictive models could make renewable energy more valuable to utilities by reducing the need for excessive backup power. “If you could predict the wind output 15 minutes ahead, you could reduce that capacity requirement, which would reduce costs,” says Erik Ela, a senior engineer at the National Renewable Energy Laboratory.
If a plant’s operators could more accurately forecast the output of renewable power sources, they’d have less reason to rely on energy storage, which is typically needed now to provide a smooth flow of power into the transmission grid. “In the industry, storage is seen as the next disruptive technology,” says Michael Valocchi, vice president in IBM’s energy and utilities consulting business. “(But) if I can really predict in this manner, it’s not that I don’t need storage, but it makes storage less important.”
Utilities often rely on specialized companies to produce wind and solar forecasts based on weather models and other meteorological data, including anemometers on wind turbines. But wind measurements taken from turbines are often unreliable because energy has already been extracted from the incoming wind, and because vibrations affect readings, says IBM researcher Lloyd Treinish, the chief scientist of IBM’s weather modeling system. For its project in China, IBM analyzed data from all the turbines to come up with a more accurate representation of actual wind speed and direction, he says.
IBM also built a meteorological model specific to this site in northern China and installed video cameras to track the movements of clouds to inform solar forecasts. The entire data set is fed into a supercomputer to generate the forecasts.
There are a number of other efforts to improve weather forecasting via better data collection and analysis (see “Sharper Computer Models Clear the Way for More Wind Power”). The latest generation of wind turbines from General Electric, for example, features a control system designed to better predict power output by analyzing tens of thousands of sensor data points a second (see “Wind Turbines, Batteries Included, Can Keep Power Supply Stable”). The U.S. Department of Energy has funded a few research projects, including one at the University of California, San Diego, to capture images of clouds with special devices featuring fish-eye cameras. The project analyzes those images with algorithms to produce a prediction of how much solar power a plant might produce over the next 15 minutes.
Ela notes that while utilities’ forecasts of electricity demand have become sophisticated over the decades, their forecasts of the supply of that power are still immature.