Imagine getting a report that predicts weather conditions right outside your front door every 15 minutes. That’s happening in China, at least on wind farms where IBM is deploying technology that crunches gargantuan data flows from sensors, satellites and other sources to give granular forecasts for wind conditions at individual turbines.
As China, the US and other countries ramp up electricity production from renewable but intermittent sources like wind and solar farms, such forecasting is becoming crucial to integrating green energy into power grids. If you can forecast how windy or sunny the weather will be at a wind farm or solar power plant, you can predict how much electricity— and revenue—will be generated. And that allows utilities to avoid firing up expensive and carbon-polluting fossil fuel plants to plug the gaps in supply when a turbine farm suddenly goes slack.
“The reason that power forecasts are poor is that weather forecasts are poor,” Lloyd Treinish, chief scientist for IBM’s Deep Thunder weather forecasting project, told Quartz. “The wind coming into a wind farm is not uniform but is highly variable and if you’re looking at things only in an aggregate way that could lead to inaccurate forecasts.”
That’s a problem particularly in China, where a failure to know which way the wind is blowing has resulted in state-owned utilities letting renewable energy projects go idle, if they aren’t confident about how much electricity they are predicted to generate.
In fact, it was demand from China—which has mandated a massive expansion of renewable energy production—that prompted IBM’s weather wizards to develop the new technology, called Hybrid Renewable Energy Forecasting (HyRef).
A subsidiary of the State Grid Corporation of China, the country’s biggest government-owned utility, approached IBM about developing the technology. The company has installed HyRef at a 670-megawatt solar and wind farm with the goal of boosting by 10% the amount of electricity that can be integrated into the grid from the project.
Here’s how it works. Sensors are attached to the nacelles (enclosures for the wind turbine’s generator) of individual wind turbines that measure wind speed and direction. A small supercomputer then analyzes that data along with weather information gathered from proprietary and government sources to forecast wind conditions, and thus electricity generation, over time. The forecasts range from 15 minutes to one month into the future.
At solar power plants, video cameras scan the skies to track clouds and filter the data through imaging software to predict the amount of solar radiance that will hit photovoltaic panels.
HyRef also aims to predict rainstorms, high temperatures and other meteorological conditions that affect electricity production. For instance, while wind is as good as gold for turbine farms it can cause the frames around solar panels to vibrate, reducing electricity production.
Treinish says IBM has begun to sell HyRef in Europe and the US. While the Chinese prefer to purchase the entire system and operate it themselves, he expects other customers will prefer to let Big Blue do the work and buy the data. Potential clients include operators of big solar and wind farms whose profits depend on hitting production targets for their utility company customers.
HeyRef can also lower operating costs for wind farm operators, according to Treinish, by allowing them to schedule turbine maintenance when power production is predicted to fall off.