Babbage | Electricity forecasts

Tilting towards windmills

IBM has come up with a better way to forecast wind power

By H.G. | SEATTLE

THE 1996 Olympics in Atlanta did not all go IBM's way. For all its technical prowess, the computer giant managed to bungle the reporting of some competition results. On the plus side, it was at the Games that IBM first deployed Deep Thunder, a novel computer model which warned the organisers when and where to expect inclement weather—and correctly predicted that a thunderstorm forecast by other meteorologists would not affect the closing ceremony. Deep Thunder has since gone through countless iterations, or which the latest, called the Hybrid Renewable Energy Forecaster (HyREF) IBM unveiled on August 12th.

As its name suggests, HyREF is meant to make it easier to incorporate wind energy into the grid. Owing to Aeolian vagaries, it is hard for operators of wind farms to forecast output accurately—or indeed to work out where best to erect turbines in the first place. The ability to predict where wind will blow and how hard is therefore crucial if wind power is to live up to its boosters' hopes.

IBM's system increases this all-important predictability using a handful of sophisticated technologies. Clever sensors mounted on individual turbines gauge wind speed, temperature and direction. Their readings are combined with data from traditional measurement towers equipped with meteorological instruments, as well as past-weather data. Indeed, Brad Gammons, who runs IBM's energy and utilities arm, says that most of the progress since Deep Thunder has taken place over the last two years, mainly thanks to the rapidly growing availability of information, both real-time and historical. In particular, Mr Gammons says, this is true for China, the world's biggest greenhouse-gas emitter, but also its biggest investor in renewable energy.

Perhaps unsurprisingly, then, it was in China that IBM conducted HyREF's first big field test. The results from a big renewables demonstration project in Zhangbei are encouraging. According to IBM, its system predicted the power output (more important to wind-farm operators than weather) for up to four hours in advance with 94% accuracy, compared with no more than 80% for existing methods. HyRef also updates the forecast every 15 minutes, much quicker than traditional models. It has also improved the accuracy of 48-hour forecasts and, most impressive, perhaps, offered forecasts for overall power output a whole month in advance. The upshot, IBM boasts, was that HyREF increased the amount of renewable energy generated at Zhangbei that was integrated into the grid by a tenth, or enough to power 14,000 homes. Little wonder, then, that countries which have made big bets on wind energy, such as Denmark, Germany and America, are interested.

For HyREF really to live up to its name, though, it needs to do more than foretell breezes. So IBM is now extending it to solar energy, using sky-facing cameras, called nephographs, which record cloud movement, and thus cloud cover. Even then, of course, no amount of widgets will be enough to ensure a renewable-energy future in the absence of political will. That is something not even IBM's clever gizmos can predict.

Discover more

And it’s goodnight from us

Why 10, not 9, is better than 8

For Microsoft, Windows 10 is both the end of the line and a new beginning


Future, imperfect and tense

Deadlines in the future are more likely to be met if they are linked to the mind's slippery notions of the present