Google, DeepMind Target Machine Learning at Boosting Wind Energy Value

Wind energy adoption is further along than you might think: Iowa, for example, is expected to generate about half its electrical power from wind within a few years; in total, according to the U.S. Department of Energy, about 90 gigawatts of wind power has been installed nationwide. It’s not just plentiful wind but also wind consistency that’s a prime factor in the success of wind turbine installations. While Iowa and other Great Plains states are relatively ideal for this renewable, other locales must account for periods of calm while demand for electricity remains constant.

That’s the problem taken on by Google and Alphabet sister firm DeepMind, a London-based AI R&D company. They’re using machine learning to develop better wind forecasts that will boost the value – and, it’s hoped, the proliferation – of clean, carbon-free wind energy.

In a blog post yesterday, Google and DeepMind said that last year the two companies started applying machine learning algorithms to Google’s 700 megawatts of wind power capacity in the central U.S., enough electricity needed by a medium-sized city.

“Using a neural network trained on widely available weather forecasts and historical turbine data,” wrote the authors*, “we configured the DeepMind system to predict wind power output 36 hours ahead of actual generation. Based on these predictions, our model recommends how to make optimal hourly delivery commitments to the power grid a full day in advance. This is important, because energy sources that can be scheduled (i.e. can deliver a set amount of electricity at a set time) are often more valuable to the grid.”

Witherspoon and Fadrhonc said initial feedback has been encouraging – “machine learning has boosted the value of our wind energy by roughly 20 percent, compared to the baseline scenario of no time-based commitments to the grid.”

“We can’t eliminate the variability of the wind,” they said, “but our early results suggest that we can use machine learning to make wind power sufficiently more predictable and valuable. This approach also helps bring greater data rigor to wind farm operations, as machine learning can help wind farm operators make smarter, faster and more data-driven assessments of how their power output can meet electricity demand.”

Google announced last April it had achieved 100 percent renewable energy purchasing and claims that it’s the world’s largest corporate buyer of renewables. Now the company is striving to source carbonfree energy on a 24×7 basis, which means sourcing enough carbon-free energy to match its electricity consumption in all places, at all times.

“Such an approach looks markedly different from the status quo, which, despite our large-scale procurement of renewables, still involves carbon-based power,” the company said in a white paper released last October. “Each Google facility is connected to its regional power grid just like any other electricity consumer; the power mix in each region usually includes some carbon-free resources (e.g. wind, solar, hydro, nuclear), but also carbon-based resources like coal, natural gas, and oil. Accordingly, we rely on those carbon-based resources — particularly when wind speeds or sunlight fade” (see above).

* Sims Witherspoon, program manager, DeepMind; Will Fadrhonc, Carbon Free Energy Program lead, Google