A new path to commercialization? Google uses AI to improve wind power generation efficiency

Latest update time：2019-02-28

Reads：

▲Click above Leifeng.com Follow

Text | Dazhuang Travel

Report from Leiphone.com (leiphone-sz)

Leifeng.com reported that the impact of technology on human life should be subtle, and AI is constantly progressing in this direction. Today, Google announced that its wind farm has used DeepMind's AI software, which has greatly improved the efficiency of wind power generation.

Using DeepMind’s machine learning algorithms, Google has been able to predict wind output at farms, allowing it to fine-tune power output, a much smarter way for wind farms to operate than the standard, non-time-based power delivery method.

Google said that the software increased the "value" of wind power by 20% through time-based forecasting methods, but they did not clearly explain whether the so-called "value" refers to the amount of electricity generated or the value of the extra electricity produced.

At present, most of Google's wind farms are deployed in the Midwest of the United States, where data centers of technology giants are concentrated. However, the search giant did not specify which wind farm the software was used on.

Last year, Google said it had finally achieved its energy transition, with the search giant now using 100% renewable energy. This milestone was made possible by Google's massive investments in solar and wind farms, which provide a steady supply of electricity for Google's data centers.

Unlike solar energy, it is not easy to capture electricity from the wind because the power generation, storage and transmission of each wind farm are constantly changing. Whether electricity can be generated, when it can be generated, and how much electricity can be generated are all up to the mood of the weather. Google pointed out that "the ever-changing nature of wind power makes it an unpredictable form of energy, and therefore it is not as reliable as those forms of energy that can deliver energy at fixed times."

"We can't eliminate the variability of wind energy, but Google has found through practice that we can use machine learning to make wind energy predictable and more valuable." DeepMind product manager Sims Witherspoon wrote. Will Fadrhonc, director of Google's carbon-free energy project, also said in a blog post: "This approach also allows wind farm operations to have more rigorous data, because machine learning can help operators make faster and smarter assessments to determine whether the farm's power output can meet the power needs of the data center."

In fact, this is not the first time that DeepMind's AI has been behind the scenes. In 2016, Google proudly announced that with the help of DeepMind's AI, they reduced the power consumption of data centers by 15 percentage points. Last year, Google further delegated power to the AI system. In 2017, a report pointed out that DeepMind was discussing cooperation with the National Grid of the United Kingdom, and they wanted to balance the contradiction between supply and demand of electricity in the United Kingdom.

This magical operation actually created a win-win situation for Google and DeepMind.

DeepMind has always been a research pedant. They are good at spending money, but they are worried about how to get continuous cash flow. In 2017, their losses reached 368 million US dollars. Even a company as rich as Google is unwilling to continue bleeding. If DeepMind's software can get out of the scope of the laboratory and be used in actual scenarios, they can have a continuous and stable income to make up for the financial holes of their own R&D projects.

Via. The Verge

END -

◆ ◆ ◆