Carbon-free technologies such as renewable energy assist in countering climate change, but several of them have not attained their total potential. Think of wind power, over the prior decade, wind farms have become a critical source of carbon-free electricity as the pricing of turbines have hit all time lows and adoption, as  result, is booming. Although, the variable nature of wind makes it an unpredictable energy source – less advantageous that one can dependably provided power at a particular time. 

In looking for a solution to this issue, previous year, Google and DeepMind began application of machine learning algorithms to 700 megawatt of wind power capacity in the central United States. These wind farms – an aspect of Google’s international fleet of renewable energy products, cumulatively produce as much electricity as is required by a medium-sized city. 

Leveraging a neural network that has received training on broadly available weather forecasts and previous turbine information, the DeepMind system was setup to forecast wind power output 36 hours prior to actual production. On the basis of these forecasts, the model indicates how to make optimal hourly delivery commitments to the power grid a complete day prior. This is critical, as energy sources which can be scheduled, (i.e., can provide a set amount of electricity at a particular time) are typically more valuable to the grid.  

Even though they continue to keep improving the algorithm, the leveraging of machine learning across the wind farms has generated positive outcomes. Up till date, machine learning has improved the value of wind energy by approximately 20%, in contrast to the baseline scenario of no time-based commitments to the grid. 

We cannot eradicate the variability of the wind, but the preliminary results indicate that we can leverage machine learning to make wind power adequately more foreseeable and worthwhile. This strategy also assists in bringing improved data rigor to wind farm operations, as machine learning can assist wind farm operators make intelligent, quicker, and more data-based assessments of how their power output can meet the demands of electricity. 

The hope is that this variant of machine learning strategy can fortify the business case for window power, and drive subsequent  adoption of carbon-free energy on electric grids internationally. Scientists and practitioners across the energy industry are producing novel ideas for how society can make the most of variable power sources such as solar and wind.  

Google just accomplished 100% renewable energy purchasing and is now looking to source carbon-free energy on a 24x7x365 basis. The partnership with DeepMind to produce wind power more predictable and worthwhile is a sure step toward that goal. While a lot is pending to be done, this step is a meaningful one – for Google, and more critically, for the environment.