Machine Learning Boosts Wind Farm Efficiency

Research is going on around the world when it comes to making renewable energy more efficient. Wind power is one of the green energy generation methods that are being worked on, and wind farms are a big part of the carbon-free energy generation scheme. But they come with challenges.

A big challenge for wind farm operators is that the variable nature of wind makes this type of energy generation unpredictable. To help combat that unpredictability, DeepMind and Google have applied deep learning algorithms to 700 megawatts of wind power in the central US.

The wind farms are part of the Google fleet of renewable energy projects and can generate enough electricity to power a medium-size city. Google used its neural network trained on widely available weather forecasts and historical turbine data to configure DeepMind to predict wind power output for 36 hours ahead of actual power generation.

These predictions allow the model to recommend optimal hourly delivery commitments to the power grid a day in advance. That allows energy resources to be scheduled making wind power more valuable to the grid. Google notes that it is continuing to refine the algorithm it uses in the system.

So far machine learning has boosted the value of Google's wind energy by as much as 20% compared to a scenario where no time-based commitments to the grid are made. Google says it can't eliminate wind energy variability, but it can use machine learning to make wind power more predictable and valuable.