Google’s DeepMind uses machine learning algorithms to predict output of wind farms 36 hours in advance, helping to optimize power allocated to grid

Briefing

Google’s DeepMind uses machine learning algorithms to predict output of wind farms 36 hours in advance, helping to optimize power allocated to grid

February 28, 2019

Briefing

  • Wind Power Prediction AI – Google’s DeepMind developed machine learning algorithms that can predict output of wind farms 36 hours in advance
  • Recommendation – AI can recommend or schedule optimized wind power delivery times allocated to grid one day ahead
  • AI Training – Trained AI on weather forecasts and historical turbine data
  • Testing – Applied algorithms to 700 megawatts of Google’s wind farms starting in 2018
  • Results – Algorithms boosted value of Google’s wind farms by 20%, compared to baseline scenario without time-based predictions
  • Benefits – Helps wind farms be more data driven and faster in making power allocation decisions for energy grid, reducing variability and unpredictability in wind power generation

Accelerator

Business Model and Practices

Business Model
and Practices

Sector

Energy, Utilities

Function

Operations, Manufacturing and Production

Organization

DeepMind Technologies Ltd., Google Inc.

Source

Original Publication Date

February 26, 2019

Leave a comment