Mon compte person ENGIE

My notifications

News See all the news
Using AI and Weather Forecast To Optimize Renewable Energy Output
New energies 30/06/2020

Using AI and Weather Forecast To Optimize Renewable Energy Output

Will it be sunny tomorrow? And the wind: from what direction and what speed will it blow? These are questions typically fit for planning a weekend escapade or for a fisherman or farmer. But they are also key to maximizing the value of renewable energy.

Through a combination of AI, computing and more accurate weather forecasts, a range of studies have granted access to more precise wind and sun power data

As we aim to accelerate the transition towards a carbon-neutral economy, the performance of such solutions as wind turbines and solar power is a major innovation challenge to achieve a sustainable future. 

We know by now that these natural elements that technology turns into energy can be intrinsically inefficient. By their nature, they are intermittent, bound to never be available 24/7, year-round: as the wind stills and the sun comes and goes. If we aren’t able to predict how much energy a plant will be able to produce, it is harder to make the most of this energy and make it reliable enough to supplant more polluting, but consistent, sources.

Fortunately, recent developments are enabling scientists and businesses to optimise the use and production of renewables. Through a combination of Artificial Intelligence, computing and more accurate weather forecasts, a range of studies have granted access to more precise wind and sun power data, fuelling a flurry of projects to improve the efficiency of renewables. Here’s a taste of some that are worth tracking:


At Google, energy is in high demand. Its portfolio uses as much power as two times the city of San Francisco, according to Forbes, and the company has long tried to pivot to renewable energy. 

  • In a recent project, the company has crunched weather data and power data from 700 megawatts of wind energy sourced in the central United States, and then used machine learning to predict wind production more accurately. The company gained better insight into how much power wind turbines will produce — 36 hours in advance.
  • More recently, Google has started to schedule its endless data-crunching tasks around green energy production patterns. Basic services that users require all the time — like Search, Maps and YouTube — continue running around the clock. But the company has started to align the timing of many other computing tasks with times of high supply of solar and wind power, thus making greater use of renewable energy.


Wind power is surging around the world — growing 19% in 2019 alone, thanks to several major projects in the US and China. 

  • In the state of New York, Cornell University scientists produced the first detailed study of how wind power can expand to providing 20% of all of the country’s supply by 2030.
  • The scientists ran year-long computationally demanding simulations to build a range of scenarios achieving the goal, then sought to understand the most efficient ones to maximise turbines’ energy output while minimizing negative effects on the local climate.
  • Rather than installing more turbines, competing for land with other economic activities, the scientists said that deploying larger, next-generation turbines can maximise yields and efficiency while having small effects on the climate. 


Like wind turbines, the output of solar energy systems also depends largely on the weather — in this case, cloud cover. A recent study proposed a new method to use data from recently launched NASA satellites to predict the optical effects of clouds and the output of solar panels around the world.

  • Three properties of a cloud can affect the amount of sunlight that passes through it: its height, i.e. the altitude of its top; its thickness, the difference between the altitude of the top and bottom; and its optical depth, i.e. how its composition absorbs or modify the light before it reaches the earth’s surface. 
  • The researchers used satellite tools to build a model estimating the clouds’ height, thickness and optical depth. Then, they coupled this with temperature and humidity data obtained from weather stations at the ground level. It allowed them to produce accurate real-time estimates of cloud optical properties — optimising solar forecasting and ensuring we can make the most of solar power. 

Other news on related topics

Sign up for the ENGIE Innovation Newsletter


By continuing to browse, you accept that ENGIE uses cookies for the purpose of saving information about your browsing on the Site. Cookies help make your browsing experience easier and also make it possible to gather statistics about who visits the Site with the goal of improving its quality. For more information, click here.
Consult cookies policy

close icon