Google algorithms predict energy production from windmills a day in advance

DeepMind, a sister company of Google, says it has deployed artificial intelligence at a number of windmills in the US. This makes it possible to predict how much energy will be produced 36 hours in advance. This would increase the value of the wind energy.

The DeepMind researchers state that in their search for a solution to the unreliability of wind energy, they started applying deep learning algorithms to wind turbines last year. This involves a total of seven hundred megawatts of wind energy capacity that is supplied by windmills somewhere in the middle of the United States. According to the researchers, the AI ​​successfully predicted energy output prematurely, up to 36 hours prior to actual energy production.

This was done by training the neural network on the widely available weather forecasts and historical data about the wind turbines. Based on this, the system was configured to predict the amount of energy to be generated 36 hours in advance. With the model, recommendations can then be made a day in advance to optimize the energy supply to the grid per hour. According to the researchers, this is important, because energy sources that can be ‘planned’, and therefore can supply a certain amount of power at a certain time, are usually more valuable for the electricity network.

According to the researchers, the algorithm is currently being further refined, but has already yielded positive results. They state that the ‘value’ of the wind energy has improved by roughly twenty percent compared to the standard scenario in which no predictions have been made of the amount of power to be produced.