Geomagnetic storms have long been totally unpredictable. Today, thanks to satellite measurements and artificial intelligence, researchers promise to be able to alert some 30 minutes before they occur. It’s already that.
If a solar storm of the magnitude at least that hit Earth in 1859 — the one that has gone down in the history books as the “Carrington Event” — were to happen again, it’s all our electrical infrastructure and our communication networks that would be threatened. And with them, entire populations. As we approach the solar maximum of cycle 25, this risk increases. And despite the efforts of scientists, the solar flares that cause these events remain difficult to predict. NASA researchers confirm today that they would have a hard time giving us a reaction time of more than… 30 minutes!
However, scientists reassure us. According to them, 30 minutes could be just enough time to prepare and avoid the worst for the most critical installations. All thanks to satellite images and artificial intelligence.
Protect critical systems with AI
The researchers applied a method called deep learning — or deep learning — to train their computerscomputers to make the link between solar wind measurements – obtained by numerous heliophysical missions of the ACE, Wind or Geotail type – and geomagnetic disturbances recorded on Earth. They then developed a model called Dagger — to Deep Learning Geomagnetic Perturbation — and able to predict where geomagnetic disturbances will strike in less than a second, and this, 30 minutes, therefore, before they occur.
The computer code of the model is accessible in open source. The researchers hope that it can be adopted by power grid operators, satellite controllers, telecommunications companies and others to apply these predictions to their specific needs. And one day perhaps, alarm sirens will sound in power stations to order the temporary shutdown of a sensitive system.
