Researchers from New York University in Abu Dhabi (NEW) developed an artificial intelligence system, which is able to predict the speed of the solar wind in four days with much higher accuracy, than the available methods. The results of the study were published in The Astrophysical Journal Supplement Series.
The solar wind is a flow of charged particles from the Sun, which can increase, causing powerful events, that disrupt the work of the Earth's atmosphere, power grid, satellites and even knock spacecraft out of orbit. A vivid example was the event 2022 year, when the solar storm destroyed 40 newly launched Starlink satellites, emphasizing the need for reliable forecasting.
How the new system works
An NYUAD team led by Dattaraj Dhuri and Shravan Hanasoge of the Space Science Center developed a neural network, trained on ultraviolet images of the Sun from the NASA apparatus Solar Dynamics Observatory and multi-year archival solar wind data. Unlike popular AI language models, which analyze the text, this system interprets the image of the Sun, recognizing patterns, preceding changes in the solar wind.
Results and implications
The model achieved an increase in the accuracy of forecasting by 45% compared to operating systems and on 20% outperformed previous AI methods. Such progress allows early warning of space weather phenomena, which are protected by satellites, navigation systems and energy infrastructure, on which modern life depends.
This breakthrough demonstrates, how machine learning can solve one of the most difficult problems in astrophysics — predicting the solar wind. More accurate predictions allow governments, researchers and industry better prepare for space weather threats, ensuring the protection of critical technology and services worldwide.
Source: https://www.spacedaily.com