AI forecasts solar wind days ahead, boosting protection of satellites and power grids

The research team at NYU Abu Dhabi, led by postdoctoral associate Dattaraj Dhuri with co-principal investigator Shravan Hanasoge at the Center for Space Science (CASS), has developed an AI model that forecasts solar wind up to four days ahead.

The system achieves 45% higher accuracy than current operational models and a 20% improvement compared to previous AI approaches. Results are published in The Astrophysical Journal Supplement Series.

“This is a major step forward in protecting the satellites, navigation systems, and power infrastructure that modern life depends on,” said Dhuri. “By combining advanced AI with solar observations, we can give early warnings that help safeguard critical technology on Earth and in space.”

Unlike language models that process text, this AI was trained to analyze ultraviolet (UV) solar images taken by NASA’s Solar Dynamics Observatory (SDO). SDO images reveal structures in the Sun’s atmosphere, such as coronal holes, where magnetic fields open and allow high-speed solar wind streams to escape.

The researchers combined these images with historical solar wind records to teach the AI to recognize visual patterns linked with changes in solar wind speed. By learning from both solar imagery and past outcomes, the model identifies signals invisible to traditional statistical methods.

The paper reports root-mean-square errors (RMSEs) of about 55–58 km/s (34–36 miles/s) and Pearson correlations of 0.78 (1-day lead) down to 0.63 (4-day lead). On unseen test data from 2019–2023, the model achieved a 4-day RMSE of 53 km/s (33 miles/s) and correlation of 0.55. These results translate to the 45% and 20% accuracy gains highlighted by NYUAD.

Predicting solar wind has long been a challenge for space scientists. The solar wind travels at speeds of 300–800 km/s (190–500 miles/s), and its behavior depends on complex interactions of solar magnetic fields and plasma.

Traditional models rely on spacecraft measurements of the Sun’s magnetic field or solar atmosphere, but these often provide only 1–2 days of warning. This limited lead time is often insufficient for satellite operators and power grid managers to prepare.

An AI model capable of extending forecasts to four days significantly improves planning horizons. It could allow operators to adjust satellite orbits, power systems, and communication networks before a storm begins.

Human reliance on space-based and electrical infrastructure means solar wind disturbances can have cascading effects. Satellites used for communications, Earth observation, and navigation are directly exposed to space weather.

On the ground, power grids are vulnerable to geomagnetically induced currents (GICs), which can overload transformers. Pipelines and undersea cables can also experience unexpected currents during strong solar storms.

The risk is not theoretical. In March 1989, a geomagnetic storm caused by a coronal mass ejection (CME) led to a nine-hour blackout in Quebec, Canada, affecting 6 million people. The most famous event, the Carrington Event of 1859, disrupted telegraph systems worldwide.

If a Carrington-class storm occurred today, it could cause widespread communication failures, satellite losses, and economic damage in the trillions of USD.

The timing of the NYUAD breakthrough is significant. The Sun follows an 11-year cycle of activity, with peaks marked by frequent solar flares, sunspots, and strong solar wind.

NOAA’s Solar Cycle Prediction Panel places the maximum of Cycle 25 around July 2025, with an uncertainty window stretching from November 2024 to March 2026. During this peak, solar storms are more likely, making accurate forecasts essential for protecting infrastructure.

Agencies such as the National Oceanic and Atmospheric Administration’s Space Weather Prediction Center (NOAA SWPC) and the European Space Agency’s Space Weather Coordination Centre (ESA SWCC) issue warnings based on solar observations. An AI system like NYUAD’s could complement these efforts, providing earlier and more accurate forecasts.

The use of AI in heliophysics reflects a broader trend across scientific disciplines. AI systems are now used to analyze astronomical images, model climate change, and predict earthquakes.

In space weather research, AI offers a way to process vast amounts of solar data from spacecraft like SDO and missions planned for the coming decade. By finding patterns beyond human recognition, these systems can extend forecast windows and reduce uncertainty.

For space agencies, utilities, and satellite operators, AI-enhanced forecasting tools could become as essential as weather satellites are for terrestrial weather.

NYU Abu Dhabi has become a significant hub for research in the Middle East. With over 90 faculty labs and 9 200 international publications, it contributes to global science in physics, biology, engineering, and space research.

Times Higher Education ranks NYU among the world’s top 35 universities, making NYUAD the highest-ranked institution in the UAE. The solar wind forecasting breakthrough underscores its role in tackling globally significant challenges.

The next steps will involve testing the AI model against real-time solar data to evaluate its performance during the coming solar maximum. Integration into operational centers such as NOAA SWPC or ESA SWCC would be a logical progression, ensuring the forecasts reach decision-makers.

As satellites multiply in low Earth orbit and global dependence on digital infrastructure deepens, the ability to forecast harmful solar wind days ahead could prove vital. AI is not a replacement for physical understanding of the Sun, but it is becoming a critical partner in safeguarding technology and society from space weather.

References:

¹ NYUAD Scientists Use AI to Forecast Harmful Solar Winds Days in Advance – NYUAD – September 16, 2025

² A Multimodal Encoder–Decoder Neural Network for Forecasting Solar Wind Speed at L1 – Dattaraj Dhuri et al. – The Astrophysical Journal Supplement Series – September 8, 2025 – https://doi.org/10.3847/1538-4365/adf436 – OPEN ACCESS

Reet Kaur

I’m a science journalist and researcher at The Watchers, contributing to the Epicenter edition, where I cover peer-reviewed scientific research and emerging discoveries across Earth and space sciences. With a background in astronomy and a passion for environmental science, I’ve worked in shark and coral conservation in Fiji, conducting reef and shark-behavior research, contributing to mangrove restoration, and earning PADI Open Water and Coral Reef Certifications. I bring a blend of scientific rigor and storytelling to illuminate the discoveries shaping our planet and beyond.

This author does not have any more posts.