Imagine the sun as a giant battery that occasionally has dangerous short-circuits called solar flares. These flares send bursts of energy toward Earth that can fry electronics and knock out power grids. This project uses deep learning—similar to how AI recognizes faces—to spot the warning signs on the sun's surface and predict these eruptions before they happen.
Current space weather forecasts rely on limited statistical models that lack accuracy and timeliness. This creates high risk for power grids, navigation systems, and spacecraft as they cannot reliably predict imminent solar flares.
A flare forecasting system using deep neural networks, including cloud-based executable services and user interfaces integrated into the PITHIA-e-Science Centre.
If you are a satellite operator dealing with unpredictable signal loss and hardware damage — this project developed deep neural networks that predict flares in real-time. This allows for proactive shielding or shutdown of sensitive instruments to prevent permanent damage.
If you are a grid manager dealing with sudden voltage surges caused by space weather — this project developed a forecasting system with uncertainty metrics. This helps in adjusting grid loads to prevent massive blackouts during solar events.
If you are a mission provider dealing with radiation risks for astronauts on missions like Artemis — this project developed a system to classify active solar regions. This ensures safer transit windows for crews traveling far from Earth's protection.
Based on available project data, no pricing is mentioned as the datasets, codes, and deep neural networks will be made openly available.
The project includes a cloud-based implementation and executable services integrated with the PITHIA-e-Science Centre, suggesting a scalable digital architecture.
The project explicitly states that the developed datasets, codes, and DNNs will be made openly available to support research and re-use.
The system is designed to be integrated into the PITHIA-e-Science Centre via executable services and preliminary user interfaces.
The project period runs from 2022-12-01 to 2026-02-28, indicating it is currently in the development and testing phase.
The consortium is purely academic, consisting of 4 partners from 4 countries (HU, IE, IT, UK). With 3 universities and 1 research institute, the project has a 0% industry ratio, meaning the current focus is on scientific validation and open-source tool development rather than immediate commercial productization.
Dublin Institute for Advanced Studies
Contact us to explore how to integrate these open-source DNNs into your risk management software.
