Inspiration

Solar storms can destroy satellites and knock out power grids, but current warning systems aren't accurate enough to help companies prepare effectively.

What it does

olarShield uses AI to predict dangerous solar flares with 87.3% accuracy and tells satellite operators, airlines, and power companies exactly what risks they face.

How we built it

We trained machine learning models on 2,026 real NASA solar flare records from 2019-2024, using features like flare intensity, duration, and location to predict threats.

Challenges we ran into

NASA's API kept crashing with 503 errors, major solar events are rare so we had a hard time collecting and training data

Accomplishments that we're proud of

We beat the industry standard of 85% accuracy by achieving 87.3% detection of major solar events, and our system actually works on real 2024 data.

What we learned

Temporal validation is crucial because you can't predict the future with past events mixed randomly, and building reliable systems requires multiple backup data sources.

What's next for SolarShield

We want to add coronal mass ejection prediction, partner with actual satellite companies for testing, and eventually build a real-time global monitoring network.

Share this project:

Updates