Ukrainian scientists won the NASA Space Apps Challenge hackathon, where more than 5500 teams competed in 10 categories. The Ukrainian team Storm Prophet, consisting of six members, emerged victorious in the "Best Use of Data" category.
The goal of their project is to predict solar storms based on raw data from the Deep Space Climate Observatory (DSCOVR) for the years 2016-2023. DSCOVR continues to operate after the end of its operational period and occasionally malfunctions, which can serve as indicators of changes in space weather, reports Alpha Centauri.
The team's project is based on Long Short-Term Memory (LSTM) networks and advanced data preprocessing. Scientists used the disturbance time of the Earth's magnetic field as an indicator of storm danger, with the efficiency of their model evaluated based on its ability to predict this value. The LSTM model demonstrated high accuracy: the long-term model showed 3% deviation from expected data, while the short-term model showed 1.8% deviation. Another Sequence to Sequence model used several hours of DSCOVR data to provide forecasts hours or days ahead. It showed a deviation of 3-5% from expected data in hourly forecasts, with the greatest deviation reaching 10% in forecasts for a month ahead.
This is the third Ukrainian victory in the NASA Space Apps Challenge. The first two are dated back to 2016. The team from the Zhytomyr Military Institute named after Korolev won in the Do not Crash My Drone category with their project DronProtect, and the Mars Hopper project won the "audience choice" award.
Comments (0)
There are no comments for now