Blowing the Wind

São Paulo | Save the Earth from Another Carrington Event!

Blowing in the Wind

High-Level Project Summary

To predict events like Carrington, we created an alert system that predicts the speed of the solar wind, from the data of Magnetic Field, Density and Ion Temperature of the solar wind, measured in real time by the Wind and DSCOVR satellites, predicting the solar wind speed one hour in advance. Our approach uses reinforcement learning to train the neural network, and who will help in the training will be any user who is interested in the subject.

Link to Final Project

https://github.com/zemarchezi/NASASpaceApps

Link to Project "Demo"

https://github.com/maetthil/NASASpaceApps/tree/main/docs https://www.canva.com/design/DAFN2R9F_ps/a7Lgx_zaL48eDTNWenn8fA/view

Detailed Project Description

You are in your house, watching TV, until BOOM all the lights and electronics go out, you have no internet or GPS signal. From the window, you can see that the entire neighborhood is going through the same thing. That's what would happen during the Carington Event, which in 1859 was the biggest solar flare ever recorded. In a world highly dependent on technology and electricity, the Carrington event could cause a catastrophe of unimaginable proportions worldwide.

Solar flares release particles that mess with the earth's magnetic field, causing satellites to stop working and intense currents to burn electricity transmission transformers. In 1989, a solar storm left Canada without electricity for 9 hours. In 2022, Starlink lost 40 satellites during its launch, caused by another solar storm. In an article published by National Geographic, if Carrington happened again as it almost did in 2012, the worldwide damage would be $1 to $2 trillion, and the effects could be felt for years.

With these events becoming more and more recurrent due to the peak of solar activity, the chances of a new Carrington happening increase. To predict events like Carrington, we created an alert system that predicts the speed of the solar wind, from the data of Magnetic Field, Density and Ion Temperature of the solar wind, measured in real time by the Wind and DSCOVR satellites, predicting the solar wind speed one hour in advance. Our approach uses reinforcement learning to train the neural network, and who will help in the training will be any user who is interested in the subject.

The solar wind speed forecast, along with other plasma parameters and real-time images of the sun will be available on an online platform to the public soon. The user will be able to make his own predictions, and if he gets it right, he will be rewarded. This information will be used to reinforce model training. This will improve the model's assertiveness and guarantee a reliable forecast for governments and technology companies, which will be able to prepare for these events and carry out decision-making aimed at minimizing damage, such as satellite shutdown and maneuvers and fault contouring. power.

Space Agency Data

We used data from NASA cdawb repositories :

Wind Mission :https://cdaweb.gsfc.nasa.gov/pub/data/wind
DSCOVR Mission: https://cdaweb.gsfc.nasa.gov/pub/data/dscovr

Canadian Space Agency:

CARIMSA Magnetometer Network: https://donnees-data.asc-csa.gc.ca/dataset/06f5e364-6e2c-4d1c-95c2-9fb7d871ca20

Brazilian Space Agency:

Embrace Magnetometer Network: http://www2.inpe.br/climaespacial/portal/pt/

Hackathon Journey

We chose to "Save the Earth from Another Carrington Event! because we thought it was the coolest and most impactful topic, combining our passion for astrophysics and machine learning.

Our solution provides a forecast of the solar wind speed and sends warnings to the government to take the necessary decisions and minimize the effects. The model uses satellite data provided by NASA and other agencies.

Our idea changes the way the neural network is trained, adding user-provided reinforcement learning, which helps spread the knowledge about Space Weather to everyone.

References