High-Level Project Summary
The model's main role is to predict the solar storms before their occurrence within 24 hours, where it trained on the scaled data before the danger by 24 hours and the danger time, the model will take the data (from the datasets provided by NASA) and compress the input data then reconstruct it once more(using auto-encoders), then if the scaled input data mean fits on the compressed data mean then there is a high chance of a solar storm impacting the magnetosphere of the earth within the next 24 hours. However, if there was a reconstructive error this means that there is no predicted danger on the earth within the next 24 hours, then the model passed by several tests to ensure its accuracy.
Link to Final Project
Link to Project "Demo"
Detailed Project Description
In 1859, history recorded the most intense geomagnetic storm, due to the colliding of a coronal mass ejection with the earth's magnetosphere, this induced a current from the electromagnetic field that caused telegraph systems in some cases giving their operators electric shocks moreover, some operator was able to continue sending and receiving messages despite having disconnected the power supplies, if an event like with was to happen again to our earth, having an early alert before it happens will reduce the damage that the earth may endure.
The model's main role is to predict the solar storms before their occurrence within 24 hours, where it trained on the scaled data before the danger by 24 hours, furthermore, the model will take the data (from the datasets provided by NASA) and compress the input data then reconstruct it once more(using auto-encoders), then if the scaled input data mean fits on the compressed data mean then there is a high chance of a solar storm impacting the magnetosphere of the earth within the next 24 hours. However, if there was a reconstructive error this means that there is no predicted danger on the earth within the next 24 hours, then the model passed several tests to ensure its validation and determine its accuracy (it passed 85%) after the AI model was complete it was warped with an API to be able to interface easily with the GUI.
With a model providing an ability to predict those solar storms within 24 hours before their occurrence, we can literally save the earth from a possible internet apocalypse, and even if another Carrington event happened, we will have time to minimize its damage on earth.
Going on with this program we used python programing language, for applying the AI model we used neural network and auto-encoders, for the API that will warp the model we used flask API, and for the GUI we made it as a desktop application by Figma.
Space Agency Data
we download data from the NASA website we download data that include information about protons' speed before the danger by 24 hours and in the danger time and it is some examples of day that happened damage (14/7/2000 and 5/11/2001)
the link:- https://cdaweb.gsfc.nasa.gov/pub/data/wind/swe/swe_h1/2022/
Hackathon Journey
The whole experience of reading about an event that happens in the sun and impact the earth was quite fascinating, as the whole team is consist of undergraduate’s computer science and AI students, therefore investing time in reading about a challenging moreover important event like Carrington event and how it happened despite being a topic irrelevant to our carrier it was remarkably interesting.
Furthermore, this was our first time participating in NASA space apps challenges, and the experience was fabulous, from making a team with different people you never knew before, passing by learning a new technology to use them with your work, ending with staying up all night working with your team, all of this was remarkable for us to remember.
References
Tags
#AI #Artificial intelligence #predict #Solar Danger #Carrington Event #NASA #Save the earth # Problem Caused from Carrington