Forecasting >100 MeV SEP Events and Intensity based on CME and other Solar Activities using Machine Learning

Author

Daniel Lee Griessler, Florida Institute of Technology

Date of Award

7-2023

Document Type

Thesis

Degree Name

Master of Science (MS)

Department

Electrical Engineering and Computer Science

First Advisor

Philip Chan

Second Advisor

Ming Zhang

Third Advisor

Marius C. Silaghi

Fourth Advisor

Philip J. Bernhard

Abstract

There is a severe risk for astronauts and machinery from high intensity Solar Energetic Particle (SEP) events which can be mitigated through accurate forecast of their presence and peak intensity. By using characteristics of CME and other space weather phenomena, machine learning techniques have the potential to classify and predict the peak intensity of SEP events. The extreme scarcity of SEP events in current datasets poses a challenge to traditional machine learning techniques. In this work, we first demonstrate classifier machine learning techniques that can achieve an F1 score of 0.800 in forecasting SEP events. We then propose techniques for forecasting SEP peak intensity including Combining Richardson forecast (RC), learning Richardson Error (RE), and integrating retraining with DenseLoss (DL+rRT+AE). Finally, we demonstrate through DL+rRT+AE that we can achieve the same F1 score of 0.800 for forecasting SEP peak intensity.

Comments

Copyright held by author

Recommended Citation

Griessler, Daniel Lee, "Forecasting >100 MeV SEP Events and Intensity based on CME and other Solar Activities using Machine Learning" (2023). Theses and Dissertations. 1318.
https://repository.fit.edu/etd/1318