Lightning Prediction Using Recurrent Neural Networks

Abstract

In this paper, hourly statistics were generated from the raw EFMs data set used in Hill. Input variables were generated from surface observations from every station within 50 miles of CCAFS and then combined with the EFM statistics for the same time periods. This combined data set was used to create Long Short-term Memory (LSTM) Neural Networks designed to capture trends within the data for each observation. A variety of different LSTM model structures were created and trained to see which model structure performed best when predicting lightning around CCAFS, KSC, and PAFB. By utilizing design of experiments techniques, optimal parameters for the LSTM model structures are narrowed down providing a solid baseline for future endeavors in predicting lightning.

Document Details

• Document Type: Technical Report
• Publication Date: Mar 21, 2019
• Accession Number: AD1077559

Entities

People

• Dominick V. Iii Speranza

Organizations

• Air Force Institute of Technology

Tags

Communities of Interest

• Energy and Power Technologies
• Space
• Weapons Technologies

DTIC Thesaurus Topics

• Air Force
• Air Force Facilities
• Algorithms
• Artificial Neural Networks
• Atmospheric Sciences
• Basic Programming Language
• Computer Languages
• Computer Programming
• Data Mining
• Data Science
• Data Set
• Data Sets
• Department Of Defense
• Detection
• Digital Data
• Experimental Design
• Governments
• Information Science
• Machine Learning
• Neural Networks
• Programming Languages
• Recurrent Neural Networks
• Statistical Analysis
• Statistics
• United States
• United States Government
• Warning Systems
• Weather Forecasting
• Weather Stations

Readers

• Atmospheric Science/Meteorology
• Computational Modeling and Simulation

Technology Areas

• AI & ML
• AI & ML - Bayesian Inference
• AI & ML - Neural Networks