Lightning severity classification utilizing the meteorological parameters: A neural network approach

Author

Azhar Omar, M. ; Khair Hassan, M. ; Che Soh, Azura ; Kadir, M. Z. A. Ab

Author_Institution

Dept. of Electr./Electron., Univ. Putra Malaysia, Serdang, Malaysia

fYear

2013

fDate

Nov. 29 2013-Dec. 1 2013

Firstpage

111

Lastpage

116

Abstract

This paper presents a technique of predicting lightning severity on daily basis by using meteorological data. The data used is supplied by Global Lightning Network (GLN) from WSI Corporation. The input of the system consists of seven meteorology parameters which had been provided by Malaysia Meteorology Service with minimal fees. Input parameters are the Minimum Humidity, Maximum Humidity, Minimum Temperature, Maximum Temperature, Rainfall, Week and Month. The output of the system determines the severity of lightning predictions in three stages; Class1: Hazardous; Class2: Warning; and Class3: Low Risk. Two training algorithms that have been tested in this study namely the Gradient Descent with Momentum Backpropagation (traingdm) and the Scaled Conjugated Gradient Backpropagation (trainscg). The traingdm has indicated better accuracy of 70% compared to the trainscg whilst in contrast; trainscg has demonstrated approximately 4 times faster training compare to traingdm.

Keywords

backpropagation; conjugate gradient methods; geophysics computing; lightning; meteorology; neural nets; GLN; Malaysia Meteorology Service; WSI Corporation; global lightning network; gradient descent with momentum backpropagation; input parameters; lightning severity classification; lightning severity prediction; meteorological data; meteorological parameters; meteorology parameters; neural network approach; scaled conjugated gradient backpropagation; traingdm; training algorithms; trainscg; Accuracy; Artificial neural networks; Backpropagation; Humidity; Lightning; Training; Artificial Neural Network; Backpropagation; Lightning severity prediction; Scaled Conjugated Gradient;

fLanguage

English

Publisher

ieee

Conference_Titel

Control System, Computing and Engineering (ICCSCE), 2013 IEEE International Conference on

Conference_Location

Mindeb

Print_ISBN

978-1-4799-1506-4

Type

conf

DOI

10.1109/ICCSCE.2013.6719942

Filename

6719942