The use Of Intelligent Algorithms For locating Cloud To ground lightning Strokes Based On Electromagnetic model Of lightning Return Stroke Channel Mansour Nejati Jahromi Abstract

finickyontarioΤεχνίτη Νοημοσύνη και Ρομποτική

29 Οκτ 2013 (πριν από 4 χρόνια και 8 μήνες)

134 εμφανίσεις

The use Of Intelligent Algorithms For locating Cloud
To ground lightning Strokes Based On
Electromagnetic model Of lightning Return Stroke

Mansour Nejati Jahromi

In order to utilize the intelligent methods in the lightning
locating and benefit the advantages of these methods,
two factors of data gathering and preventive the locating
errore are the significant points. With this aim, presented
thesis stems from three fundamental subjects; modeling,
electromagnetic field measurement and intelligent
lighting locating.
In this thesis, for the modeling of the problem, a new
method is presented to model a Cloudto- Ground (CG)
lightning return-stroke channel. In this method, the
lightning channel is approximated by a lossy monopole
antenna with distributed resistive-inductive load above a
lossy ground. To determine the temporal-spatial current
distribution along the channel, the governing electric
field integral equation (EFIE) in frequency domain is
solved numerically by method of moment. As opposed to
the conventional models, the proposed model enables
one to assign the realistic value of unity to the relative
permittivity of the surrounding medium. In addition, the
channel speed can be appropriately adjusted using the
distributed inductive loads along the channel. The latter
enables one to assign a realistic propagation speed to the
channel current. In addition, the inclusion of lossy
ground in the proposed model further enhances its
superiority over the conventional models for accurate
modeling of a lightning channel. In fact, a lossy ground
produces a nonzero horizontal component of electric
filed whose value is found to be closer to its actual value
for having a more realistic current propagation speed
along the channel. The error in simulation is reduced by
this quality. To demonstrate the validity of the model, the
current distribution along the channel and the radiated
electromagnetic field are computed and compared with
those obtained using the conventional AT (Antenna
Model) models. A comparison of the results confirms the
validity of the proposed model. It is also shown that, as
opposed to the conventional AT models, the proposed
model is capable of predicting the zero-crossing feature
of electromagnetic fields waveform in a far range as
observed in all measurement data.
In measurement, the system is used as a narrow-band
lightning locating network using the radiated
electromagnetic fields due to a cloud-to-ground (CG)
lightning return stroke channel (RSC). The network is
composed of three measurement stations, located at the
vertices of a triangle covering the region of interest. The
stations are networked together using a low cost and
access-on-demand dial-up connection. Each station
utilizes a wavelet-based compression algorithm to duly
reject unwanted signals recorded during a lightning flash
while extracting the time of arrival, direction of arrival,
polarity and peak values of electromagnetic field
associated with the first and subsequent RSCs. To
determine the location of a CG lightning stroke, a
timedifference- of-arrival approach is used.
In the intelligent lightning locating subject, two
parameters of incidence angle and distance from the
station are needed in order to estimate the location where
the lightning strikes the ground. Incidence angle is
obtained from the ratio of the tangential components of
the measured magnetic field and the distance from the
station is found from the electrical field In order to
estimating distance from measured electromagnetic
fields, three methods, namely, “iterative algorithm”,
“multi-layer perceptron neural network algorithm” and
“wavelet network algorithm” methods are used. Results
show that iterative method can estimate distances but it
has slow convergence speed. On the other hand, in neural
network, although the training phase is time consuming,
obtaining the results is fast in the test phase. Therefore,
this method is more desirable for real time applications.
Wavelet network results didn’t show any advantageous
over multilayer perceptron network results taking into
account that the convergence speed of wavelet network is
even lower than multi-layer perceptron network. It is
observed that using two stations along with intelligent
methods leads to locating error reduction of 3%.
Key words: Intelligent algorithms , Locating,
Electromagnetic model, Return stroke, Lightning
channel, Single station, Measurement, Lightning,
Electromagnetic field, Locating, Time-of arrival,
Wavelet, Narrow-band