Documente Academic
Documente Profesional
Documente Cultură
tex
%% V1.4
%% 2012/12/27
%% by Michael Shell
%% See:
%% http://www.michaelshell.org/
%% for current contact information.
%%
%% This is a skeleton file demonstrating the use of IEEEtran.cls
%% (requires IEEEtran.cls version 1.8 or later) with an IEEE conference paper.
%%
%% Support sites:
%% http://www.michaelshell.org/tex/ieeetran/
%% http://www.ctan.org/tex-archive/macros/latex/contrib/IEEEtran/
%% and
%% http://www.ieee.org/
%%*************************************************************************
%% Legal Notice:
%% This code is offered as-is without any warranty either expressed or
%% implied; without even the implied warranty of MERCHANTABILITY or
%% FITNESS FOR A PARTICULAR PURPOSE!
%% User assumes all risk.
%% In no event shall IEEE or any contributor to this code be liable for
%% any damages or losses, including, but not limited to, incidental,
%% consequential, or any other damages, resulting from the use or misuse
%% of any information contained here.
%%
%% All comments are the opinions of their respective authors and are not
%% necessarily endorsed by the IEEE.
%%
%% This work is distributed under the LaTeX Project Public License (LPPL)
%% ( http://www.latex-project.org/ ) version 1.3, and may be freely used,
%% distributed and modified. A copy of the LPPL, version 1.3, is included
%% in the base LaTeX documentation of all distributions of LaTeX released
%% 2003/12/01 or later.
%% Retain all contribution notices and credits.
%% ** Modified files should be clearly indicated as such, including **
%% ** renaming them and changing author support contact information. **
%%
%% File list of work: IEEEtran.cls, IEEEtran_HOWTO.pdf, bare_adv.tex,
%% bare_conf.tex, bare_jrnl.tex, bare_jrnl_compsoc.tex,
%% bare_jrnl_transmag.tex
%%*************************************************************************
% *** Authors should verify (and, if needed, correct) their LaTeX system ***
% *** with the testflow diagnostic prior to trusting their LaTeX platform ***
% *** with production work. IEEE's font choices can trigger bugs that do ***
% *** not appear when using other class files. ***
% The testflow support page is at:
% http://www.michaelshell.org/tex/testflow/
%\usepackage{stfloats}
% stfloats.sty was written by Sigitas Tolusis. This package gives LaTeX2e
% the ability to do double column floats at the bottom of the page as well
% as the top. (e.g., "\begin{figure*}[!b]" is not normally possible in
% LaTeX2e). It also provides a command:
%\fnbelowfloat
% to enable the placement of footnotes below bottom floats (the standard
% LaTeX2e kernel puts them above bottom floats). This is an invasive package
% which rewrites many portions of the LaTeX2e float routines. It may not work
% with other packages that modify the LaTeX2e float routines. The latest
% version and documentation can be obtained at:
% http://www.ctan.org/tex-archive/macros/latex/contrib/sttools/
% Do not use the stfloats baselinefloat ability as IEEE does not allow
% \baselineskip to stretch. Authors submitting work to the IEEE should note
% that IEEE rarely uses double column equations and that authors should try
% to avoid such use. Do not be tempted to use the cuted.sty or midfloat.sty
% packages (also by Sigitas Tolusis) as IEEE does not format its papers in
% such ways.
% Do not attempt to use stfloats with fixltx2e as they are incompatible.
% Instead, use Morten Hogholm'a dblfloatfix which combines the features
% of both fixltx2e and stfloats:
%
% \usepackage{dblfloatfix}
% The latest version can be found at:
% http://www.ctan.org/tex-archive/macros/latex/contrib/dblfloatfix/
% *** Do not adjust lengths that control margins, column widths, etc. ***
% *** Do not use packages that alter fonts (such as pslatex). ***
% There should be no need to do such things with IEEEtran.cls V1.6 and later.
% (Unless specifically asked to do so by the journal or conference you plan
% to submit to, of course. )
\begin{equation}
%
% paper title
% can use linebreaks \\ within to get better formatting as desired
% Do not put math or special symbols in the title.
\title{Breakdown Prediction of Oil Insulation}
% for over three affiliations, or if they all won't fit within the width
% of the page, use this alternative format:
%
%\author{\IEEEauthorblockN{Michael Shell\IEEEauthorrefmark{1},
%Homer Simpson\IEEEauthorrefmark{2},
%James Kirk\IEEEauthorrefmark{3},
%Montgomery Scott\IEEEauthorrefmark{3} and
%Eldon Tyrell\IEEEauthorrefmark{4}}
%\IEEEauthorblockA{\IEEEauthorrefmark{1}School of Electrical and Computer
Engineering\\
%Georgia Institute of Technology,
%Atlanta, Georgia 30332--0250\\ Email: see
http://www.michaelshell.org/contact.html}
%\IEEEauthorblockA{\IEEEauthorrefmark{2}Twentieth Century Fox, Springfield, USA\\
%Email: homer@thesimpsons.com}
%\IEEEauthorblockA{\IEEEauthorrefmark{3}Starfleet Academy, San Francisco,
California 96678-2391\\
%Telephone: (800) 555--1212, Fax: (888) 555--1212}
%\IEEEauthorblockA{\IEEEauthorrefmark{4}Tyrell Inc., 123 Replicant Street, Los
Angeles, California 90210--4321}}
% no keywords
% For peer review papers, you can put extra information on the cover
% page as needed:
% \ifCLASSOPTIONpeerreview
% \begin{center} \bfseries EDICS Category: 3-BBND \end{center}
% \fi\begin{abstract}
\begin{abstract}
Investigate the modelling and analysis of the electric field distribution in oil
gaps, stressed by the breakdown voltages, under varying gap distance and rate of
voltage rise, with finite element method. Breakdown voltage of the oil gap is of
vital importance for the design of the transformer insulation in high-voltage. In
this report, a prediction method for the breakdown voltages of typical oil gaps
based on the dynamic electric field features and support vector machine (SVM) is
discussed. According to the finite element calculation results of dynamic electric
field distribution, the electric field values in the whole region, discharge
channel and surface of the electrode were extracted and post-processed, which
constituted the electric field features characterizing the gap structure. Then, the
breakdown voltage prediction model of the oil gap was established by using electric
field features as the input parameters to SVM, and whether the oil gap breakdown
would happen as the output parameters of SVM, which changing the regression problem
to a binary classification problem. This model was applied to predict the
dielectric strength of oil gaps of parallel plate electrodes with varying distance
and varying rate of voltage rise.
\end{abstract}
%
% For peerreview papers, this IEEEtran command inserts a page break and
% creates the second title. It will be ignored for other modes.
\IEEEpeerreviewmaketitle
\section{Introduction}
The study of the field strength distribution in electric fields, and especially in
quasi-uniformly electric fields like the fields in parallel plate electrodes, is of
great importance in high voltages. One of the most determinant factors of the
dielectric strength of the insulating materials is the field strength distribution
inside the mass of the materials, when stressed by high voltages.
The experimental measurement of the field strength in small oil gaps stressed by
high voltages is a very difficult process and not quite accurate. The analysis of
the proper models of the air gap electric fields using Laplace and Poisson
equations for general 2 or 3-dimensional fields with a theoretical mathematical way
is more accurate but in some cases it leads to difficulties and is time consuming.
The most convenient way¬¬¬ is to use numerical procedures.
This report investigates the modelling and analysis of the dynamic electric field
distribution in oil gaps, stressed by the breakdown voltages, under flat plate
electrode with finite element method.
% Note that IEEE typically puts floats only at the top, even when this
% results in a large percentage of a column being occupied by floats.
% An example of a floating table. Note that, for IEEE style tables, the
% \caption command should come BEFORE the table. Table text will default to
% \footnotesize as IEEE normally uses this smaller font for tables.
% The \label must come after \caption as always.
%
%\begin{table}[!t]
%% increase table row spacing, adjust to taste
%\renewcommand{\arraystretch}{1.3}
% if using array.sty, it might be a good idea to tweak the value of
% \extrarowheight as needed to properly center the text within the cells
%\caption{An Example of a Table}
%\label{table_example}
%\centering
%% Some packages, such as MDW tools, offer better commands for making tables
%% than the plain LaTeX2e tabular which is used here.
%\begin{tabular}{|c||c|}
%\hline
%One & Two\\
%\hline
%Three & Four\\
%\hline
%\end{tabular}
%\end{table}
% Note that IEEE does not put floats in the very first column - or typically
% anywhere on the first page for that matter. Also, in-text middle ("here")
% positioning is not used. Most IEEE journals/conferences use top floats
% exclusively. Note that, LaTeX2e, unlike IEEE journals/conferences, places
% footnotes above bottom floats. This can be corrected via the \fnbelowfloat
% command of the stfloats package.
\section{Methodology}
The methodology has been divided into two sections namely, experiential methodology
and prediction methodology.
\subsection{Experimental Methodology}
The breakdown of soybean oil was tested in a standard dielectric oil tester (BAUR
DPA 75 C). Two standard-compliant parallel plate circular test electrodes shape
under the standard ASTM D 877. The parallel plate circular electrode shape was
selected to study the breakdown phenomenon under quasi-uniform electric field. A
test voltage rise for e.g. 5 kV/sec was applied and continuously increased to one
of the electrode while keeping the other electrode grounded until the breakdown
occurs. The experiment was performed under two conditions. Firstly by varying the
gap distance between the electrodes at constant rate of voltage rise. Secondly, by
varying the rate of voltage rise between the electrodes at constant gap distance.
After filling the test cell with the test oil, a hold time of 3 minutes is given
until the start of the first breakdown test with continuous stirring. Between
consecutive breakdowns there is a pause time of 5 minutes with continuous stirring.
The mentioned parameters were maintained constant throughout the breakdown test
experiment. Ten measurements were taken for each mentioned combination of distance
and rate of voltage rise as given in Table 1 and the average breakdown RMS voltage
was measured.
\subsection{Predction Methodology}
The prediction method is based on the dynamic electric field features and SVM. The
experimental data is divided into training and test sample. The Date in bold
italics in Table 2 and Table 3 were chosen as the training samples, whereas the
rest were test samples for the respective oil test condition .The training samples
are utilized to train the SVM model, while the test samples confirm the accuracy of
the prediction model.
Breakdown voltage $U_{b}$ of the oil gap is selected for a particular gap and
voltage rise from the training Data set. The selected voltage at a value 90 \% of
$U_{b}$, named $U_{b}^-$ is unable to cause breakdown is labelled as -1, whereas
the value of 110 \% of $U_{b}$, named $U_{b}^+$ is able to cause breakdown is
labelled as +1. Hence, the regression problem is converted to a two class
classification problem. The electrode geometry is constructed in COMSOL
Multiphysics (CMP) and time dependent boundary conditions are applied to the high
voltage electrode. The simulation was run till time $t_{o}^+$ in equation (4)
\begin{equation}
t_{b}^+ = U_{b}^+ / RR
\end{equation}
CMP determines for the electric field distribution of the gap by finite element
method (FEM) and the corresponding dynamic Electric field features are extracted
after simulation for a step time of 0.1 second. The training samples are formed by
post processing the Electric field features by normalization and dimension
reduction, and the tuning parameters of the SVM model C and γ were found by error
and trial. To test the generalizability of the SVM the test electric field features
corresponding to gap distance and voltage rise were given as an input to the SVM
model, and the output SVM labels were compared with the original test labels; if
they are identical then the SVM has successfully classified the test data and gives
good generalizability. Calculation of Breakdown voltage could be achieved by
applying an initial voltage $U_{o}$, if the SVM prediction model outputs -1, then
applied voltage is set as $U$=$U_{o}$+$dU$, where $dU$=($voltage$ $rise$$\times$
$simulation$ $step$ $time$). The SVM will have different prediction model for oil
test condition one and oil test condition two. The flowchart of the prediction
model is given in Fig. 1.
\begin{figure}[!t]
\centering
\includegraphics[width=9cm,height=8cm]{Fig1.pdf}
\caption{Flowchart of prediction model.}
\label{fig_sim}
\end{figure}
\begin{table}[h]
\begin{center}
\caption{Set of electric field features.}
\label{tab:table1}
\begin{tabular}{|>{\centering\arraybackslash}m{1cm}|
>{\centering\arraybackslash}m{6.0cm}|>{\centering\arraybackslash}m{0.3cm}|}
\hline
\textbf{Spacial location} & \textbf{Electric Field Features} &
\textbf{No.}\\
\hline
Whole Area & $E_{mw}$; $E_{aw}$; $E_{dw}$; $W_{w}$; $W_{ew}$;
$V_{r90w}$; $V_{r75w}$; $V_{r50w}$; $V_{r25w}$; $W_{r90w}$; $W_{r75w}$; $W_{r50w}$;
$W_{r25w}$ & 13\\
\hline
Discharge Area & $E_{md}$; $E_{ad}$; $E_{dd}$; $W_{d}$; $W_{ed};$
$V_{r90d}$; $V_{r75d}$; $V_{r50d}$; $W_{r90d}$; $W_{r75d}$; $W_{r50d}$ & 11\\
\hline
Electrode Surface & $E_{ms}$; $S_{90}$; $S_{75}$; $S_{50}$;
$S_{25}$; $S_{r90}$; $S_{r75}$; $S_{r50}$; $S_{r25}$ & 9\\
\hline
\end{tabular}
\end{center}
\end{table}
\begin{equation}
x^\prime_{i} = \frac{x_{i}-x_{min}}{x_{max}-x_{min}} + 1
\end{equation}
$x^\prime_{i}$ is the normalized value of a certain feature $x_{i}$, $x_{min}
$ and $x_{max}$ are the minimum and maximum values of $x_{i}$. Hence, the features
have been normalized from the interval [$x_{min}$, $x_{max}$] to [1, 2]. Let this
matrix be called $C$.
\begin{equation}
m_{j} = \frac{1}{N}\sum_{i=1}^{N}{C(i,j)}
\end{equation}
Second, we calculate the zero mean matrix $B$ in equation (26).
\begin{equation}
B = C-em^T
\end{equation}
Here $e$ represents an $N\times1$ vector of all ones.
\begin{equation}
D = Y^{-1}ZY
\end{equation}
Fifth, arrange $X$ and $Y$, such that the eigen values are arranged in a descending
order, with their corresponding eigen values given by equation (29).
\begin{equation}
D (1,1) \geq D (2,2) \geq ... \geq D(d,d)
\end{equation}
Sixth, calculate the cumulative variance for each eigenvector by equation (30).
\begin{equation}
G(k) = \frac{\sum_{i=1}^{k}{D(i,i)}}{\sum_{i=1}^{d}{D(i,i)}}
\end{equation}
Thus vectors can be formed as shown in equation (31).
\begin{equation}
G = [G(1) G(2) G(3)...G(d)]
\end{equation}
The first $k$th eigen vectors till $G(k)$ $\geq$ 0.95 are chosen as chosen and the
final training sample data is dimensionally reduced and is given by equation (32)
to train the SVM.
\begin{equation}
Train Sample = B\times [G(1) G(2) G(3)...G(k)]
\end{equation}
In equation (32), the first dimension is called the first principal
component; the second dimension is called the second principal component and so on.
\subsubsection{SVM Theory}
There is a multidimensional non-linear relationship between the electric field
features, the oil gap breakdown voltage and the sample data is limited. SVM is a
machine learning algorithm developed from the optimal hyperplane in linearly
separable cases. SVM has good adaptability in the case of small samples [6]. The
statistical learning in build commands in MATLAB is used in this report to solve
the classification problem. The objective function of the optimization problem is
given by equation (2).
\begin{equation}
\label{main_eq}
%\begin{subequations}
%\begin{aligned}%{rll}
\left.\begin{array}{rll}
\textrm{optimize:}\quad & \underbrace{min}_{\text{\vec{w},b,\zeta}}=\frac{1}
{2}||\vec{w}||^2+C\sum_{i=1}^{N}\zeta_{i}, & \mallabel{a}\\
\textrm{subject to:}\quad & y_{i}(\vec{w}.\vec{x_{i}}-b) \geq 1-\zeta_{i}, \forall
i& \mallabel{b}\\
& where \; \zeta_{i} \geq0, & \mallabel{c}\\
\end{array}\right\}
%\end{align}
%\end{subequations}
\end{equation}
Where $\vec{w}$ is called the weight vector, $b$ is called the bias, is the class
label corresponding to feature vector $x_{i}$, $\zeta_{i}$ are slack variables that
permit margin failure and $C$ is a parameter which trades off wide margin with a
small number of margin failures. When this optimization problem is transformed into
the dual form, it simply changes the constraint optimization into a box constraint
given by equation (4).
\begin{equation}
\label{main_eq}
\left.\begin{array}{rlr}
\textrm{min:}\quad & \psi(\alpha) &\;\;\qquad \mallabel{a}\\
\psi(\alpha)= & X-Y & \mallabel{b}\\
X=&\frac{1}{2}\sum\limits_{i=1}^{i=N}\sum\limits_{j=1}^{j=N}y_{i}y_{j}
(\vec{x_{i}}.\vec{x_{j}})\alpha_{i}\alpha_{j} & \mallabel{c}\\
\begin{equation}
s=sgn{\sum\limits_{i=1}^{i=N}\alpha_{i}y_{i}(\vec{x}.\vec{x_{i}})}
\end{equation}
SVMs can be further generalized to non-linear classifiers. The above equation (4)
and (5) are valid for linear separable case, however most of the real world
problems are non-linear in nature. Therefore a kernel function is introduced which
will map the non-separable from the original space into a feature space which will
be linearly separable in the feature space. The dual form equation (4) and the
decision rule equation (5) are modified as shown in equation (6) and (7).
\begin{equation}
\label{main_eq}
\left.\begin{array}{rlr}
\textrm{min:}\quad & \psi(\alpha) &\;\;\quad \mallabel{a}\\
\psi(\alpha)= & X-Y & \mallabel{b}\\
X=&\frac{1}
{2}\sum\limits_{i=1}^{i=N}\sum\limits_{j=1}^{j=N}y_{i}y_{j}K(\vec{x_{i}}.\vec{x_{j}
})\alpha_{i}\alpha_{j} & \mallabel{c}\\
\begin{equation}
s=sgn{\sum\limits_{i=1}^{i=N}\alpha_{i}y_{i}K(\vec{x}.\vec{x_{i}})}
\end{equation}
\paragraph{Kernel Function}
The limited computational power of linear learning machines was highlighted in the
1960s by Minsky and Papert. In general, complex real-world applications require
more expensive hypothesis spaces than linear functions [7]. Kernel representations
offer an alternative solution by projecting the data into a higher dimension space
where it is linearly separable. The use of linear machine in the dual
representation makes it possible to perform this step implicitly. Thus kernel would
apply a fixed non-linear mapping of the data into a feature space, where it is
linearly separable as shown in fig. 1.
The kernel function chosen is radial basis function (RBF) given in equation (7),
where $\gamma$ is a positive constant, $x_{i}$ is the training observation and
$x_{j}$ be the test observation. If the test observation is far from the training
observation, then the square of its distance will be large, so $K(x_{i}$,$x_{j})$
will be small, then that respective training observation will have no role in
determining the class of the test observation. In other words, training observation
that are far from the test observation will essentially play no role in predicting
class labels from the test observation; this means that the radial basis kernel has
a very local behavior in the sense that only nearby training observations have an
effect on the class labels of a test observation.
\begin{equation}
K(x_{i},x_{j})=exp(-\gamma||x_{i}-x_{j}||^2)
\end{equation}
The advantage of using kernel which rather than simply enlarging the feature space
using functions of the original features is computation, and it amounts to the fact
that using kernels, one need only compute $K(x_{i},x_{j})$ for all $n(n-1)/2$
distinct pairs. This can be done explicitly without working in the enlarged feature
space. This is important because in many applications of SVMs, the enlarged feature
space is so large that computations are intractable. For the RBF kernel, the
feature space is implicit and infinite-dimension, so we could never do the
computation there [8].
The generalization of SVM depends on two parameters $C$ and $gamma$, $C$ parameter
gives a tradeoff between the margin and number of misclassification, whereas gamma
describes the reach of training samples to predict the class of test samples. They
are discussed in detail in the following section.
\begin{equation}
\gamma=\frac{1}{2\sigma^2}
\end{equation}
\subsection{Error Analysis}
\subsubsection{Error Analysis}
Error analysis is part of the essential procedures of the breakdown
voltage prediction of air gaps. By error analysis, the merits of the prediction
model can be evaluated scientifically so as to guide the reasonable application of
the model. Four error indexes as follows are applied in this paper to examine the
error of the prediction results.
\begin{enumerate}
\item Sum of squared error (SSE).
\begin{equation}
e_{SSE} = \sum_{i=1}^{N}{(A_{i}-
P_{i})^2}
\end{equation}
\item Mean square error (MSE).
\begin{equation}
e_{MSE} =\frac{1}{N}
\sqrt{\sum_{i=1}^{N}{(A_{i}-P_{i})^2}}
\end{equation}
\subsection{Experimental Results}
The data for varying electrode gap is shown in Table XX and Table YY, due to the
stochastic behavior of breakdown mechanism, Weibull distribution is used to provide
the scale parameter for the breakdown process.
% \usepackage{multirow}
% \usepackage{multirow}
\documentclass{article}
\usepackage{array,multirow,graphicx}
\begin{document}
\begin{table}[!t]
\begin{center}
\caption{Breakdown voltage for ramp rate at 5
kV/sec.}
\begin{tabular}{|c|c|c|c|c|c|c|}
\hline
\multirow{2}{*}{\textbf{S. No. }} & \multirow{2}{*}{} &
\multicolumn{5}{c|}{\textbf{Ramp Rate 5 kV/sec}} \\
\cline{3-7}
& \multicolumn{1}{c|}{} & \textbf{2.5 mm} & \textbf{3.5 mm} &
\textbf{4.5 mm} & \textbf{5.5 mm} & \textbf{6.5 mm} \\
\hline
1 & \multirow{10}{*}{\rotatebox[origin=c]{90}{\textbf{Breakdown
Voltage (kV)}}} & 25.5 & 37.7 & 48.1 & 44.5 & 53.9 \\
\cline{3-7}
\cline{1-1}
2 & & 25.0 & 28.6 & 49.0 & 56.0 & 52.5 \\
\cline{3-7}
\cline{1-1}
3 & & 38.1 & 40.6 & 51.2 & 39.9 & 52.0 \\
\cline{3-7}
\cline{1-1}
4 & & 30.4 & 34.9 & 51.7 & 66.4 & 47.3 \\
\cline{3-7}
\cline{1-1}
5 & & 28.9 & 36.7 & 51.6 & 34.7 & 64.6 \\
\cline{3-7}
\cline{1-1}
6 & & 23.8 & 26.0 & 35.0 & 53.4 & 46.2 \\
\cline{3-7}
\cline{1-1}
7 & & 29.4 & 35.8 & 46.3 & 46.9 & 48.9 \\
\cline{3-7}
\cline{1-1}
8 & & 29.5 & 30.5 & 33.0 & 45.1 & 46.8 \\
\cline{3-7}
\cline{1-1}
9 & & 28.3 & 48.7 & 34.3 & 33.2 & 49.1 \\
\cline{3-7}
\cline{1-1}
10 & & 20.0 & 42.5 & 43.8 & 63.8 & 53.8 \\
\hline
\end{tabular}
\end{table}
\end{document}
\documentclass{article}
\usepackage{array,multirow,graphicx}
\begin{document}
\begin{table}[!t]
\begin{center}
\caption{Breakdown voltage for electrode gap at 4 mm.}
\begin{tabular}{|c|c|c|c|c|c|c|}
\hline
\multirow{2}{*}{\textbf{S. No. }} & \multirow{2}{*}{} &
\multicolumn{5}{c|}{\textbf{Ramp Rate 5 kV/sec}} \\
\cline{3-7}
& \multicolumn{1}{c|}{} & \textbf{1 kV/s} & \textbf{3 kV/s}
& \textbf{5 kV/s} & \textbf{7 kV/s} & \textbf{9 kV/s} \\
\hline
1 & \multirow{10}{*}{\rotatebox[origin=c]{90}
{\textbf{Breakdown Voltage (kV)}}} & 37.5 & 39.2 & 47.1 & 47.6 & 43.4 \\
\cline{3-7}
\cline{1-1}
2 & & 21.5 & 32.5 & 41.3 & 45.4 & 46.2 \\
\cline{3-7}
\cline{1-1}
3 & & 25.3 & 46.1 & 40.7 & 47.9 & 46.6 \\
\cline{3-7}
\cline{1-1}
4 & & 38.3 & 36.7 & 40.1 & 37.5 & 46.6 \\
\cline{3-7}
\cline{1-1}
5 & & 28.9 & 38.2 & 41.4 & 46.7 & 43.4 \\
\cline{3-7}
\cline{1-1}
6 & & 39.3 & 42.1 & 44.4 & 47.8 & 44.2 \\
\cline{3-7}
\cline{1-1}
7 & & 36.1 & 40.2 & 43.1 & 43.4 & 42.0 \\
\cline{3-7}
\cline{1-1}
8 & & 24.3 & 43.6 & 41.8 & 40.9 & 41.5 \\
\cline{3-7}
\cline{1-1}
9 & & 32.9 & 41.3 & 37.7 & 40.5 & 50.4 \\
\cline{3-7}
\cline{1-1}
10 & & 36.9 & 40.0 & 40.4 & 49.1 & 46.8 \\
\hline
\end{tabular}
\end{table}
\end{document}
\begin{table}[!t]
\begin{center}
\caption{Weibull parameters for electrode gao of 4 mm.}
\label{tab:table1}
\begin{tabular}{|>{\centering\arraybackslash}m{2.0cm}|
>{\centering\arraybackslash}m{2.0cm}|>{\centering\arraybackslash}m{2.0cm}|}
\hline
\textbf{Ramp Rate (kV/s)} & \textbf{Scale Parameter (kV)} &
\textbf{Shape Parameter}\\
\hline
1 & 35.1558 & 4.4384\\
\hline
3 & 41.8107 & 10.1795\\
\hline
5 & 43.0897 & 15.4425\\
\hline
7 & 46.6006 & 10.8403\\
\hline
9 & 47.6363 & 15.8739\\
\hline
\end{tabular}
\end{center}
\end{table}
\begin{table}[!t]
\begin{center}
\caption{Weibull parameters for ramp rate of 5 kV/sec.}
\label{tab:table1}
\begin{tabular}{|>{\centering\arraybackslash}m{2.0cm}|
>{\centering\arraybackslash}m{2.0cm}|>{\centering\arraybackslash}m{2.0cm}|}
\hline
\textbf{Electrode Gap (mm)} & \textbf{Scale Parameter (kV)} &
\textbf{Shape Parameter}\\
\hline
2.5 & 30.0751 & 5.5813\\
\hline
3.5 & 39.1742 & 5.2175\\
\hline
4.5 & 48.1195 & 5.2388\\
\hline
5.5 & 53.1825 & 4.1453\\
\hline
6.5 & 54.2403 & 8.8027\\
\hline
\end{tabular}
\end{center}
\end{table}
\begin{figure}[!t]
\centering
\includegraphics[width=8cm,height=6cm]{Weibull_2.pdf}
\caption{Weibull plot for varying electrode gap at 5 kV/s.}
\label{fig_sim}
\end{figure}
\begin{figure}[!t]
\centering
\includegraphics[width=8cm,height=6cm]{Weibull_ramp_2.pdf}
\caption{Weibull plot for varying ramp rate at 4 mm gap.}
\label{fig_sim}
\end{figure}
\begin{table}[!t]
\begin{center}
\caption{Weibull parameters for electrode gap of 4 mm.}
\label{tab:table1}
\begin{tabular}{|>{\centering\arraybackslash}m{2.0cm}|
>{\centering\arraybackslash}m{2.0cm}|>{\centering\arraybackslash}m{2.0cm}|}
\hline
\textbf{Ramp Rate (kV/sec)} & \textbf{Scale Parameter (kV)} &
\textbf{Shape Parameter}\\
\hline
1.0 & 35.1558 & 4.4384\\
\hline
3.0 & 41.8107 & 10.1795\\
\hline
5.0 & 43.0897 & 15.4425\\
\hline
7.0 & 46.6006 & 10.8403\\
\hline
9.0 & 47.6363 & 15.8739\\
\hline
\end{tabular}
\end{center}
\end{table}
\subsection{SVM prediction}
\begin{figure}[!t]
\centering
\includegraphics[width=8cm,height=6cm]{SVM_Gap.pdf}
\caption{Weibull plot for varying ramp rate at 4 mm gap.}
\label{fig_sim}
\end{figure}
\begin{figure}[!t]
\centering
\includegraphics[width=8cm,height=6cm]{SVM_Ramp.pdf}
\caption{Weibull plot for varying ramp rate at 4 mm gap.}
\label{fig_sim}
\end{figure}
The SVM prediction model was trained with features of 3.5 mm and 5.5 mm and tested
for 2.5 mm, 4.5 mm and 6.5 mm for varying electrode gap. Similarly model was
trained with features of 1 kV/sec and 7 kV/sec and tested for 3 kV/sec, 5 kV/sec
and 9 kV/sec for varying ramp rate , and the output labels of the test data were
compared to the original test data, the tuning parameters of SVM were found by
trial and error. Table T and K shows the predicted values along with error
calculation. the comparison of experimental values and predicted values is shown in
Fig. d and fig. j.
\begin{table}[!t]
\begin{center}
\caption{Weibull parameters for ramp rate of 5 kV/sec.}
\label{tab:table1}
\begin{tabular}{|>{\centering\arraybackslash}m{2.0cm}|
>{\centering\arraybackslash}m{2.0cm}|>{\centering\arraybackslash}m{2.0cm}|
>{\centering\arraybackslash}m{2.0cm}|>{\centering\arraybackslash}m{2.0cm}|}
\hline
\textbf{Gap (mm)} & \textbf{Exp. (kV)} & \textbf{Predicted (kV)}
& \textbf{MAPE} & \textbf{SSE}\\
\hline
2.5 & 30.0 & 30.5 & 5.5813 & 5.5813 \\
\hline
5.5 & 48.0 & 46.5 & 5.5813 & 5.5813\\
\hline
6.5 & 54.5 & 58.5 & 5.5813 & 5.5813\\
\hline
\end{tabular}
\end{center}
\end{table}
\section{Conclusion}
SVM sucessfully predicted breakdown voltages for varying electrode gap and varying
ramp rate based on the electric field features of the oil insulation.
% conference papers do not normally have an appendix
% references section
\bibitem{IEEEhowto:kopka}
H.~Kopka and P.~W. Daly, \emph{A Guide to \LaTeX}, 3rd~ed.\hskip 1em plus
0.5em minus 0.4em\relax Harlow, England: Addison-Wesley, 1999.
\end{thebibliography}