Project 1

Download as pdf or txt
Download as pdf or txt
You are on page 1of 7

International Electrical Engineering Journal (IEEJ)

Vol. 5 (2014) No.4, pp. 1328-1334


ISSN 2078-2365
http://www.ieejournal.com/

ANN based Discrimination of Inrush


Current and Fault Current in Power
Transformer
P.Ganapathi1, K.Bharanidharan2
1
2

Muthayammal Engineering College,Rasipuram ,Tamil Nadu,India


Muthayammal Engineering College,Rasipuram ,Tamil Nadu,India
1
ganapathip.8@gmail.com,2email2bharani@gmail.com

Abstract- In a power system, transformers and other electrical


equipment have to protect not only from short circuit, but also
from unusual operating conditions, such as over loading, and
dissimilar fault protection. The increased growth in power
systems both in size and difficulty has brought the need for
fast and reliable protection scheme for main equipments like
transformer. The power transformer protective relay should
block the tripping during magnetizing inrush and rapidly
initiate the tripping during internal and external faults. The
concept of magnetizing inrush current and over-excitation
phenomena as they belong to causes of the protection of maloperation. This project work proposes for Identification of
Inrush Current & Fault Current of power transformer for
proper protection scheme. In the proposed algorithm, Feed
Forward Back Propagation Network (FFBPN) are used as a
classifier and address the challenging task of detecting
magnetizing inrush from fault current. The algorithm is
evaluated using MATLAB simulation. The results verify that
the proposed technique is faster, stable and more reliable
recognition of transformer inrush and fault condition.
Keywords- Inrush current, fault current, BPNN, Transformer
protection, differential protection.

I. INTRODUCTION
The power transformer is a part of electrical equipment
that needs continuous monitoring and fast protection
because it is very costly and an essential element for a
power system to perform effectively. Power transformer
internal faults may cause extensive damage And or power
system instability. Thus, differential transformer protection
schemes are used to avoid interruption of the power supply
and catastrophic losses. The most common protection
technique is the percentage differential protection, which
provides discrimination between the fault (internal fault and
external fault) or a normal operating condition. Usually,

differential relays compare the currents from all terminals


of the protected transformer to a predetermined threshold
and, in the case of an internal fault, the equipment is
disconnected from the power supply. However, some
operation conditions can cause differential currents and
they deserve special attention. Some examples of this (with
the presence of magnetizing inrush currents) are
energization, over excitation and sympathetic inrush. In
order to avoid the relay misoperation in these cases, it is
necessary to differentiate inrush current from fault current.
The literature shows that the method based on the harmonic
restraint has been commonly used to solve this problem
[16] [2]. In the case of energization, the second harmonic
component is larger than in a typical fault current. In
modern transformer due to presence of high permeability
core material there is presence of second harmonic content
compare to fault current so based on harmonic tripping
operation may become mal operation [3].
As mentioned before, a sympathetic inrush current also
cause relay mal operation. In order to improve power
transformer protection, various methods were developed for
accurate and efficient discrimination. Here ANN technique
has been applied for power transformer security because
of its potential of highly non-linear mapping feature. But
in these proposed methods the extraction techniques are
based on either time or frequency domain signals, but it
becomes very important to extract both time and frequency
features of the signal for accurate discrimination between
an internal fault and other operating conditions [ 1 1 ] [ 1 3 ] .
II. INRUSH CURRENT
Transformer Inrush current is classified into three
types. That is energization inrush, recovery inrush and
sympathetic inrush. When a transformer is energized a
transient current up to 10 to 20 times larger than the rated
transformer current can last for more than a few cycles.

1328
Ganapathi and Bharanidharan

ANN based Discrimination of Inrush Current and Fault Current in Power Transformer

International Electrical Engineering Journal (IEEJ)


Vol. 5 (2014) No.4, pp. 1328-1334
ISSN 2078-2365
http://www.ieejournal.com/
Decay of inrush current depends on the series resistance on
the transformer winding. In a transformer winding
resistance will damp out the inrush. High inrush happens
when the primary winding is connected at an instant around
the zero-crossing of the primary voltage. When dispersion
occurs for part of half cycles only, then harmonic rich
waveforms can be generated, and can cause lot of problems
to other interrelated equipment.
For large transformers by means of low winding
resistance and high inductance, these inrush currents can
last for several seconds until the transient has died away
(decay time proportional to ~XL/R) and the regular AC
symmetry is established. Generally inrush current contain
d.c. offset, odd harmonics, and even harmonics.

can be used to calculate the peak value of the first cycle of


the inrush current. This equation is as follow

(2)

Where, Vm maximum applied voltage; L air core


inductance of the transformer; R tot resistance of the
transformer; BN normal rated flux density of the
transformer core; BR remnant flux density of the
transformer core; BS saturation flux density of the core
material.
As seen from the equations (1) and (2), the value of
inrush current is dependent to the parameters of transformer
and operating conditions. So a comprehensive analysis for
finding the relations between the inrush current
characteristics and these factors are needed. Factors
influencing the magnitude and duration of magnetizing
current inrush are,
1. Transformer size: as size of transformer increases,
the inrush current also increased.
2. Type of core: if the core is made up of material
having good permeability magnetic then the inrush current
routinely decreased.
3. Residual flux of transformer: presence of residual
flux also increases the magnetizing current.
4. Switching at Instant: if the transformer is switched
on at the instant when the voltage wave is transitory
through zero value, then the magnetizing inrush current at
that inrush will be high.

Fig.1. Generation of inrush current in transformer.

The general equation that for the amplitude of


inrush current as a meaning of time can be expressed as

i(t )

2vm
* kw * ks * (sin(wt ) e
zt

( t t o )

. sin )

(1)

where vm maximum applied voltage; Zt total impedance


under inrush including system; energization angle; t
time; t0 point at which core saturates; time constant of
transformer winding under inrush conditions; function
of t0; Kw accounts for 3 phase winding connection; Ks
accounts for short-circuit power of network.
For the use of scheming a protective system for
transformer, the maximum value of inrush current is an
important factor [4]. In these cases, a simplified equation

III. ARTIFICIAL NEURAL NETWORK


A neural network is a dominant data modelling tool
that is able to capture and signify complex input / output
relationships[1]. The motivation for the progress of neural
network technology came from the wish to develop an
artificial system that could perform Intelligent tasks
related to those performed by the human brains. We look,
see, pay attention and then recognition. Neural network
theory studies both:
pre attentive and attentive
processing of stimuli
The basic unit of neural networks, the artificial
neurons, stimulates the four basic functions of natural
neurons. Artificial neurons are much simpler than the

1329
Ganapathi and Bharanidharan

ANN based Discrimination of Inrush Current and Fault Current in Power Transformer

International Electrical Engineering Journal (IEEJ)


Vol. 5 (2014) No.4, pp. 1328-1334
ISSN 2078-2365
http://www.ieejournal.com/
biological neurons[8]. The belowr fig 2 shows the
fundamentals of an artificial neuron

Fig.2. Artificial Neuron

This section gives the formula used in the model of an


artificial neuron: A single neuron may be written accurately
as mentioned in(3).
y=f (bias+w1*x1+w2*x2+w3*x3++wn*xn)
(3)
Where y is output, x1, x2,xn are the inputs and f(x)
is the activation function. And the formula used for this job
is called the sigmoid.
F(x) = 1/ (1+ exp (-x))
IV.

(4)

BACKPROPAGATION LEARNING

Neural networks are typically organized in layers.


Layers can be made up of number of interconnected nodes
which contain an activation function. Parameters are
presented to the network via the input layer, which
communicate to one or more hidden layers where the actual
processing is done via a system of weighted links. The
hidden layers after that link to an output layer where the
answer is output as shown in below (Fig. 3)

using a two phase propagate adopt cycle. When a Back


Prop network is cycled, the activations of the input units are
propagated forward to the output layer throughout the
connecting weights. Like the perceptron, the net input to a
unit is determined by the weighted sum of its inputs. After
an input pattern has been applied as a motivation to the first
layer of network unit, it is propagated through each upper
layer until an output is generated. This output pattern is
then compare through the desired output sample and an
error signal is computed for each output unit.
The error signals are then transmitted backward from
the output layer to each node in the intermediate layer that
contributes directly to the output. However, each unit in the
intermediate layer receives only portion of the total error
signal, based roughly on the relative contribution the unit
made to the original output. Based on the error signal,
connection weights are then updated by each unit to cause
the network to converge towards a state that allows all the
training patterns to be encoded [5][6][8].
The Back propagation Algorithm (BPA), also called as
the Generalized Delta Rule (GDR), Which provides a way
to estimate the gradient of the error function efficiently
using chain rule of isolation. This paper is concerned with
the delta rule. With the delta rule, as with other types of
back-propagation, learning is a supervised process that
occurs with each cycle or epoch (i.e. in every time the
arrangement is presented with a new input pattern) during a
forward activation flow of outputs, and the backward error
propagation of weight adjustments [6][7]. Within each
hidden layer node is a sigmoidal activation function which
is differentiable and therefore continuous everywhere.
Sigmoid function polarizes network activity and helps
stabilize it.
Yj=1/(1+exp(-Si))

Fig.3. BPNN Architecture

This neural network is mostly trained by Back


propagation algorithm. In Back propagation the network
learns a predefined set of input output examples pairs by

(5)

Back propagation performs a gradient descent within


the solutions vector space towards a Global Minimum
beside the steepest vector of the error space. Learning in a
back propagation network is in two steps. First each
pattern Ip is presented to the network and propagated
forward to the output. Second, a method called gradient
descent is used to minimize the total error on the patterns in
the training set. In gradient descent, weights are malformed
in proportion to the negative of an error derived regarding
each weights. The global minimum is the theoretical
solution with the least possible error. The error itself is a
hyper paraboloid but is rarely smooth.

1330
Ganapathi and Bharanidharan

ANN based Discrimination of Inrush Current and Fault Current in Power Transformer

International Electrical Engineering Journal (IEEJ)


Vol. 5 (2014) No.4, pp. 1328-1334
ISSN 2078-2365
http://www.ieejournal.com/
Fig.4.BPNN Flowchart

V. CIRCUIT EXPLANATION & ANALYSIS


All the three circuit breakers are not blocked at the
same instant rather they are closed when the relevant
voltage waveforms are enchanting their maximum value in
order to reduce the inrush effect. The 3 fault is applied
for 5 cycles i.e. from 25/60 cycles to 30/60 cycles (0.4s0.5s). The Circuit Breakers are closed at their first
switching timings. For the first switching timing 25/60
(phase A) the energy source is now connected to the power
transformer through CBs. Now a 3 fault is applied.
From the current measurement, currents through all the
three lines are tapped to MATLAB work space.

Fig.5.Simulink representation of

3 Transformer

A. TRAINING PERFORMANCE BY ANN

1331
Ganapathi and Bharanidharan

ANN based Discrimination of Inrush Current and Fault Current in Power Transformer

International Electrical Engineering Journal (IEEJ)


Vol. 5 (2014) No.4, pp. 1328-1334
ISSN 2078-2365
http://www.ieejournal.com/

Fig.6 Normal current Training Performance

Fig.8.Line-Ground Fault Current Training

Fig.7.Inrush Current Training Performence

Fig.9.Line-Line Fault Current Training Performance


Fig.10.Waveform of Inrush & Fault current from phase a

B. OUTPUT CURRENT DISCRIMINATION BY


FFBPN TOPOLOGY WITH MATLAB

Fig.11.Waveform of Fault current from phase b

1332
Ganapathi and Bharanidharan

ANN based Discrimination of Inrush Current and Fault Current in Power Transformer

International Electrical Engineering Journal (IEEJ)


Vol. 5 (2014) No.4, pp. 1328-1334
ISSN 2078-2365
http://www.ieejournal.com/
to improve the performance & rectify error by
analyzing of Mean, Standard Deviation, Maximum
Norm & THD standards are taken from the inrush
& fault current. The facility of the new scheme will
express by simulating various cases on a distinctive
power system and the proposed algorithm is also
tested data collected from a simulation model. In
this new algorithm is performance was very crucial.
The proposed approach is considered a general tool
because it can be easy implemented on the
fashionable MATLAB software.

REFERENCES

Fig.12. waveform of steady state current (after clearing fault)

VI.

RESULTS AND DISCUSSION

From the figures (Fig 6, 7, 8, 9) we


observe the fault and inrush current. From the ANN
performance (V-A),the normal current reach the
best value in very few test iteration. And then faults
are takes more test iterations than normal current
value.
The MATLAB waveform shows (V-B)
that inrush is very short period and which is so
peak. And the fault current is in 0.4s-o.5s as
smaller than inrush and differs from steady state
current. In the fig (12) shows the steady state
current wave which is appears after clearing the
fault. From the section V(A,B) the classification of
waves and magnitudes, duration of faults are
noticeable.

VII. CONCLUSION
This work presents a new approach in
differential protection for power transformer vastly
enhanced
performance
over
conventional
techniques. The conventional harmonic control
technique may fail because high second harmonic
components are generated through faults and low
second-harmonic components are generated during
magnetizing inrush with such high permeability
core materials of power transformer. In this paper

[1] T. Shanthi, A.S.Vanmukhil Photovoltaic Generation


System with MPPT Control Using ANFIS IEEJ Trans vol 04
no 03 pp 1105-1115
[2] Myong-chul Shin, member, IEEE, Chul-Won Park, and
Jong-Hyung Kim, Fuzzy Logic-Based Relaying For Large
Power Transformer Protection IEEE Trans.Power Del, Vol.18,
No.3, July.2003
[3] X.Lin, J.Huang, L.Zeng and Z.Q. Bo, Analysis of
Electromagnetic Transient and Adaptability of SecondHarmonic Restraint Based Differential Protection Of UHV
Power Transformer, IEEE Trans. Power Del., vol.25, no.4, pp.
2299-2307, Oct.2010.
[4] M.Jamali, M.Mirzaie and S.AsgharGholamian, Calculation
and Analysis of Transformer Inrush Current Based on
Parameters of Transformer and Operating Conditions, ISSN
1392-1215 electronics and electrical engineering.
[5] SRParaskar, M.A.Beg and G.M.Dhole Discrimination
between Inrush and Fault in Transformer: ANN Approach,
International Journal of Advancements in Technology ISSN
0976-4860.
[6] Mudita Mathur, Vinay Barhate ANN based Technique for
Discrimination between Magnetizing Inrush and Internal Fault
Currents in Transformer,
International Journal of Science and Research (IJSR), India
Online ISSN: 2319-7064.
[5] D. Q. Bi et al., Correlation analysis of waveforms in non
saturation zone-based method to identify the magnetizing inrush
in transformer, IEEE Trans. Power Del. , vol. 22, no. 3, pp.
13801385, Jul. 2007.
[6] . L. Orille - Fernndez, N. K. I. Ghonaim, and J. A.
Valencia, A FI-RANN as a differential relay for three phase
power transformer protection, IEEE Trans. Power Del. , vol.
16, no. 2, pp. 215218, Apr. 2001.
[7] Application Manual of RE 521*2.3 (Differential
Protection Terminal),ABB Relay Catalogue-1MRK 504 021UEN.[On-line]. Available: www.abb.com
[8] E. C. Segatto and D. V. Coury, A differential relay for
power trans-formers using intelligent tools, IEEE Trans. Power
Syst., vol. 21, no.3, pp. 11541162, Aug. 2006.
[9] P. L. Mao and R. K. Aggarwal, A novel approach to the
classification of the transient phenomena in power transformers
using combined wavelet transform and neural network, IEEE
Trans. Power Del. , vol.16, no. 4, pp. 654660, Oct. 2001.
[10]G. Baoming, A. Almeida, Z. Qionglin, and W. Xiangheng,
An equivalent instantaneous inductance-based technique for
discrimination between inrush current and internal faults in

1333
Ganapathi and Bharanidharan

ANN based Discrimination of Inrush Current and Fault Current in Power Transformer

International Electrical Engineering Journal (IEEJ)


Vol. 5 (2014) No.4, pp. 1328-1334
ISSN 2078-2365
http://www.ieejournal.com/
power transformers, IEEE Trans. Power Del., vol. 20, no. 4,
pp. 24732482, Oct. 2005
[11]C. Cummins, D. C. Yu, and L. A. Kojovic, Simplified
artificial neural network structure with the current transformer
saturation detector provides a good estimate of primary
currents, in Proc. IEEE Power Eng. Soc. Summer Meeting,
2000, vol. 3.
[12] J. Ma, Z. Wang, Q. Yang, and Y. Liu, Identifying
transformer inrush current based on normalized grille curve,
IEEE Transactions on Power Delivery, vol. 26,no. 2, pp. 588
595, 2011.
[13] S. A. Saleh and M. A. Rahman, Testing of a wavelet
packet-transform based differential protection for resistance
grounded three-phase transformers, IEEE Trans. Ind. Appl.,
vol. 46, no. 3, pp. 11091117,May/Jun. 2010.
[14] R. A. Turner and K. S. Smith, Resonance excited by
transformer inrush current in inter-connected offshore power
systems, in Conf. Rec. IEEE IAS Annu. Meeting, Edmonton,
Canada, Oct. 2008.
[15] G. B. Kumbhar and S. V. Kulkarni, Analysis of
sympathetic inrush phenomena in transformers using coupled
field-circuit approach, in Proc. IEEE Power Eng. Soc. General
Meeting 2007, Jun. 2007, pp.16.
[16] Ali Hooshyar, Majid Sanaye-Pasand, Seneior member,
IEEE, Saeed Afsharnia, Mahdi Davarpanah, and Bashir Mahdi
Ebrahini Time-Domain Analysis Of Differential Power Signal
To Detect Magnetizing Inrush In Power Transformers,IEEE
Trans.Power Del., Vol. 27, no.3, July.2012.

1334
Ganapathi and Bharanidharan

ANN based Discrimination of Inrush Current and Fault Current in Power Transformer

You might also like