Intelligent“Transformer Faults monitoring system

"using FLC and ANN techniques.

V. T. Barhate Dr. K. L. Thakre Dr. S.S. Limaye
(Dept. of Electrical Engg.) (Prof. in Electrical Engg.) (Prof & Principal)
S.R.K.N.E.C, Nagpur V.N.I.T., Nagpur S.R. K.N. E.C., Nagpur
[email protected] [email protected]

1.INTRODUCTION on the application and the importance of the transformer.

Maintaining the health and reliability of the power Transformers are protected primarily against faults
transformer has been a concern for many years. For this and overloads. The type of protection used should minimize
reason, maintenance engineers would periodically take the time of disconnection for faults within the transformer
transformers and circuit breakers off-line, in order to assess and to reduce the risk of catastrophic failure to simplify
whether the equipment is operating normally. With this eventual repair. Any extended operation of the transformer
method, there are still catastrophic failures. under abnormal condition such as faults or overloads
Transformer is an indispensable part for any power compromises the life of the transformer, which means
system and hence its protection becomes prime necessity adequate protection should be provided for quicker isolation
and the selection of the method used for protection becomes of the transformer under such conditions. Various schemes
a crush. The universally available protection schemes for power transformer protection are:
sometimes fail for the excitation of the transformer at no
load due to the heavy magnetizing inrush current in the 1. Percentage Differential Protection
primary. Hence to avoid the tripping of relay during this 2. Over current protection of transformer
condition, it is necessary to distinguish between the 3. Over-fluxing protection
magnetizing inrush current and internal fault current. A 4. Hottest-Spot Winding Temperature Protection :
majority of researches are being carried out to build an 5. Sensitive ground fault protection to limit
algorithm using wavelet transforms and / or Artificial transformer damage
Neural Networks, Fuzzy Logic Techniques for efficient
2.1 PERCENTAGE DIFFRENETIAL RELAY:
discrimination between magnetizing inrush and internal
The disadvantage of the current differential protection is
faults.
that current Transformers must be identical; otherwise there
Advanced simulation techniques and recently introduced
will be current flowing through the current relays for faults
artificial neural networks with tremendous training
outside of the protected zone or even under normal
capability combined with fuzzy logic approaches to power
conditions. Sensitivity to the differential current due to the
Transformer protection will provide means to enhance the
current transformer errors is reduced by percentage
classical protection principles and facilitate faster , more
differential relays. In percentage differential relays, the
secure and dependable protection for power transformer.
current from each current transformer flows through a
Due to the numerous benefits of digital relaying in terms of
restraint coil. The purpose of the restraint coil is to prevent
economics, performance, reliability and flexibility,
undesired relay operation due to current transformer errors.
significant efforts have been made towards the development
The operating coil current | I1 - I2 | required for tripping is a
of digital relaying algorithms. Numerous algorithms for the
percentage of the average current through the restraint coils.
differential protection of power transformers have been
It is given by
proposed. There are attempts to develop various techniques
Idiff > k (I1 + I2)/2
to detect a magnetizing inrush current using ANN and
Where, k is the proportion of the operating coil current to
Differential protection using Fuzzy logic. Generally, an
the restraint oil. For example if k = 0.1, the operating coil
acceptable protection scheme involves features: reliability,
current must be more than 10% of the average restraint coil
cost, simplicity to use and high speed of operation.
current in order for the relay to operate.
2. OVERVIEW OF PROTECTION SCHEMES
The type of protection of the transformers varies depending

ICACCT, Nov 08th 2008

 The reason for this current rush is to be found in
Fig2.1 CIRCUIT FOR THE PERCENTAGE DIFFERENTIAL RELAY
characteristic shape of magnetism curve of transformer core
2.2 OPERATING CONDITION steel, which is shown in, and from this it will be seen that
As stated above the relay operates when the the no load current at unsymmetrical core flux is increased
differential current exceed some predetermined value. The very high as compared with current under symmetrical core
driving equation for the relay is as :- flux. The initial value of this inrush current is principally
I1 –I2 > B (I1 +I2)/2 determined by the point of voltage wave at which switching
Where, in occurs, but it is also partly dependent on magnitude and
I1= primary current polarity of residual flux, which may be left in the core after
I2=secondary current previous switching out. This residual flux is influence by
(I1+I2)/2 = restraining current transformer core material characteristic, core gap factor,
I1 – I2 = differential current winding capacitance, circuit breaker, chopping
B= bias characteristics and other capacitances connected to the
2.3 MAGNETIZING INRUSH CURRENT transformer. As shown in Fig, below at the instant of
It is often noticed when switching in a no load transformer switching in, if the voltage be zero, the residual flux will be
an initial current rush greatly in excess of magnetization maximum and the peak transient core flux will be more than
current and normal full load current. This may cause twice of normal condition flux and this produces a high
incorrect operation of conventional over current protection magnitude and asymmetrical inrush current.
and fuses. Also produced magnetic force may cause
mechanical damage to transformer winding. This is result of
non-linearity of core magnetization curve. No load
transformer switching may create large asymmetric flux and
saturation of winding core of transformer. This saturation
creates high magnitude asymmetrical current with a high
harmonic content and a high direct current component.

3. BACKGROUND FOR MODEL DESIGN

This paper is an attempt to develop a simulation model in
MATLAB using Fuzzy logic and Neural Network Tool Box
along with Power System Equipments using SIMULINK
which would distinguish between the magnetizing inrush
current and internal fault current to avoid the tripping of
relay .There is a provision for detection of other faults by
simply adding Input signals such as signal through
Temperature sensors etc.A Case study of 50 MVA power
transformer from Wardha City 220/66 KV Substation is

ICACCT, Nov 08th 2008

considered and all its specifications are simulated in
External control of switching times
MATLAB environment.
Breaker resistance Ron = .001 Ω
Block Diagram shown in Fig3.1. exhibits
Snubber resistance Rp = 1 MΩ
components used for simulation of Protection of
Snubber capacitance = 1mF
Transformer under case study. Current transformers on both Current transformers: - A two winding saturable
sides i.e. primary and secondary CT’s are used to obtain transformer has been used with following specifications.
Operating and Restraining currents using circuits for
CT ratio (primary)=10.497/1
calculating Average And Difference of primary and
CT ratio(secondary)=278.24/.577
secondary currents. These Operating and Restraining
Nominal power & frequency = 10 VA,50 Hz
currents are given as an input to Fuzzy Information System
Winding 1 parameters = V1(rms)= 1
(FIS) using Fuzzy Logic Tool Box. FIS model represents pu,R1(pu)=.02,L1(pu)=.08pu
membership functions for input and output. Here Inputs are
Winding 2 parameters=V2
Operating and Restraining currents with fuzzified state (rms)=10.497pu,R2(pu)=.02,L2(pu)=.08pu
developed according to operating and non-operating regions 3 phase RLC series load with Yg connection
of Characteristics of Differential Relay shown in Fig.3.1
Phase to phase voltage = 415 V
The output of FIS is Trip signal for Circuit Breaker which is
Active power = 200 KW
generated by writing several Fuzzy Rules using knowledge Relay Characteristics for operating Current Versus
base of Fuzzy Associated Matrix (FAM). Restraining current is drawn with reference to chosen
The present scheme not only caters for Differential transformer of 50 MVA, CT’s ratio , driving equation I1 –I2
protection of Power Transformer for various internal faults > B (I1 +I2)/2. It is possible to draw the relay characteristics
but also avoids any malfunction of relay due to Magnetizing for different percentage Bias values of B. Here it is drawn
Inrush Current which occurs during excitation of without % Bias. Fuzzy Membership functions for operating
transformer under no load i.e. while putting the Transformer Current and Restraining current are assigned based on this
in service. This is achieved by using Neural Network Tool Characteristic(Fig 4.1). The MATLAB Simulink Model is
Box . The Primary current is taken as sample and Neural as shown in Fig.4.2
network is trained for pattern recognition algorithms .It
recognizes and identifies the Short circuit current and 5. RESULTS
1. No fault condition : Primary current Approx 173 Amps. And Secondary
magnetizing inrush current using ANN and enables the current Approx.600Amps is snown in Figure given below:
Fuzzy Controller only if there is a Short Circuit current
which is generally Sinusoidal in nature. Thus this scheme
avoids the mal operation of Differential Relay by
identifying magnetizing inrush current.
The provision of identifying Hottest-Spot Winding
Temperature using suitable sensors and comparator is kept
and FIS can take care of protecting Transformer since
separate membership function is defined to identify action
to be taken for various temperature ranges.
4. IMPLEMENTATION (SIMULINK MODEL)
4.1 Three phase transformer:-
MVA Rating : 40 / 50 MVA, 50 Hz
Voltage : 220 / 66 kV
Current : 131.2 / 437.38A 2. Three phase-to-ground Fault simulated after 20 msec : Fault current at
Make : Crompton Greaves, Bombay Primary side approx. 3500 Amps. lasted for 1 Cycle and Circuit Breaker
Year of Manufacture : 1991 opened is shown in Figure given below:
Cooling Provided : ONAN : 40MVA
ONAF : 50 MVA
Guaranteed Temp Oil : 500C
Rise : Winding : 550C
Connection Symbol : YN yno
Untanking Mass : 46800 kg
Total Oil : 25900 / 29600 kg
Total Mass : 9700 kg
Heaviest Package : With oil : 79700 kg
Without oil : 59000 kg
HV WT1 CT / 1.8, 1.9, 2.1, 7.5VA, Class 5 connect
Ratio : terminals 1V1S1 & 1V1S3
Circuit breaker : -

Initial status = CLOSED

Switching of all phases

ICACCT, Nov 08th 2008

3. Waveforms of Differential and Restraining current for Three phase-to- 6. Magnetizing Inrush current pattern identification is done. first for
ground fault and second waveform shows Trip signal (active low) after Fault current having sinusoidal current waveforms Neural Network gives
20 msec. is shown in Figure given below: enable to Fuzzy logic and hence Differential Relay is activated only under
fault condition. And secondly when ever there is non-sinusoidal waveform
the Neural network identifies it as a magnetizing Inrush current and
therefore Neural Network does not give enable signal to fuzzy logic and
there is no mal-operation of Relays is shown in figure given below:

4. Line-to-ground fault (in phase A ) Fault current approx. 3400 Amps is

shown in Figure give1n below:

5. External Fault : If the fault is outside the unit protection zone then
Differential Relay does not operate and Fault currents at primary and
secondary are very large is shown in figure given below:
Fig 4.1: Relay Characterstics for operating current Vs Restraining current

REFERENCES
6. Waveforms of Differential and Restraining current for External fault 1. V. T. Barhate & Sangeeta H. Deshmukh, “Neruro Fuzzy based
and second waveform shows Trip signal is not activated is shown in differential protection of transformer”, PECA-IFTOMM 2006,
figure given below: International Conference 12th July 2006.
2. V.T. Barhate & etal., “Fuzzy logic, an alternate tool for protection
against internal faults in transformer”, ICACCT-07,Panipat.
3. S. E. Zocholl, Armando Guzmán, and Daqing Hou, “Transformer
modeling as Applied to Differential Relaying,” Proceedings of the 22nd
Annual Western Protective Relay Conference, Spokane, WA, Oct, 1996.
4. M.C. Shin, C.W.Park and J.H.Kim, " Fuzzy Logic-Based Relaying for
large Power Transformer Protection, " IEEE Transaction on Power
Delivery, Vol. 18, No. 3, pp. 718-724, July 2003.
5. Xu, W., Wang, D., Zhou, Z., and Chen, H. 1997. Fault Diagnosis of
Power Transformers: Application of Fuzzy Set Theory, Expert Systems,
and Artificial Neural Networks. IEEE Proceedings of Science,
Measurement, and Technology, 144(1), pp. 39-44.
6. Khorashadi – Zadeh,H. “Fuzzy – neuro approach to differential
protection for power transformer” TENCON 2004. IEEE 10 Conference
21-24 Nov 2004, PP 279-282

ICACCT, Nov 08th 2008

 Magnetizing Inrush Sample

220 / 66 KV
KKV
Transformer NEURAL Enable
NETWORK
Source
Bus
For Pattern
Feeder I1 50 MVA Recognizing of
I2
Magnetizing FUZZY
220 KV LOGIC
Three Temperature Controller
Phase
(I1+I2) / 2
Restraining Differential
Current
Relay

Characterist
Fault I1-I2
Circuit
Breaker Simulator Operating ics
Current
Trip
Signal
Output

Figure 3.1: Schematic for Transformer protection using Fuzzy-Neuro techniques.

Scope
Ip & Is

Ip

RMS 1.915e-005
Trip RMS 0.5437
com RMS1 Di splay
a R in S1
Ip Is
A A RMS2 Displ ay1
b Y in Rout A a Rin S1
B B
S1 Load
c Bin Y out B b Y in Rout R in
C C
Rout A
3Ph 220kV Bus CB I1 Bout C c Bin Y out Y in
Y out B
Primary CT 's I2 Bout Bin C
50MVA, 220 / 66kV Bout
Power T ransformer A Secondary CT 's
Overload ,
B T emperature rel ay

C
C
A
Internal Fault
B
A1
i C
C
+
-
External Fault
I1

I2

Di ffrenti al Relay
Ires

Idiff

Ires I dif f
Other Faults

y {1} p{1}

Digital Trip
Output Trip Faul t Monitor

Di gital Output I dif f

Fuzzy Logic
St andard Dev ices Controll er
Parallel Port [378h]

I res Continuous
Trip
pow ergui
Idiff / Ires / T rip

VARIOUS PROTECTION SCHEMES OF POWER TRANSFORMER

[ USING FUZZY CONTROLLER & ARTIFICIAL NEURAL NETWORK ]

Figure 4.2: MATLAB Simulation Model

ICACCT, Nov 08th 2008