Prediction Point of Fault Location On Its Campus P

IOP Conference Series: Materials Science and Engineering
PAPER • OPEN ACCESS
Prediction Point of Fault Location on Its Campus Power Grid by using

Neural Artificial Method
To cite this article: A Syamsudin et al 2021 IOP Conf. Ser.: Mater. Sci. Eng. 1096 012068
View the article online for updates and enhancements.
This content was downloaded from IP address 213.182.203.163 on 18/03/2021 at 14:13

ICIMECE 2020 IOP Publishing
IOP Conf. Series: Materials Science and Engineering 1096 (2021) 012068 doi:10.1088/1757-899X/1096/1/012068
Prediction Point of Fault Location on Its Campus Power Grid

by using Neural Artificial Method
A Syamsudina*, I Hafidza,b, D Rahmatullahc, D A Asfania, I M Y Negaraa
a
Electrical Engineering Departement, Institut Teknologi Sepuluh Nopember, Surabaya, 60111
Indonesia
b
Electrical Engineering Departement, Institut Teknologi Telkom Surabaya, 60231 Indonesia
c
Electrical Engineering Departement, Universitas Hang Tuah, Surabaya 60111 Indonesia
* Corresponding author. E-mail address: [email protected]
Abstract The dispersed electric load connected to the power system leads to various nominal
current and direction. However, to protect the plan optimally, relay settings must be updated
according to its configuration. This paper investigates the prediction of fault location point for
Directional Overcurrent Relay (DOCR). The system used Institut Teknologi Sepuluh Nopember
(ITS) Campus electricity system connected to the grid utility. Artificial Neural Network (ANN)
include data combination of power flow and short circuit as input data, can determine the
appropriate fault location of the system. From the result, the data set in the master control has a
smaller composition than by performing manual looking for tables. From the simulation result,
313 testing data obtained an average error of 0.002614377 so that the test results are quite close
to the target data. Another advantage is that fewer data must be entered in Master control when
using ANN, 136 data, compared using a lookup table, 1512 data. Through this method, the user
can predict the fault location quickly and accurately.
1. Introduction
It is necessary to analyze the settings and coordination of the relay, especially in the coordination of
overcurrent protection relay to improve the reliability of the protection system. Update relay setting and
good relay coordination can prevent damage network and electrical equipment in the event of a fault
and restrict the power supply from being cut off in other areas.
Protection coordination has a vital role in guaranteeing continuity of energy sources, so in the design
of protection must be guaranteed the reliability, selectivity, flexibility, and speed of the relay operation
to isolate the affected area without affecting the affected other regions. Some of the impacts can affect
changes to short circuit breakdowns and errors in the termination network. Therefore, new protection
scheme is carried out and calculated by considering the generator location by determining the time dial
setting and the current pickup so that it can have minimum total operating time for relay. The conditions
of each network topology that occur next specify the value of each relay setting related to topological
terms. Coordination can be done using a hierarchical control system (hierarchical control) where there
are several levels in the settlement process the problem by processing, sensing and adjusting, monitoring
and supervision, using devices to find out.
One of the difficulties that occurred during the protection scheme study was the determination of the
primary relay operating time and backup relay. The backup relay must work as fast as possible to back
Content from this work may be used under the terms of the Creative Commons Attribution 3.0 licence. Any further distribution
of this work must maintain attribution to the author(s) and the title of the work, journal citation and DOI.
Published under licence by IOP Publishing Ltd 1
up the central relay by taking into account the CTI (Coordination Time Interval) parameter so that the
arc flash sparks interference can be minimized. This condition explains that the failure in the protection
strategy is caused by the results of the operating time of the backup to the primary relay does not include
the exact CTI value.
Artificial Neural Network is an algorithm that has several things to improve the optimization process
as a solution to the problem of the high level of difficulty of coordinating protection in a Loop and
Radial system. These are the velocity parameters and the best position exchange of each objective
function.
Several methods are used to optimize some protection parameters to reduce difficulties in the study
Loop and Radial systems. Optimized parameters include the TDS (Time Dial Setting) value, CTI
(Coordination Time Interval), and operating time for each primary and backup relay. Some of the ways
that have been done to optimize the parameters are LP (Linear Programming), Artificial Neural
Network, NLPP (Nonlinear Programming Problem), GA (Genetic Algorithm), and EA (Evolutionary
Algorithm). Artificial Neural Network (ANN) is an algorithm that functions to be more adaptive in
detecting interference and changing combinations in a Loop and Radial system.
2. Fundamental of Power System
2.1. Fault Modelling

When an interruption occurs in a system will flow a large current leading to the point of interference.
The interference current has a value higher than the maximum allowable ray current, so there is an
increase in temperature on the equipment that can cause equipment damage. Disturbances that often
occur in electric power systems are overload and short circuit.The three-phase short circuit is a short
circuit that involves the three-phase. The two-phase short circuit is a short circuit that occurs between
two-phase without being connected to the ground. One-phase short circuit affects a zero sequence
impedance, and the magnitude of this short-circuit current depends on the grounding system used. The
amount of each phase short circuit current above can be calculated respectively by the following
equation:
𝑉
𝐼𝑠𝑐 3∅ = 𝑋𝐿𝑁 (1)
1
𝑉𝐿𝐿 √3 𝑥 𝑉𝐿𝑁
𝐼𝑠𝑐 2∅ = = (2)
𝑋1 + 𝑋2 2 𝑥 𝑋1
3 𝑥 𝑉𝐿𝑁
𝐼𝑠𝑐 1∅ = (3)
𝑋1 + 𝑋2 + 𝑋0+ 3 𝑥 𝑍𝑔
Where𝑉𝐿𝑁 is line to neutral nominal voltage, 𝑋0 is zero reactance,𝑋1 is a positive reactance sequence
while 𝑋2 is a negative reactance sequence. Overload problems occur when the electric current flowing
in the power supply system is greater than the nominal current permitted through the channel (I> In). If
this fault occurs, then the current passing through electrical appliances, such as transformers, generators,
motors, and other electrical appliances), will exceed the current of the equipment capacity. This
condition will damage the equipment.
2
Generator
Utility System
Transformator
Switchgear
Fault Location
M M
Synchronous Induction Motor
Motor
Figure 1. Direction of current flow when a short circuit occurs
2.2. Overcurrent Relay

The overcurrent relay is a relay that operates when the current flowing in a power system conduit
exceeds the specified current value constraint. The relay will work when If> Ip, and it will not work
when If<Ip. Where Ip represents the value of current expressed to the secondary winding CT (Current
Transformer) and If represents the rated fault current.
t(s)
Extremely
inverse
Very inverse
inverse
I(A)
Figure 2. Characteristic of Standard Inverse, Very Inverse dan Extremely Inverse Curve
The overcurrent tuning setting limit is that the releases do not work at the maximum load. Therefore,
the current setting must be greater than the maximum load current. The limits of setting value is 1.05-
1.3 I_FL. In the overcurrent relay, the magnitude of the pickup flow is determined by selecting the tap
value using the following equation:
Tap = Iset / CT primary (4)
Besides that, to adjust the size of the tap also requires setting the dial time. Setting the time value for
each inverted relay curve can use equation (5) and Table V.
𝑘×𝑇
td = 𝐼 ∝ (5)
𝛽 × [(𝐼𝑠𝑒𝑡) −1]
Where td is Time operation (second), T is Time dial, I is Current value (Ampere), Iset is Pickup Current
(Ampere), k is inverse coefficient 1 , α inverse coefficient 2, and β is Coefficient inverse 3.
Table 1. Coefficient of Relay Curve
3
Curve Type Coefficient

k  
Standard Inverse 0,14 0,02 2,970
Very Inverse 13,50 1,00 1,500
Extremely Inverse 80,00 2,00 0,808
The overflow relay of the instant part will act if there is more current flowing beyond the allowable
limit. In determining this instant pickup setting is used Isc min. It represents as two-phase short circuit
current at minimum generation. So, the settings are set:
Iset ≤0,8 x I_ (sc min) (6)
Coordination of relays is performed to avoid miscoordination, such as trips simultaneously. In
accordance with the IEEE 242 standard on the time difference between the primary relay and backup
relay, the allowable Δt is 0.2 - 0.5 seconds. With the following equipment specifications, CB opening
time: 0.04 - 0.1s (2-5 cycles); Overtravel from Relay: 0.01 s; and Safety factor: 0.12 - 0.22 s. For the
microprosessor-based relay, Overtravel time from relay can be ignored. Thus, the required time is 0.2-
0.3s.
2.3. Directional Relay

Directional relay is needed where the relay current usually does not work selectively. This equipment is
made just like conventional current, with the addition of direction elements that can determine the
direction of the fault. Directional elements work based on the phase shift between the magnitude of the
polarization and the magnitude of the operation. For forward condition, I (current) lagging against V
(voltage), and for a reverse condition, I lead to V.
3. Research Methodology
3.1. Campus Power Grid

In general, the campus electricity section can be described as follows. The southern ring ITS electricity
network at ITS Sukolilo Campus consists of 5 Sub Stations (SS) namely SS-1 Electrical Engineering
that supplies 800 kVA transformers, SS-2 Physics Engineering that supplies 630 kVA transformers, SS-
3 Chemical Engineering that supplies 630 kVA transformers, SS-4 BAUK that provides 630 kVA
transformers and FMIPA SS-5 that supplies 630 kVA transformers.
● Lump load 1: Sub Station 1 in Electrical Engineering Faculty with capacity 0.717 MVA,
● Lump load 2: Sub Station 2 in Engineering Physics Faculty with capacity 0.25 MVA,
● Lump load 3: Sub Station 3 in Chemical Engineering Faculty with capacity 0.329 MVA,
● Lump load 4: Sub Station 4 in General Support Division with capacity 0.225 MVA,
● Lump load 5: Sub Station 5 in Science and Statistics Faculty with capacity 0.271 MVA.
4
Figure 3. ITS Campus Electrical System

In the southern ring power network, the ring network is indeed designed, so that it is easy to maintain
the electricity network and quickly solve the problem if there is damage to one of the networks or
substations. However, with the aging of the electricity network, there are electrical disturbances in
several substations, and the safety equipment in each Sub Station uses only the Direct Switch on the
incoming Cubicle and the fuse with SF 6 gas in the Out Going Cubicle. After simulating the operation
of the southern ring network system on the ETAP program, the values of short circuit current and Power
flow are obtained. After that, several system configurations, namely: Loop and Radial systems, with
various kinds of the fault where impedance 0; 0.1; 1; 10; 100 are used to represent fault type.
3.2. Artificial Neural Network

Artificial Neural Network (ANN) is a network of a group of small processing units that are modelled
from human neural networks. ANN has an adaptive system so that its structure can change to solve
problems based on information flowing through the network. Also called ANN is a tool to model non-
linear statistical data.
Figure 4. Basic Concept of Neuron abd Architecture of Neural Network

This is very useful for complex modelling relationships between inputs and outputs to find data
patterns.Fig.4 describes Neural Network architecture, which consists of Neurons as a basic part of
5
processing. The input used both during learning and in recognizing an object. Weight, the burden that
always changes every time, given input as a learning process. Processing Unit, the place where an object
recognition process takes place based on the loading given. Output comes from the introduction of an
object.
Backpropagation is a supervised learning algorithm. They are generally used by perceptron with
layers to change the weights associated with neurons. The hidden layer of the backpropagation algorithm
changes the output error to change the values of the weights in the backward direction. The advance
propagation stage must be done first to obtain an error. At the time of propagation forward, neurons are
activated by using the activation function that can be differentiated as follow.
1
𝑦 = 𝑓(𝑥) = 1+𝑒 −𝜎𝑥 (7)
Where
𝑓′(𝑥) = 𝜎𝑓(𝑥) {1 − 𝑓(𝑥)} (8)
Multiple neurons can be combined in layers to form networks, and network architecture can contain
one or more layers. Given the connectivity of neurons in the network, this combination can be divided
into two types: Feedforward network, Recurrent Network or Feedback Network.
The architecture used in this study is a feedforward network with new instructions, which consist of
2 layers: one input layer, one hidden layer, and one output layer. The learning process, in neural
networks, is a procedure to modify the weights and biases of the network. The training algorithm forces
the network to produce specific responses to specific inputs. There are two types of learning rules as
Supervised Learning and Unsupervised Learning. This study uses supervised learning as a learning
method. This method is monitored and provided with input-output data sets (called data training) of
appropriate network behaviour. As input is applied to the network, the network output is compared to
the target output. Learning rules are used to adjust network weights and biases, so that network output
is closer to the target.
4. Simulation Results
The results of the calculation and modeling are the result of a combination of generation inputs that will
get the value of the pickup current and time dial setting, the distribution system with the generator is
divided into 2 generation topologies. Topology 1: Utility PLN electricity source, with Ring system
configuration, Topology 2: Utility PLN electricity source, with Radial system configuration. The
operating scheme for the transformer is varied as described follow. Scheme 1: All transformers operate,
and Scheme 2 to 7, only one transformer does not operate respectively from transformer 1 to 6.
Table 2. Power Flow Calculation in Campus Electrical System
Bus Number Configura
1 2 3 4 5 6 7 8 9 10 11 12 13 tion
1.7 0.2 -1.1 -3 -4.5 -8.6 8.6 4.5 3 1.1 -0.2 -1.7 10. 0
3
10-5 -1.6 -2.9 -4.7 -6.2 - 10.3 6.2 4.7 2.9 1.6 10-5 10. 1
10.3 3
1.6 10-5 -1.3 -3.2 -4.6 -8.8 8.8 4.6 3.2 1.3 10-5 -1.6 10. 2
3
10-5 10-5 10-5 -1.9 -3.3 -7.5 7.5 3.3 1.9 10-5 10-5 10-5 10. 3
3
4.7 3.2 1.9 10-5 -1.4 -5.6 5.6 1.4 10-5 -1.9 -3.2 -4.7 10. 4
3
6.2 4.6 2.3 1.4 10-5 -4.1 4.1 10-5 -1.4 -2.3 -4.6 -6.2 10. 5
3
6
10.3 8.8 7.5 5.6 4.1 10-5 10-5 -4.1 -5.6 -7.5 -8.8 10.3 10. 6
3
The results of the Power Flow recapitulation of all system configurations at Campus Electric System
is examined in the calculation, as showed in Table2. Each relay has its own current data, which functions
to secure the system. There is a target with a value of 1 to present the location of the disturbance, i.e.,
on Bus 1 in Table 3. read the value of the relay short circuit currents 1 to 13 wherein the Target column
is location bus disturbance. Data input into the LookUp Table for ANN training and testing process can
be explained as follows. There are six channels configurations, each with a different relay current.
Table 3. Sample Current Groupings Calculation in Power Grid
Relay Targe
t
1 2 3 4 5 6 7 8 9 10 11 12 13
0.00 3 6 10 13 22 -22 -13 -10 -6 -3 -0.001 270 1
1 0
0.00 3 6 10 13 22 -22 -13 -10 -6 -3 -0.001 268 1
1 0
0.00 3 6 10 13 21 -21 -13 -10 -6 -3 -0.001 256 1
1 0
Table 4. Recapitulation of MSE Data from 15 Neurons
Neuron Training MSE Neuron Training MSE
1.1 0.29662 9.1 0.000099141
1 1.2 0.18644 9 9.2 0.00032123
1.3 0.48236 9.3 0.00020867
2.1 0.13605 10.1 0.00012695
2 2.2 0.000233 10 10.2 0.00034181
2.3 0.000297 10.3 0.00012468
3.1 0.000311 11.1 0.00026662
3 3.2 0.000223 11 11.2 0.00015278
3.3 0.000158 11.3 0.00021889
4.1 0.000443 12.1 0.00030035
4 4.2 0.000189 12 12.2 0.00025893
4.3 0.000171 12.3 0.0003603
5.1 0.00042 13.1 0.0041437
5 5.2 0.00016 13 13.2 0.00016305
5.3 0.00018 13.3 0.0086157
6.1 0.000157 14.1 0.00021459
6 6.2 0.00061 14 14.2 0.00017118
6.3 0.000373 14.3 0.0003378
7.1 0.000274 15.1 0.00031087
7 7.2 0.000576 15 15.2 0.00037695
7.3 0.000199 15.3 0.00022398
8.1 0.000141
8 8.2 0.000135
8.3 0.000319
7
The type of interphase fault is represented by an impedance of 0.1 and 1 because there are likely to
be significant differences in the representation of ground fault. 0 means no fault. In this discussion there
are 12 possible locations of disturbances on the bus, with details of 1 bus there are two possibilities of
2 areas of disturbance, so 6 buses multiplied by 2 fault locations. The incoming currents from relays 1
to 12 connected to the channel are entered sequentially, while relay 13 is a relay owned by PLN. The
calculation of error data is grouped in tables 4.5. Each neuron trained the neuron results, experiments,
and the mean square error of the training. The division of relay settings based on this source is used as
an accurate calculation for all loading conditions.
Figure 5. Target results compared
Figure 6. Best Training Performance Neuron Error 9
8
Figure 7. Mean Square Error Neuron 9

Figures 5, 6, 7 explain the Figure of the results of training ANN Neuron 9 and Mean Square Error
Neurons 9. Circle o symbolizes actual data, and (*) symbolizes the results of neural network training.
The x-axis is the amount of current data processed, and the y-axis is the number of buses as many as 12
buses. The MSE value represents the mean square error. The x-axis describes the target of the training,
while the y-axis describes the training output. Figure 4.19 represents. Referring to the ANN simulation,
the best performance was obtained for 9 neuron data, namely 0.000099141 on the iteration of epoch
500, and the results of training calculations R = 0.99995.
5. Conclusion
Based on the simulation results and prediction analysis, Artificial Neural Network can simplify
identification and accelerate the prediction fault points in the ITS electrical system. In this process, data
form power flow and short circuit is trained three times, starts from the 1st neuron to the 15th neuron, to
obtain the smallest error. The calculation results obtained MSE value of the smallest neuron at number
9 with trial 1. It means, the best performance was obtained for 9 neuron data, namely 0.000099141 with
epoch iteration 500 and R 0.99995. The testing process is done by using 20% data, and this is done to
prove the value of Weight and Bias can determine the location of the system with different data. From
313 testing data, obtained an average error 0.002614377 approaching the target data. In this advantage,
Master control can use smaller amount of data to be processed. In detail, data that is needed by Master
control with the Artificial Neural Network is 136 data, while lookup table is 1512 data.
Acknowledgments
The author would like to thank the Institut Teknologi Sepuluh Nopember Surabaya Campus for
helping funds and always supporting him while studying
References
[1] Hussain, M. H., Rahim, S. R. A., & Musirin, I. “Optimal overcurrent relay coordination: A
review”. Procedia Engineering, 53, 332–336. 2013.
[2] Anderson, P.M, “Power System Protection”. Canada: Jhon Wiley & Sons, Inc, 1998.
[3] Sanderson, J. V. H. “Book Review: Recommended Practice for Protection and Co-Ordination of
Industrial and Commercial Power Systems”. International Journal of Electrical Engineering
Education (Vol. 15). 1978.
9
[4] Dugan, R. C., & McDermott, T. “Distributed generation operating conflictsfor distributed
genration for interconnected utility distribution system”. Industry Applications Magazine,
IEEE, 8(2), 19–25. 2002.
[5] Ackermann, T., Andersson, G., & Söder, L. “Distributed generation: A definition. Electric Power
Systems Research” 57(3), 195–204. 2001.
[6] M. Negnevitsky and V. Pavlovsky, “Neural Networks Approach to Online Identiﬁcation of
Multiple Failures of Protection Systems,” IEEE Trans. Power Del., vol. 20, pp. 588–594, Apr.
2005
[7] M. Biswal, S. M. Brahma, and H. Cao, “Supervisory protection and automated event diagnosis
using PMU data,” IEEE Trans. Power Del., vol. 31, pp. 1855–1863, Aug 2016.
[8] P. Kundu and A. K. Pradhan, “Enhanced protection security using the system integrity protection
scheme (SIPS),” IEEE Trans. Power Del., vol. 31, pp. 228–235, Feb. 2016.
[9] Pico, J. D., Celeita, D., & Ramos, G. “Protection Coordination Analysis under a Real-Time
Architecture for Industrial Distribution Systems Based on the Std IEEE 242-2001”. IEEE
Transactions on Industry Applications, 52(4), 2826–2833. 2016.
[10] Alam, M. N., Gokaraju, R., & Chakrabarti, S. “Protection Coordination For Networked
Microgrids Using Single And Dual Setting Overcurrent Relays”. IET Generation,
Transmission and Distribution, 14(14), 2818–2828. 2020.
10

Prediction Point of Fault Location On Its Campus P

Uploaded by

Copyright:

Available Formats

Prediction Point of Fault Location On Its Campus P

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Prediction Point of Fault Location On Its Campus P

Uploaded by

Copyright:

Available Formats

IOP Conference Series: Materials Science and Engineering

PAPER • OPEN ACCESS

Prediction Point of Fault Location on Its Campus Power Grid by using

View the article online for updates and enhancements.

This content was downloaded from IP address 213.182.203.163 on 18/03/2021 at 14:13

Prediction Point of Fault Location on Its Campus Power Grid

A Syamsudina*, I Hafidza,b, D Rahmatullahc, D A Asfania, I M Y Negaraa

2. Fundamental of Power System

2.1. Fault Modelling

Figure 1. Direction of current flow when a short circuit occurs

2.2. Overcurrent Relay

Curve Type Coefficient

2.3. Directional Relay

3.1. Campus Power Grid

Figure 3. ITS Campus Electrical System

3.2. Artificial Neural Network

Figure 4. Basic Concept of Neuron abd Architecture of Neural Network

𝑓′(𝑥) = 𝜎𝑓(𝑥) {1 − 𝑓(𝑥)} (8)

Figure 5. Target results compared

Figure 6. Best Training Performance Neuron Error 9

Figure 7. Mean Square Error Neuron 9

You might also like