Recognition and Simulation of Electromagnetic

Transients in Power Systems of Southern Vietnam

Huu Phuc Nguyen Nhan Bon Nguyen, Huy Anh Quyen,
Department of Electrical and Electronic Engineering, Ho Chi Department of Electrical and Electronic Engineering, Ho Chi
Minh City University of Technology Minh City University of Technical Education
Ho Chi Minh city, Vietnam Ho Chi Minh city, Vietnam
[email protected] [email protected]; [email protected]

Abstract—This paper presents a research of electromagnetic

transients in power systems based on the three issues: I. INTRODUCTION
identification of voltage transient disturbances, analysis of power The work performed in this research is focused on
system transient, and modeling of distributed-parameter transient phenomena recognition (using Wavelets technique
transmission lines. The first issue includes (a) recognizing the and artificial intelligent techniques), power system transient
presence of power quality disturbances in supply voltage and (b)
simulation for prediction of voltage and current values and
classifying any existing deviations from a particular type. Various
distributed-parameter transmission lines modeling. Therefore,
techniques using artificial neural networks and Fuzzy associative
memories, Neuro-Fuzzy systems based on Wavelet transform are the focus of this work mainly concentrates on the following
used to identify power quality disturbances. The second issue is to topics:
analyze the power system transients using the wavelets technique x Identification of power system transient phenomena
as opposed to the trapezoidal rule method. A novel developed
algorithm has the following properties: 1) it is time-domain x Simulation and analysis of power system transients
working, so that circuit nonlinearity can be handled and
numerical accuracy can be well controlled, unlike the frequency- x Modeling distributed-parameter transmission lines
domain method where numerical error may become uncontrollable The paper includes 6 sections, in which section I is
during the inverse L a p l a c e transform; 2) the wavelets property introduction, section II is Identification of transient
of localization in both time and frequency domains makes a
phenomena. Modeling of distributed-parameter transmission
uniform approximation possible, which is generally not achieved
in time-stepping methods. The last issue is to model a
lines is presented in section III, while section IV is about
transmission line based on passive network synthesis for simulation and analysis of typical power systems. Section V is
representing the nonlinear frequency dependence of the the application of the developed method in a part of Vietnam
characteristic impedance matrix function. The synthesized power system. The Conclusion is in section VI.
passive network of interconnected RLC elements preserves at
any frequency the inherent stability property of this matrix II. IDENTIFICATION OF POWER SYSTEM TRANSIENTS
function for any power transmission circuits. The network
parameters are identified using the constrained optimization by In this paper, a comprehensive comparison of various
which the synthesized network response is matched with the techniques based on the use of wavelets technique and artificial
actual response derived from the characteristic impedance intelligences such as Probabilistic Neural Network, Fuzzy
matrix. With the trapezoidal rule of numerical integration Logic and Neuro-Fuzzy method is made.
applied to individual elements of the synthesized network,
discrete time-domain recursive sequences are derived, and then A. Multi-Resolution Analysis Technique
included in the overall transmission circuit model. The network The first main characteristic in Discrete Wavelet Transform
synthesis procedure developed is applied to a representative
(DWT) is the Multi-Resolution Analysis (MRA) technique
transmission line, and comparative studies indicate high accuracy
that can decompose the original signal into several other
being achieved. The research is then applied and validated to a
southern of Vietnam power system in study.
signals with different levels (scales) of resolution (Fig.1).
From these decomposed signals, the original time-domain
Index term—Power system transient, Electromagnetic transient signal can be recovered without losing any information.
analysis, passive circuits, power system, transmission circuit, z-
transform, artificial neural networks and Fuzzy associative The recursive mathematical representation of the MRA is as
memories, Neuro-Fuzzy, wavelet transform, Multi-Resolution follows:
Analysis
u J w J 1 † u J 1 w J 1 † w J 2 † ˜ ˜ ˜ † w J n † u J n (1)

978-1-4673-4584-2/12/$31.00 ©2012 IEEE 556 IPEC 2012

Where: implementation time is of 0.45(s), outputs are types of
uJ+1: approximated version of the given signal at scale J+1. disturbances.
wJ+1: detailed version that displays all transient phenomena of
the given signal at scale J+1. B.2. Fuzzy Logic System
† : denotes a summation of two decomposed signals. Time FIS has two properties: YES and NO; Amplitude FIS
has six properties of disturbance: sag, swell, interruption, pure
N: the decomposition level. sin, capacitor swichings; Output FIS gives the results; and 22
rules will classify 11 voltage disturbances.
Employing the DWT technique to analyze the energy of the
distorted signals, time and amplitude of the distorted signal is Fuzzy Logic System has the following characteristics:
presented in [4, 5]. implementation time is of 0.90 (s), outputs are types of voltage
disturbances.

Fig.1 Two Decomposed Filters of Wavelet Transform

Fig.3 Outputs (11 types of disturbances)

B. Results and Discussion
There are 241 samples of disturbances under consideration,
in which sudden sags, sudden swells, interruptions, harmonics, B.3. Neuro - Fuzzy Logic System
flickers, capacitor switchings, pure sin. These samples of This classifier system has four inputs such as high
capacitor switchings have 5 types: prestrike, restrike, isolated frequency FIS, low frequency FIS, time FIS, amplitude FIS.
capacitor, magnification, back to back switching. All of the Results shows that it is as efficient as other classifiers
samples are trained in AI systems. The results are detailed as mentioned above. Neuro-Fuzzy Logic System has the
follows. following characteristics: learning time is of 2 hours,
implementation time is 1.2(s), The outputs are types of voltage
B.1. Probabilistic Neural Network disturbances.
Signal

DWT - 13 levels
w1,w2,…,w13

Disturbance Duration Energy Distribution

tt , A P1D,P2D,…,P13D

PNN : Probabilistic Neural Network

Hidden Layer
H1
Input Layer Output Layer
X1 H2 Y1

X2 H3 Y2
H4
Xi Yj

Radial Basis Layer Hh Competitive Layer

1 6 2 3 7 4 5

Capacitor Harmonics Voltage Voltage Pure Sin Inter- Flicker

Switching Sag Swell ruption

Fig.2 Diagram of the proposed classifier

Inputs are 15 values of disturbances: 13 levels of energy Fig.4 Output FIS Membership
distribution, transient duration and amplitude. PNN has the
following characteristics: learning time is of 0.23(s),

557
C. Assessment vk (Z )  Z p (Z )i k (Z ) Fp (Z ) ª¬ v j (Z )  Z p (Z )i j (Z ) º¼ (2)
The identification and classification of power quality is a
vital task for the precise monitoring of power systems. In this
research, a comprehensive comparison of various techniques v j (Z )  Z p (Z )i j (Z ) Fp (Z ) ª¬ v k (Z )  Z p (Z )i k (Z ) º¼ (3)
based on the use of wavelet technique and artificial
intelligences such as Probabilistic Neural Network, Fuzzy
Logic and Neuro-Fuzzy method is made. It is shown that the In (2) and (3):
three methods of classification have their own inherent
disadvantages and advantages. Moreover, the three proposed Z : angular frequency independent variable.
methods of classification of disturbances demonstrate the v k (Z ) , i k (Z ) : vectors of phase-variable voltages and
potential for the development of fully automated monitoring
currents in the frequency-domain at terminal k of the line or
systems with on-line classification capabilities.
cable, respectively. The current flow reference direction is that
into the line or cable at terminal k.
III. MODELLING DISTRIBUTED-PARAMETER TRANSMISSION
LINES v j (Z ), i j (Z ) : vectors of phase-variable voltages and
The present research makes a contribution to addressing a currents in the frequency-domain at terminal j of the line or
stability issue in the phase-domain distributed-parameter cable respectively. The current flow reference direction is that
transmission line or underground cable model for EM transient into the line or cable at terminal j.
analysis. The paper develops a method to represent the
characteristic admittance (or impedance) matrix of a Z p (Z ) : the phase-variable characteristic impedance
distributed-parameter transmission circuit in terms of a passive matrix in the frequency-domain of the line or cable.
network comprising the interconnected resistors, inductors and
capacitors. In this way, the synthesized network is guaranteed Fp (Z ) : the phase-variable forward impulse response
to be stable always. The parameters of the network, i.e. matrix in the frequency-domain of the line or cable.
resistances, inductances and capacitances are determined by
constrained optimization in which the differences between the Subscript p: identifier for the pth transmission line or
synthesized network admittances and the calculated underground cable.
characteristic admittance matrix elements are minimized,
subject to the constraints that the network parameters are to be The expressions for Z p (Z ) and Fp (Z ) are given in [6]
positive. The stability is achieved without any need for
approximation as the real part of the admittance matrix of the The scattering equations in (2) and (3) form the basis for
synthesized network which comprises only passive circuit deriving the discrete-time-domain model used in EM transient
elements is always positive definite. The stability criterion in analysis. The key requirements are to synthesize s -plane or
relation to the synthesis of characteristic admittance matrix z -plane functions to represent the matrix functions
function [6] is, therefore, automatically fulfilled. The Z p (Z ) and Fp (Z ) . Once the s -plane or z -plane functions
synthesized network representing the characteristic admittance
have been synthesized, the frequency-domain equations in (2)
matrix (or impedance) is incorporated in the transmission line
and (3) are transformed in a straightforward manner to those in
or underground cable model expressed in the final form in the
the discrete-time domain.
discrete-time domain in a straightforward manner, through the
application of the trapezoidal rule of integration to individual
inductors and capacitors in the network. B. Transmission line energisation
The modelling and formulation developed and presented in
Representative results are presented in the paper which [6] have been implemented in software based on which
includes the comparisons between the synthesized functions switching transients study related to the energisation of the
and those obtained from transmission line parameter 220 kV Nhabe - Phumy transmission line in the system of Fig.
calculations, and EM transients arising from transmission line 9 is carried out. This configuration and basic conductor data
energisations. are those in [6]. The fault level of the source in Fig. 5 is
10,000 MVA. The time step length of 50 P s is adopted for the
A. Scattering equations in the frequency-domain
switching transients study.
Expressed in the frequency-domain, the scattering
equations in the phase-variable form for a transmission line or The preparatory phase prior to the transients study involves
underground cable connected between nodes k and j are the syntheses of z -plane rational functions for representing
the phase-domain forward impulse response matrix, and the
given in [6]: passive network which represents the phase-variable
characteristic impedance matrix of the transmission line. The

558
synthesized functions and network admittance when expressed 2.5
in the frequency-domain compare closely with the actual 2 proposed model
values as evaluated from the transmission line parameters. 1.5
frequency transform

Voltage [p.u]
0.5

-0.5

-1

-1.5
Fig. 5. A power system diagram for switching transients study.
-2

-2.5
0 0.002 0.004 0.006 0.008 0.01 0.012 0.014 0.016 0.018 0.02
Time [s]
C. Assessment
Figure 8: Phase “a ” switching transients in simultaneous-pole-closing.
The research has addressed a key issue related to stability in
modeling a distributed-parameter transmission circuit in the
phase-domain for EM transient analysis in power systems. The IV. POWER SYSTEM SIMULATION
passivity requirement in representing the phase-domain
This research will focus on the FWCM for linear ODE’s.
characteristic impedance matrix for time-domain EM transient
The extension of the FWCM to nonlinear systems is discussed
simulation is fulfilled in the most straightforward manner
in paper [5], due to limitations in space here. Therefore, we
through the use of passive RLC circuits interconnected
shall emphasize methodology and potential application of
(Fig. 6), and having their parameters directly identified to
electromagnetic in power system.
match the nonlinear frequency dependence of the matrix
function. The interconnected RLC circuits which are
synthesized are an inherently stable network. As indicated in A. Methodology and Algorithm
the paper, very high accuracy in addition to the stability has The idea behind the wavelet collocation method is quite
been achieved in the synthesis. simple [5]. The unknown function is expanded by a wavelet
series and let the expansion satisfy the original ODE at the
chosen collocation points, to determine the expansion
coefficients. Unlike the wavelets used in the image processing,
we need to construct the wavelets which are not only able to
represent a signal in the interior of an interval, but also the
arbitrary boundary conditions (including the signal’s derivative)

Before applying FWCM to any practical problem, the

following basic problems need to be addressed; the
introduction of approximation subspaces, the construction of
the basic functions, the treatment of initial conditions, the
expansion of the known functions, and the transform between
the expansion coefficients and functions at the collocation
points. Fig. 9 shows a proposed algorithm to apply dynamic
Fig. 6. Block diagram of an RLC network representing the characteristic
impedance matrix.
time-domain simulation in power systems.
B. Result and Discussion
-5
x 10
-0.5
actual data Result shows the magnitude of expansion coefficient of
-1
synthesized data different level wavelets. The coefficients of the third-level
wavelets were very small. In practical computation, we can
check the magnitude of the coefficient against a given param-
Susceptance [S]

-1.5

eter for the accuracy measurement and need not proceed to a

-2 higher level once the coefficients are small. Notice also, that
even at the same level, we do not have to insert wavelets over
-2.5
the entire interval. We need only add more wavelets at the
location where accuracy has not been attained. Due to the
-3
1000 2000 3000 4000 5000 6000 7000 8000 9000 10000
wavelet localization property and [5], we can actually have an
Frequency [Hz]
Fig. 7. Comparison between the susceptance of the synthesized circuit adaptive implementation of the FWCM by inserting wavelets
between phase nodes a and b (as referred to in Fig. 10) and the actual level by level and location by location. Such an adaptive
susceptance. capability is crucial to the practical circuit simulation and the

559
realization of the uniform approximation. It is seem that Transient based-wavelet techniques are fast. Results from
accuracy in the time range concerned has been increased. IEEE power systems (fig. 10) show that time consuming of
wavelet technique is short than that of other method. Moreover
Start level of j changes easily to adapt specified cases.
Wavelet technique improves accuracy and time consuming
in power system transient. The simulation results are
considered as the same as those of MATLAB program or ATP-
EMTP program. It proved there is an accurate and a practical
Set up derivative equation of wavelet-based methodology.
power system VHancock 132
VFieldale 132
1.4 V
Roanoke 132
V
Blaine 132
VReusens 132
1.2 VRoanoke 1.0
VRoanoke 33
VRoanoke 11

No 1 V
V
Hancock 33

Hancock 11
V
Bus 14 33
0.8 VBus 15 33
VBus 16 33
VBus 17 33
0.6 VBus 18 33

Require level of V
V
Bus 19

Bus 20
33

33

correct? 0.4 VBus 21

VBus 22
33

33
VBus 23 33
0.2
VBus 24 33
VBus 25 33
V
Bus 26 33
0
V
Yes 0 2 4 6 8 10
time (s)
12 14 16 18 Cloverdle 33 20
VCloverdle132
VBus 29 33

Fig. 10: Results of short - circuit voltage of IEEE system 30 buses.

Solve by Waveles methodology The research is the new idea to use the wavelet technique
for power system transient. The experimental data are
presented simply for demonstrating the feasibility and
properties of the wavelet-based method, not for showing its
efficiency in practical power system transient simulations.
Many implementation details and more benchmark testing are
Solution for transient of time
domain in power system
needed before the method becomes mature. The wavelet-based
methodology has demonstrated a promising direction in the
development of high-speed transient simulation of power
system. We are investigating the use of this method to develop
some practical transient simulations in power systems.
Yes V. APPLICATION IN SOUTHERN POWER SYSTEM
Does a power OF VIETNAM
system Three proposed topics above apply at 500kV NHABE
configuration substation in the Southern power system of Vietnam with
change?
assessments as following:
- Recognition of power system transient phenomena.
No - Switching power transmission lines (in section III).
There are results as following:

Output final exam

results

End

Figure 9: A proposed based-wavelets-technique algorithm.

Fig. 11: Schema location of NHABE substation of 220-550kV Southern

power system of Vietnam.

560
 VI. CONCLUSION
The paper presents wide range of power system transients
such as identification, analysis, modeling power transmission
line… Some contributions of the research of power system
transients can be listed:
- Detection and time power system transient phenomena
with noise.
- Recognition specified capacitor switchings.
- Identification and classification power system
disturbances with high accuracy.
- A proposed based wavelet analysis in discrete- time-
domain simulation.
Figure 12: Short-circuit voltages at 220kV bus at NHABE substation.
- The passivity requirement in representing the phase-
domain characteristic impedance matrix for time-domain EM
transient simulation is fulfilled in the most straightforward
manner through the use of passive RLC circuits interconnected,
and having their parameters directly identified to match the
nonlinear frequency dependence of the matrix function. The
interconnected RLC circuits which are synthesized are an
inherently stable network. As indicated in the paper [6], very
Figure 13: Identification short-circuit of 220kV Nhabe - PhuMy power
high accuracy in addition to the stability has been achieved in
transmission line. the synthesis.
These contributions above have been applied at Nhabe
Substation in southern power system of Vietnam with good
simulation analysis. We are investigating the use of this method
to develop some practical transient simulations of power
systems.

REFERENCES
[1] Allan Greenwood, “Electrical Transient in Power System”, Jonh Willey
and Sons, Inc. 1991
[2] Neville Watson, Jos Arrillaga, “Power Systems Electromagnetic
Figure 14: Identification harmonic fault at 220kV NHABE bus. Transients Simulation”, IET, London, United Kingdom, 2007
[3] Nguyen Huu Phuc, Truong Quoc Khanh, Nguyen Nhan Bon, “Discrete
Mẫu
Wavelet Transform Technique Application in Identification of Power
Quality Disturbances”, pp159-164, ISEE2005 in October 2005,
Phân tích DWT 13 mức Hochiminh City University of Technology.
w1 , w2 , …, w13
[4] Bon N. N, Phuc H. N., Anh Q. H., “Proposal Wavelet-Technique-Based
Algorithm to Evaluate Electromagnetic Transients in Power Systems, ”
Khoảng thời gian
quá độ
13 mức năng lượng
DWT hệ số D the 10th Postgraduate Electrical Engineering and Computing Symposium
tt ,A P1D , P2D , …, P13D in Australia, Vol. 10 pp 230-236 Octorber 2009.
Lớp ẩn
[5] Bon N. N, Phuc H. N., Anh Q. H., Anh V. T., “Assessment of Power
H1 Quality Disturbances Using Various Wavelet-Based Artificial Intelligent
Lớp ngõ vào
X1 H2
Lớp ngõ ra
Y1
Techniques, ” the 10th Postgraduate Electrical Engineering and
H3
Computing Symposium in Australia, Vol. 10 pp 40-46 Octorber 2009
X2 Y2

H4 [6] T. T. Nguyen, V. L. Nguyen, Bon N. Nguyen “Phase-Domain

Xi Yj Transmission Circuit Model for Electromagnetic Transient Analysis-
Representation of Characteristic Impedance Matrix by a passive
Hh
Network”, IEEE General Meeting in June 2010 in USA.
1 2 3 4 5 6 7 8 [7] Phuc H. N., Bon N. N, Anh Q. H., “Assessment and Analysis of
NM NM NM NM Đóng Họa Mất Sin
NB-NSG NB-Q7 NB-NT NB-PM cắt tải Tần Điện chuẩn
Electromagnetic Transients in Power System ” the 2011 International
Symposium on Electrical & Electronic Engineering, Ho chi Minh city,
Vietnam, November 8-9, 2011 (ISEE2011+ASAC2011)
Figure 15: Neural network model for transient phenomena recognition at
NHABE Substation.

561