Energy and Power Engineering, 2013, 5, 922-926
doi:10.4236/epe.2013.54B176 Published Online July 2013 (http://www.scirp.org/journal/epe)
Study on Voltage Sag Detection of Wind Power System
Based on HHT
Yanqing Li, Ting Yue, Hongling Xie, Feilong Wang
Hebei Provincial Key Laboratory of Power Transmission Equipment Security Defense,
North China Electric Power University, Baoding, China
Received December, 2012
ABSTRACT
For the output of wind power system has the characteristics of randomness, volatility and intermittence, the voltage of
wind power system fluctuates frequently and voltage sag is one of the most common voltage fluctuations in wind power
system. For the problem of voltage sag of wind power system, the limitations of the detection methods such as the
square detection method, the half-wave RMS detection method and wavelet transform are summed up, and a new de-
tecting method named Hilbert-huang Transform(HHT) is put forward in this paper, which can detect the voltage sag
accurately and timely. In order to solve the problem of end effect in the process of empirical mode decompostion
(EMD), a self-adaptive method named improved waveform matching is applied in dealing with the end issue. Voltage
fluctuations are reflected by two parameters named voltage amplitude and frequency of each intrinsic mode function
(IMF) in HHT. The practicality of the method is verified by Matlab simulation.
Keywords: HHT; Wind Power System; Voltage Sag; Detection
1. Introduction
With the development of the society, we are facing a
problem of a gradually redu ce of pr imary energ y reserves,
it is essential for us to develop other forms of energy. In
such a background, renewable and clean wind gradually
cause concern at home and abroad in the related field,
large-scale development and utilization of wind power is
considered to be an effective measure to solve the energy
crisis and environment pollution problem. Our country
has massive land, abundant wind power, so if we make
full use of this energy, not only can we create consider-
able economic benefits, but we can also effectively re-
lieve the pressure caused by the shortage of traditional
energy, and we also can provide long-term effective en-
ergy supply[1].
The development of new energy resources in China,
such as wind power, conforms to the strategic require-
ments of sustainable development. However, wind power
is vulnerable to natural climate influence. The output of
wind power system has the characteristics of randomness,
volatility and intermittence. Wind power system voltage
fluctuations may happen at any time, which is difficult to
control. Voltage sag is one of the most common forms of
voltage fluctuation. It has serious impact on the power
quality of wind power system, causes great harm to the
sensitive loads, brings great difficulties to wind power
integration, and seriously restricts the large-scale devel-
opment of wi n d po wer.
At the present, there are several common methods for
voltage sag detection in wind power system, such as the
square detection method, the half-wave RMS detection
method and wavelet transform detection method. To
some extent, these methods can detect wind power sys-
tem’s voltage sag, but there are also certain deficiencies
in these methods. The square detection method ignores
the frequency shift component at the moment of volt-
age’s sag[2]. The half-wave RMS detection method
needs to take sampling d ata that is half a cycle in orde r to
get the conclusion, which can’t guarantee the real-time.
So this method can only be used for occasions that real
time requirement is not high[3]. Wavelet transform de-
tection method is appropriate for wave signal which con-
tains one, two or above two frequency. But it requires
synchronous signal and carrier signal the same phase,
same frequency and to have strict frequency division. It
also demands energy concentrated wavelet to improve
the detection’s precision[4]. By a two-stage decomposi-
tion of voltage sag detection, Daubechies wavelet is able
to detect the voltage sag’s start-stop moment and drop
amplitude accurately by making a two-stage decomposi-
tion for voltage sag. But the choice of wavelet base is a
big problem. At present, how to select the wavelet base is
not unified or have a pri nci ple[5].

Y. Q. LI ET AL. 923
We need to detect real-time wind power system volt-
age sag’s happen time and amplitude effectively, so that
we can take certain compensation measures and improve
the wind power system’s power supply reliability and
power quality. This paper presents a improved HHT me-
thod to detect voltage sag of the wind power system. This
method is based on the waveform matching method for
end processing. And the simulation results verify the
practicability.
2. Method Introduction
2.1. HHT Introduction
Hilbert-Huang transformation(HHT) is a new developed
method for signal analysis. This method consists the em-
pirical mode decomposition(EMD) and Hilbert transfor-
mation[6]. Through the EMD decomposition, signal can
be decomposed into a series of intrinsic mode function
(IMF). The intrinsic mode function is a signal which is
approximate to single-frequency components, which
means at all times, there is only one signal frequency
component. For each intrinsic mode function on Hilbert
transform, we can get the instantaneous spectrum of each
IMF.
1) EMD process
According to the maximum point and the minimum
point of the signal (),
x
t
1()
we can find out the average of
the upper envelope
x
t an d the lower envelope 2()
x
t.
112
1[() ()]
2
x
txt   (1)
Then calculate the difference between ()
x
t and 1

as 1

:
1
()xt 1

  (2)
If 1

satisfies the two conditions of IMF（ The
number of the points which past the extreme point and
zero point is the same or differ at most one. The signal
is symmetric about time axis）, 1

is the first IMF
component of ()
x
t[7]. If 1

doesn’t satisfy the two
conditions of IMF, Put 1

as raw data. Repeat the
above process k times, get 11(1)1k ,
kk





then judge
whether each screening results are IMF components us-
ing
D
S.
2
1( 1)1
2
01( 1)
() ()
()
nkk
Dtk
tt
St





 (3)
in Formula (3),
D
S can be determined according to the
actual requirements. If 1k

satisfies the requirement of
D
S, then make 1k1,

  and 1

is the first IMF compo-
nent of signal ()
x
t. Separate 1

from ()
x
t as Formula
(4):
1
()rxt
Take 1 as the new r()
x
t
, and repeat the above process, we can get 234 ,,

 … Stop until n r is monotonic or
n
r is very small. The result of the decomposition is as
follows:
1
() ()
n
i
in
x
tt 

r


 (5)
2) Hilbert transformation
Do Hilbert transformation on ()
it

which is a IMF
component as follows:
() 1
() i
i
H
t
t

d





 (6)
Its inverse transform is as follows:
() 1
() i
i
H
t
t

d






 (7)
We can get analytic signal ()
A
t:
()
()()()() i
j
t
ii ii
AttjHt ate

   (8)
The instant amplitude:
22
()() ()
iii
attH t  ,
The phase:
()
( )arctan.
()
i
ii
H
t
tt




The instant frequency:
() 1
() 2i
i
dt
ft dt


 (9)
2.2. Introduction of End Effect
In the decomposition process of empirical mode, we need
to calculate the local average of signal according to the
envelope in the calculating process of the IMF. What's
more, the upper and lower envelope can respectively be
obtained by making cubic spline interpolation algorithm
on signal’s local maximum and local minimum[8]. Due
to the signal’s two endpoints are not necessarily the ex-
treme point which couldn’t satisfy the requirements of
interpolation, so it may bring some error, this situation is
so-called end effect.
2.3. Introduction of Self-adaptive Method
Named Improved Waveform Matching
In the aspect of end effect’s inhibition, common methods
include mirror continuation method, continuation method
based on neural network, continuation method based on
polynomial fitting, etc. These methods can inhibit end
effect to a certain extent. But they also have some prob-
lems[9]. In order to detect voltage sags of wind power
system quickly, a self-adaptive method named improved
1

 (4)

Y. Q. LI ET AL.
924
waveform matching is applied in dealing with the end
issue. The core idea of the waveform matching method is
as follows: According to the law of nature signal, we
assume that the development and change of the signal is
always according to certain rules. Signal’s development
trend at boundaries will also be reflected in inner signal,
especially for the regularity strong signal, this feature
will be more obvious. In order to test the true extent for
continuation wavefo rm, we need to introduce the con cept
of waveform matching degree to test the authenticity of
the continuation waveform. Assume 12 are two
data sequences whose length both are N. 11
22 are two points on . Then the
waveform matching degree of 12 which rela-
tive to 1 can be obtained according to the follow-
ing steps.
() ()ft ft,
St
12
() ()ft ft,
() ()ft,
(, ()),ft
(, ())St ft
, Sft
2
S
1) Translating coincides the and , the
new waveform is 1()ft
'
1() 1
S2 S
f
t;
2) Obtain the waveform matching degree of 1
2 which relative to according to formula
(10).
()ft,
()ft 12 , SS
'
122 1
1
( ()()S)(()())
N
j
mdft ftfjfj



，， 2 (10)
Concrete steps of waveform matching method are as
follows:
1) Obtain all extreme points i
M
of original signal
, put the maximum points into set{} and the
minimum points into set {};
()ft,maxi M
,mini
2) The first minimum point is 0
M
M
and the first
maximum point is i
M
. The distance between 1
M
and
at the start time is ()ft0
s
d. The length of 0
s
d
M is l; 3) The waveform matching degree of all respect
to ,maxi
0
s
d is ;
i
4) The minimum wave band of i md is i
md
s
d.If
i
md l

 (

is a constant which is determined by the
matching accuracy requirements), take i as wave-
form continuation of ’s left end. Otherwise, process
according to the step (5);
md
()ft
5) Specify the maximum and the minimum of the
endpoint directly. Take the average value of two adjacent
maximum points at original signal’s left-most derivation
as the maximum at left end. Take the average value of
two adjacent minimum points at original signal’s left-
most derivation as the minimum at left end.
3. Example Simulation
In fact, due to natural environment’s influence on the
voltage of wind power system, fluctuations are more
complex. This paper does some simulation analysis to
voltage sags of wind power system voltage fluctuation.
During the non-sag, we assume the wind power system
voltage waveform always remain unchanged(amplitude
is rated voltage and its frequency is 50 Hz). Voltage sag
usually refers to root-mean-square voltage quickly drop
to 90% - 10% of the rated voltage, and then quickly re-
store to the normal voltage. Its typical duration is 0.5-30
period [10]. Accordingly, we assume that a wind power
system voltage RMS temporarily reduces to 70% of the
rated voltage, and the duration is five period. The voltage
function is shown in Equation (11), and the voltage sag
waveform is shown in Figure 1.
cos(100 ),
() 0.7cos(100 ),0.10.2
tothers
ut tt





 (11)
For the problem of voltage sag, this paper firstly apply
waveform matching data continuation subroutine within
Matlab to the extension of original signal, which can
effectively avoid the end effect in the process of HHT
transform. Then call the procedure about Hilbert-huang
Transform to do HHT treatment for the continuation sig-
nal, and output the amplitude / time, frequency / time
curve of IMF1. Finally, according to the amplitude/time,
frequency/time curve of IMF1, we got the voltage am-
plitude and voltage frequency of wind power system when
voltage sag happened, which could help to judge the time
and the amplitude about voltage sag. HHT simulation
results are shown in Figure 2 and Figure 3. And the si-
mulation is after the endpoint processing based on the
waveform matching adaptive.
In the simulation, we assume a voltage sag happen af-
ter 0.1 - 0.2 s. We can conclude that the voltage is only
Figure 1. Voltage sag waveform.
00.05 0.10.15 0.20.25 0.3 0.350.4 0.450.5
-2
0
2
00.05 0.10.15 0.20.25 0.3 0.350.4 0.450.5
-0.1
0
0.1
00.05 0.10.15 0.20.25 0.3 0.350.4 0.450.5
-0.05
0
0.05
00.05 0.10.15 0.20.25 0.3 0.350.4 0.450.5
-0.2
0
0.2
Time
Figure 2. EMD decomposition results.

Y. Q. LI ET AL. 925
0.62 times the rated voltage at the moment of 0.1s or 0.2s
from the HHT simulation results; At the moment near 0.1
s or 0.2 s, the instantaneous frequency also has great
changes. The highest frequency is up to 78.64 Hz, the
minimum frequency is only 19.8 Hz. We can judge the
situation of voltage sags in wind power system effec-
tively and in real time according to the mutations of the
voltage parameter.
If the wind power system voltage RMS temporary de-
cline to 80% of the rated voltage, and the duration is 0.5
a period, then voltage function is shown in formula12,
voltage waveform is shown in Figure 4.
cos(100 ),
() 0.8cos(100 ),0.20.21
tothers
ut tt





 (12)
Voltage sag parameters of a wind power system are as
follows: the root-mean-square voltage temporarily reduce
to 80% of the rated voltage, the duration is 0.5 periods.
HHT simulation results after self-adaptive endpoint
treatment based on waveform matching is shown in Fig-
ure 5, Figure 6.
0 0.1 0.2 0.3 0.4
0
50
100
time
frequency
0
0.1 0.2 0.3 0.4 0.5
0.8
1
1.2
1.4
time
amplitude
Figure 3. IMF1’s amplitude, frequency variation overtime.
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5
-2
-1.5
-1
-0.5
0
0.5
1
1.5
2
time
amplitude
Figure 4. Voltage mutation waveform.
00.05 0.10.15 0.20.25 0.3 0.350.4 0.450.5
-2
0
2
00.05 0.10.15 0.20.25 0.3 0.350.4 0.450.5
-0.05
0
0.05
00.05 0.10.15 0.20.25 0.3 0.350.4 0.450.5
-0.05
0
0.05
00.05 0.10.15 0.20.25 0.3 0.350.4 0.450.5
-0.01
0
0.01
Time
Figure 5. EMD decomposition results.
00.1 0.2 0.3 0.4
0
50
100
time
frequency
00.1 0.2 0.3 0.4 0.5
0.7
0.8
0.9
1
1.1
time
amplitude
Figure 6. IMF1 amplitude, frequency variation over time.
The HHT simulation results show that in the h alf cycle
start at 0.2 s, the wind power system voltage changes. At
the time of 0.2 s, the minimum voltage of wind power
system is 0.77 times the rated voltage. At the time of 0.2
s, instantaneous frequency also have great changes, The
highest frequency is 68.4 Hz, the minimum frequency is
only 32.2 Hz. We can judge the situation of voltage sags
in wind power system effectively and in real time ac-
cording to the mutation s of the voltage parameter.
4. Conclusions
This paper applies Hilbert-huang transform method (HHT)
to the real-time and accurate detection of the voltage sag
for wind power system. In order to solve the endpoint
effect when do the EMD decomposition, we use a wave-
form matching self-adaptive data continuation technol-
ogy to manage endpoint waveform. Matlab example si-
mulation results show that the HHT can detect the situa-
tion of voltage sags in wind power system effectively and

Y. Q. LI ET AL.
926
in real time, can accurately judge the moment of voltage
sag and the voltage amplitude after sag and voltage sag’s
duration. Export voltage of wind power system is influ-
enced by the natural environment, so the export voltage
is fluctuating. Practical problems are more complex than
the research of voltage sag in this paper and it is more
difficult to detect voltage fluctuation. The detection me-
thod used in this paper is especially for voltage sag of
wind power system. In order to strengthen the control of
the wind power system, we also need to do further re-
search on other volt age fl uct u at i on pr o bl ems.
REFERENCES
[1] Y. Zhang and G. D. Gao, “Wind power generation,” Bei-
jing, China Power Press, 2002.
[2] Y. Y. Hu, S. S. Niu, Z. R. Liang, etc. “Of Voltage Flicker
Measurement Algorithm,” Journal of Low Voltage Elec-
trical Appliances, 2012, pp. 32-36.
[3] S. J. Hu, J. L. Li and M. L, “Test Method of Wind Power
System’s Implementation LVRT Grid Voltage Drop.
High Power Communication Technology, Vol. 2, 2008,
pp. 17-21.
[4] X. P. Wei, Y. H. Xu and X. G. Xiao, “Voltage Fluctua-
tion and Flicker Detection Methods in Electric Measure-
ment and Instrument., Vol. 46, 2009, pp. 1-6.
[5] Y. C. Liu, “Analysis and Study of Voltage Sag’s Detec-
tion Method,” Changsha: hunan university, 2007.
[6] T. Y. Li, Y. Zhao, N. Li N, etc. “New Method for the
Detection of Power Quality Based on the HHT,” Chinese
Journal of Electrical Engineering, Vol. 17, 2005, pp.
32-56.
[7] Y. M. Zhong, S. R. Qin and B. P. Shang, “Research of
Hilbert - Huang Transform,” Journal of Vibration and
Impact, Vol. 21. 2002, pp. 13-17.
[8] D. J. Yu, J. S. Cheng, Y. Yang and H. Hilbert, “Trans-
form Method for Mechanical Fault Diagnosis,” Beijing:
Science Press, 2006.
[9] C. X. Shao, J. Wang, J. F. Fan, etc., “An adaptive EMD
endpoint continuation method,” Journal of Electronics,
Vol. 35, 2007, pp. 1944-1947.