EJEEE
https://doi.org/10.62909/ejeee.2024.006 Edison Journal for Electrical and Electronics Engineering
Article
Mitigation The Impact of Voltage-Dip on The DFIG Using DVR
Based on ANFIS Tool
Batoul Amen Alsati1, *, , and Ghassan Issa Ibrahim1,
1 Department of Electrical Engineering, Faculty of Engineering Homs University, Syria; batolalsati@gmail.com
* Correspondence: batolalsati@gmail.com; Gassan12@gmail.com
Abstract: The Doubly Fed Induction Generator (DFIG) performance affected by several states such
as the stator voltage variation which has been concentrated by many researchers. When a voltage-
dip happens, both stator and rotor currents increase to a high level. Additionally, the DC-link volt-
age is also increased to a high value. The researchers proposed multiple approaches to mitigate the
serious effects that associated with the voltage-dip, but each method addresses one problem for a
specific fault condition without taking into account addressing all parameters for various condi-
tions, which shows the limitations of the methods used. Moreover, many schemes also neglect the
shifting from generator to power consumer in some severe fault conditions. In this paper, Dynamic
Voltage Restorer (DVR) integrated with adaptive-neuro-fuzzy-inference system (ANFIS) is imple-
mented to enhance the DFIG operation during 50% symmetric-dip of the rated value, the torque
ripples are reduced to 10% in this case compared with the crowbar (142.34%). The second tested
case was 90% symmetric-dip, the torque ripples were 190.80%, 11.39% for Crowbar and DVR+AN-
FIS respectively. For the third case asymmetric-dip, the torque ripples were 140.42%, 10.25% respec-
tively. Other parameters such as THD%, are also compared. The results are shown that crowbar
circuit connection to the DFIG did not mitigate the over-currents. Additionally, the rotor-side con-
verter (RSC) control is lost. The rotor-currents are reduced, and the DC-link voltage () is in-
creased from 1150 V to 1200 V. However, the proposed DVR based on (ANFIS) controllers provides
the best performance by predicting the voltage-dip occurrence, the rotor-currents are reduced to the
normal values comparing with Crowbar method, the ) is also reached to the nominal value
(1150 V).
Keywords: ANFIS, Crowbar, DFIG, Wind.
1. Introduction
Recently, the environmental problems have led the world to discover new sources
that are considered environmentally friendly energy sources. Wind-energy is one of the
most desired and powerful solutions. Generating power based on wind is gradually
growing everywhere. Wind system needs a generator to convert the mechanical power to
an electrical one. The doubly-Fed induction generator (DFIG) is the first choice for power
systems due to its small size, its converters with small power-ratings also provide. High
ratings of power (MVA). It is also able to operate with both variable wind-speed as well
as fixed frequency [1]. However, DFIG has a sensitivity to the sudden disturbances that
happen in the grid. This is due to the connection of the stator directly to the grid; thus, a
high current is created due to the voltage-dip. Due to the electromagnetic-coupling be-
tween the rotor and stator windings, over-currents are also created in the DFIG rotor. This
in turn impacts passively on the rotor-side converter (RSC). Figure 1 explains the structure
of the wind system using DFIG. Voltage-dip also causes over-voltage in the DC-Link [2].
Citation: Alsati, B. A., and Ibrahim,
G. I., Mitigation the Impact of Volt-
age-Dip on The DFIG Using DVR
Based on ANFIS Tool. 2025. 3. p. 46-
64
Academic Editor: Prof. Dr. Sergey
Shabunin
Received: 4/9/2025
Revised: 16/10/2025
Accepted: 27/10/2025
Published: 2/11/2025
Copyright: © 2025 by the authors.
Submitted for possible open access
publication under the terms and
conditions of the Creative Commons
Attribution (CC BY) license
(https://creativecommons.org/license
s/by/4.0/).
EJEEE 2025, Vol.3 47 of 64
The novel grid-codes recommended remaining the wind-turbines connected to the grid
in the presence of voltage-dips by improving their performance under transient states [3].
Voltage-dip makes the wind turbines (WT) system interrupted frequently. For the large
WT systems, the interruption leads to preventing the grid from multi-megawatts of power
[4]. Previously, the level of WTs penetration was small compared with classical generation
units [5], thus the WTs disconnections from the grid are allowed under voltage-dips to
prevent the negative effects. Nowadays, the WTs penetration is obviously increased in the
power system. Several researchers provide fault ride-through strategies. Some of these
strategies improve the performance partially, and the others improve the performance
completely. Several researchers tried to combine these strategies to provide powerful per-
formance. A traditional crowbar protection is applied to protect the RSC against asym-
metric and symmetric voltage, the Vdc of the DC-Link is improved at the asymmetric
type, but there is no improvement to the rest of the parameters for other types of faults
[6]. Under voltage-dip, this crowbar provides a path to allow the over-currents to pass
through. In this case, the crowbar disconnects the RSC to protect it from the over-currents
and this is considered the major disadvantage of this scheme [7]. When RSC is discon-
nected, the DFIG changes to a squirrel cage induction generator and receives reactive
power. Additionally, Crowbar protection requires a suitable instant to be deactivated.
This is necessary to decrease the consumption of reactive power and recover the DFIG
control as mentioned in [8], the results explained that only rotor-current and dc voltage
are improved under symmetric faults, the researcher did not treat the other types of faults.
Additionally, the torque ripples are not taken into consideration. Ref [9] studied DFIG
performance with the crowbar circuit under only a symmetric dip. The selected protection
did not improve the rotor-current and other DFIG parameters. Ref [10] used STATCOM,
that is installed on the grid-side converter, sharing with a super-capacitor which is in-
stalled on the DC-Link to enhance the DFIG performance. The outcomes illustrated that
STATCOM reduced only the peak of Vdc but did not mitigate the ripples, the method is
tested only for asymmetric dip. Ref [11] Compared between SVC, STATCOM, SDSC tech-
niques to mitigate the fault effects, these methods are tested for only symmetric-dips. The
results showed that SVC and STATCOM reduced the overvalues of the Vdc and rotor-
currents, but they have not reduced the oscillations. The SDSC improved the DFIG's pa-
rameters and mitigated the ripples. Ref [12] Applied DVR circuit sharing with synchro-
nous-reference frame control approach to compensate for the voltage-dip on the DFIG
stator, the method is only applied for the asymmetric-dip. This method proved its ability
to reduce the overcurrent and mitigate the Tem's ripples. Ref [13] Investigated the DVR
performance to compensate the symmetric-dip based on several control schemes; the
studied schemes did not mitigate the ripples. Ref [14] Proposed DVR based on active-
disturbance rejection control to enhance the LVRT needs; only the symmetric case is stud-
ied and also in this case, the torque's oscillations are not cancelled.
Problem of study:
From previous studies, it is found that each of them used a method to treat one type
of fault without the rest of the types such as [8, 10, 13], and in each of these studies some
parameters were improved. For example, ref [13] improved the current but did not men-
tion the torque, and some of them reduced the capacitor voltage but did not eliminate the
oscillations, and the rest of the studies are the same. Additionally, those methods also
neglect the shifting from generator to power consumer in some severe fault conditions.
Therefore, since the electrical grid is vulnerable to multiple types of faults, an effec-
tive method must be employed to ensure optimal generator performance. This minimizes
fault currents, protects the rotor windings and the RSC, and maintains the nominal values
of both the capacitor voltage and electromagnetic torque. Furthermore, many studies have
not addressed the problem of torque ripples, nor have they addressed the problem of the
machine's shifting from power delivery to power consumption from the grid during
faults. Based on the above, this study designed a DVR that integrates with ANFIS control-
lers, which are trained on a wide range of grid data. This enables it to adapt to all possible
EJEEE 2025, Vol.3 48 of 64
fault types, while also offering flexibility and the ability to optimize generator parameters
and adjust its ability to remain in power delivery mode during faults.
Aims of Study:
This study aims basically to mitigate the voltage-dip effects by designing a DVR in-
tegrated with ANFIS controllers to predict the voltage-dip occurrence and mitigate its ef-
fects, and remove the limitations of the other methods that mentioned in previous studies.
The proposed scheme is compared with Crowbar for the following cases:
1-Symmetric voltage dip (50%) of the nominal value.
2- Symmetric voltage dip (90%) of the nominal value.
3-Asymmetric voltage dip
The proposed controller based on ANFIS must be able
to adapt to various types of grid failures in short time, ensuring that the DFIG does not
malfunction or degrade its performance, while also improving signal quality and reduc-
ing distortion and ripples.
DFIG is represented according to Park's transformation to convert three phase mag-
nitudes to () coordinates as following [14]:
DFIG's voltages according to the d-axis
(1)
(2)
DFIG's voltages according to q-axis
(3)
(4)
: are the stator (voltage, current and flux) respectively according to the
direct axis.
: are the stator (voltage, current and flux) respectively according to the
quadrature axis.
: the stator-resistance.
: are the rotor (voltage, current and flux) respectively according to direct-
axis [15].
: are the rotor (voltage, current and flux) respectively according to the
quadrature axis [15].
: rotor speed
: is the rotor-resistance.
& are the rotor and stator inductances respectively
: the DFIG's leakage-coefficient.
The electromagnetic-torque of DFIG is formulated as follows:
(5)
EJEEE 2025, Vol.3 49 of 64
Where P: Poles' DFIG number.
DFIG's Converters control schemes based on Space-Vector PWM:
Based on Fourier series, the three-phase voltage which is used to feed the 2-level RSC is
written as follows:
(6)
The value of specifies the modulation-index m that is used in PWM methods.
There are eight potential switches sequences corresponding to the voltage sectors that are
shown in figure 2 [16]:
{V1, V2, V3, V4, V5, V6}: the active-vectors
{V0, V7}: the zero-vectors.
Figure 1. DFIG structure
Figure. 2 Switching Sectors
2. Materials and Methods
Crowbar: This method consists of controllable resistors that are inserted between the
RSC and DFIG's rotor to limit the high rotor currents during the transient case. Crowbar
is activated when a voltage dip occurs and disconnects the RSC by inserting resistors to
the three-phase rotor windings in series. After recovering the voltage, it reconnects RSC
again to the rotor-winding [6]. In this study, the resistor value of crowbar resistor is (0.2)
ohm. Figure 3 shows the Crowbar representation in MATLAB.
DVR: This strategy is characterized by complexity and high cost. It has the same
structure as STATCOM-device [17]. Figure 4 explains the connection of the DVR to the
grid. DVR represents a signal amplifier that amplifies the voltage signal to enhance the
low voltage ride-through of DFIG and keep the DFIG voltage at the desired level. By this
way, DFIG will remain connected to the power system at both steady and fault states.
During the voltage compensation by DVR, the DFIG provides active power to the grid by
DVR [18].
EJEEE 2025, Vol.3 50 of 64
Figure. 3 Crowbar diagram
Figure 4. DVR connected to the grid
The DVR injects the desired voltage that is generated by the converter through
a coupling-transformer. The injected voltage is controlled to mitigate the effects of faults
and enhance the LVRT capability of DFIG. Thereby, any disturbance in the grid is com-
pensated by a suitable voltage. The DVR has two modes to operate as follows: standby
mode where the injection voltage equals zero and the semiconductors do not switch in
this mode and the boost mode where the injection voltage is larger than zero and the DVR
injects voltage to mitigate the voltage-dip effects. A closed-loop control is used to control
the injection of DVR, according to the reference-frame. The classical control scheme
for the DVR depends on PI controllers in general due to their simplicity [16]. The gains of
traditional PI controllers are constant and do not have the capability to control the DVR
injection process effectively to meet the variations of voltage dip[16].
ANFIS controllers: The traditional Fuzzy-logic controllers rely on the relation be-
tween the parameters and the membership-functions. Additionally, the rules of the con-
trollers depend on the expert opinion. By this way, the parameters changing is imple-
mented by an error and trial method with no optimization approach [19]. Therefore, mul-
tiple approaches have been proposed to overcome these problems such as Particle Swarm
[18], neural-networks, Genetic Algorithms [20]. ANFIS is an algorithm that employs the
fuzzy system to convert the input into a demanded output. This is implemented by inter-
connected ANN elements and data connections [6]. A SUGENO type is selected for the
ANFIS controller, which combines the properties of the traditional fuzzy principles and
the neural network. This strategy decreases the required time for determining the rules
and ranges. It also optimizes the outcomes. ANFIS consists of 5 layers that are intercon-
nected as follows:
1. Fuzzification: this layer is responsible to obtain the inputs' membership function.
2. Rule: this layer multiples the inputs and gives the outputs to the following layer.
3. Normalization.
4. Defuzzification.
5. Output.
To design this controller, some steps should be implemented in MATLAB. The re-
quired data for training (inputs and outputs) is obtained from the grid voltage and ex-
ported to the workspace in MATLAB. After that, data should be exported to the ANFIS
toolbox to start the training process. GUSSMF memberships and the hybrid methods are
used. When the ANFIS designing is completed, it is applied on the DVR controller, figures
(5- a, b), show the structure of the used ANFIS. Table 1 represents the parameters that are
used to design the ANFIS controllers.
EJEEE 2025, Vol.3 51 of 64
(a)
Figure 5. the Memberships of (a) the input1 (b) the input 2
Figure 6 presented the block diagram of the control loops based on ANFIS which is
applied to drive the DVR. ANFIS provides more accuracy to the DVR by predicting the
voltage-dip. When the DVR becomes able to predict the voltage-dip, a proper level of
voltage is provided to the grid by the DVR which enhances the DFIG performance.
Figure 6. Structure of the control scheme in DVR
Figure 8. Structure of ANFIS for two inputs and one
output
The controllers are trained to maintain the stator voltage at its nominal value, thus
improving the performance of the machine and ensuring that it is not affected by faults
caused by the grid, regardless of its type. The ANFIS training is implemented by collecting
the data (inputs & outputs) from the simulations that represent the DFIG behavior during
several types of voltage dips with traditional PI controllers. The inputs are collected from
the instantaneous values of the load voltage after converting it from a three-phase quan-
tity to a two-phase quantity according to the transformations (dq), and the outputs are
represented by reference signals that are used to implement the control loops for the DVR.
Table 2 shows the basic parameters of the DFIG to implement this study.
Table 1. Parameters of the designed ANFIS
Parameter
Value
Inputs
2
Outputs
1
Table 2. Parameters of DFIG
Parameters
Values
(PU)
EJEEE 2025, Vol.3 52 of 64
Rules
25
Epochs
3
MF of the input 1
5
MF of the input 2
5
MF of the output
14
Inference system type
SUGENO
Method
Hybrid
(Pu)
The surface curve of input 1(error voltage) and input 2(changes in the error), and the
output for the first controller in this study are shown in figure 7. These curves represent
the basic used data to train the controller. Figure (7) represents the relation between the
inputs of the second controller depending on ANFIS, and illustrates that the controller
linked the input data to the output depending on two factors, the Fuzzy rules as well as
the Memberships. The surfaces present the nonlinear features that have peaks and bot-
toms, this proved the ability of the ANFIS controller to adapt to complicated inputs and
capture the changes in the dataset. These curves implies that the inputs' variations impact
the output. The various gathering of the inputs is causing varying levels of smoothness or
unexpected transformations in output. The curves also show that there is no discontinuity,
this means that there are no overfitting problems, and that the training data and fuzzy
rules are appropriate.
Figure 7. 3-D representation of the inputs and outputs of the ANFIS controller
The controller is managed by 25 fuzzy rules that are represented by figure (9), the
design is implemented by a hybrid ANN structure that combines the fuzzy logic basics
and a neural network. Through the training process, the relation between the inputs and
output is modified, and their impact on the output. The figure explains how the output
changes due to the input changes. This indicates the controllers' ability to adapt their per-
formance to represent the nonlinear relation. The detailed fuzzy rules for the controller
are mentioned in the appendix.
EJEEE 2025, Vol.3 53 of 64
Figure 9. The rules of inputs and output of the ANFIS after training
For DFIG control, the SVPWM is applied to the rotor-side converter (RSC). The DFIG
performance is investigated with DVR based on ANFIS-controllers during three phase
voltage-dip. In fact, speed changes are not taken into account because the focus of the
controller design is on transient cases related to faults. In these cases, wind speed changes
do not affect the design of the control scheme. If they are taken into account, the results
do not change and what will happen is only a repetition of the results. Based on the above,
the value of speed (11.5 m/s) is determined as it is the optimal speed that achieves the
maximum power obtained from the turbine
Case 1: Symmetric-dip (50 %)
The system is simulated with two various approaches to test their performance dur-
ing disturbances. The grid-side converter is also controlled by the SVPWM approach. The
disturbance is represented by a three-phase voltage-dip. It occurs at 0.44 sec and it is
cleared at 0.55 sec. In the following part, the performance of DFIG is investigated. The
value of voltage dip is 50% of the rated voltage.
3. Results
The DFIG performance is investigated under a symmetric voltage-dip with the two
schemes. The stator voltage decreases at 0.44 sec, then it is recovered to its normal value
at 0.55 sec. In the first case, the crowbar is shorting the RSC when the voltage-dip is de-
tected. Then it is deactivated when the voltage is recovered in 0.55 sec. In the second case,
it is applied to the DVR that is controlled by ANFIS-controllers; this strategy proves that
it is the best solution to eliminate the voltage-dip effects.
The DFIG performance under disturbance is investigated for two schemes. The RSC
is controlled with SVPWM integrating with (Crowbar, DVR integrated with ANFIS-con-
trollers). These strategies are tested for three fault cases as follows:
The figure 11 explained how the rotor-voltage is increased from (0.7 to 1 Pu) during
the period from 0.44 sec to 0.55 sec due to the fault, which has negative effects on the RSC
performance.
EJEEE 2025, Vol.3 54 of 64
Figure 10. The stator-voltage with 50% symmetric-dip
occurring between 0.44 to 0.55 sec
Figure 11. The rotor-voltage signal under voltage-dip
Figure 12 shows the dc-link voltages for the three conditions. Figure (12 a) shows
how the resulting () reaches to a high value (1634 V) compared with the nominal one.
For the first strategy (crowbar), the DC-link voltage () experiences high values
during the voltage-dip as shown in figure (12 b). To overcome the crowbar limitations, the
DVR with ANFIS-controllers is used. The oscillations and high values are mitigated. In
the figure (12. c) the becomes more better compared with the Crowbar.
Figure 12. DC-Link Voltages under voltage dip a)
without any method
(b) Crowbar
The rotor-currents ( for the two methods are compared as shown in figure 13. The
(crowbar) protects the RSC from the over-currents. When the voltage-dip occurs, Crowbar
is activated [9]. The figure (13 b) demonstrates that the is oscillated during the voltage-
dip with imbalance state. This influences negatively on the DFIG performance. For the
proposed method, it is clearly shown that the resulting is sinusoidal and without any
distortion as shown in figure (11 c).
The figure (13) presents that the rotor currents have a high peak at 0.44 sec.
EJEEE 2025, Vol.3 55 of 64
(c) DVR+ANFIS
Figure 13. The rotor-currents under voltage dip a)
without any method
(b) Crowbar (c) DVR+ANFIS
The is obviously oscillating at 0.44 sec when the fault is started, the negative
peak is occurred due to the over rotor-current. At the end of the fault, the increased to
a positive value (0.8 N.m), this means abnormal operation of the machine.
when the crowbar is activated. This occurs due to the RSC control losing from 0.44
to 0.55 sec, as shown in figure (14 b). When the voltage is recovered, is not returned
to its nominal value before the voltage-dip, it still oscillates with considerable ripples. Fig-
ure (14 c) shows that is improved in this case. It is clearly shown how the proposed
technique mitigates the high peaks from the. This strategy also mitigates the ripples
from during the dip; this result agrees with ref [19].
Figure 14. Electromagnetic-Torque for DFIG with a) with-
out any method
(b) Crowbar
EJEEE 2025, Vol.3 56 of 64
Furthermore, the rotor-voltage is also enhanced, this in turn improves the and
enhances the performance during the voltage-dip occurrence, as shown in figure 15.
(a) DVR with ANFIS-controllers
Figure 15. Rotor voltage for DFIG using DVR+AN-
FIS
Table 3 summarizes the results of the DFIG's parameters in the three conditions in the
case of symmetric voltage-dip (50%) as follows.
Case2: 90% Symmetric dip:
In this case, a symmetric voltage dip (90%) as a severe condition is applied, in this
case the stator voltage is decreased from 0.44 sec to 0.55 sec, then it is recovered to its
nominal value, as seen in figure 16.
In this case, the increasing of stator voltage directly influences on the rotor voltage of
the machine. Figure 17 shows how rotor voltage increases at the time from 0.44 sec to 0.55
sec. This increasing from (0.7 to 1.02) has negative effects on the RSC converter, the DC-
link between RSC and GSC.
To address the above effects, methods are commonly used to protect the machine
and improve its performance during voltage disturbances. Figure 18 shows the resulting
DC-link voltage (). The first condition (figure 18 a) illustrates how the voltage dip
causes an increase in the ( from 1150 V to 2235 V without applying any protection
tool. This increase has the ability to damage the capacitor in the dc-link. By applying
Crowbar, the value is reduced to 1950 V, but the value is still very high and much
greater than the nominal value (1150). This is explained by the limitations of this method
and its inability to handle the excess voltage resulting from the fault. The designed ANFIS
with the DVR mitigates clearly the over-voltage in the DC-link compared with the other
method and maintains the nominal voltage during the fault, figure (18 c).
Figure 16. The stator-voltage with 90% voltage-dip oc-
curring between 0.44 to 0.55 sec
Figure 17. The rotor-voltage signal under voltage-dip
EJEEE 2025, Vol.3 57 of 64
Figure 18. DC-Link Voltages under voltage dip a)
without protection
(b) Crowbar
(c) DVR+ANFIS
Figure 19. The rotor-currents under voltage dip a)
without any method
(b) Crowbar
(c) DVR+ANFIS
The rotor current during the fault exceeded the nominal value by four times (4.3 A).
During a fault, the current not only increases but also becomes distorted and non-sinus-
oidal. As the fault's end time approaches, the current decreases and returns to its nominal
value, but it continues to fluctuate for a few cycles, figure (19 a). By applying conventional
protection Crowbar, it is observed that the rotor current during the fault reduced to 0.90
A. However, the distortions and imbalance in the waveform increased during the fault.
This is because activating the crowbar leads to a short circuit in the RSC and a loss of
control. After the fault clears, it is observed that the current remains unbalanced and high
EJEEE 2025, Vol.3 58 of 64
for several consecutive cycles, figure (19 b). Figure (19 c) shows the best performance of
the controllers during fault cases and the current becomes more sinusoidal.
Figure 20 explains the resulting electromagnetic torque in the three conditions. Fig-
ure 20-a shows how the voltage dip at t=0.44 sec experiences a negative peak, this peak
occurs due to the sudden increase in the rotor current. After that the torque's value be-
comes zero (0), at t=0.55 sec Tem starts to decline, and returns to the normal value. After
the fault is removed. It can be seen that the torque needs several cycles to stabilize. Figure
(20 b) represents that the Crowbar results in greater torque distortion, although it reduces
the fault current value it leads to excessive torque ripples. It can be noted that the current
distortions and imbalance during the fault led to the torque wave being shifted to positive
values. Therefore, the machine in this case works as a motor and consumes power from
the grid instead of providing power. The proposed strategy performed effectively in this
case, treating torque oscillations during the fault and resulting in a waveform that was
almost free of oscillations and within negative limits, meaning that the machine during
the fault remained in its working state as a generator and reliably supplied its power to
the grid. Table 4 summarizes the results of the DFIG's parameters in the three conditions
in the case of symmetric voltage-dip (90%) as following.
Figure 20. Electromagnetic-Torque for DFIG with a)
without any method
(b) Crowbar
(c) DVR with ANFIS-controllers
Figure 21. The stator-voltage with Asymmetric voltage-
dip occurring between 0.44 to 0.55 sec
EJEEE 2025, Vol.3 59 of 64
Table 3. Comparison between the two ap-
plied methods in the symmetric 50% volt-
age-dip
Vdc
THD%
Tem
ripples
Without
any Pro-
tection
Method
3.4 A
1634
V
78.70%
96.73%
With
Crowbar
0.9 A
1200
V
11.74%
142.34%
With
DVR+AN-
FIS
0.89
A
1151
V
0.99%
10%
Table 4. Comparison between the two ap-
plied methods in the symmetric 90% voltage-
dip
Vdc
(V)
THD%
Tem
Ripples
Without
any Pro-
tection
Method
4.3 A
2235
V
102.94%
85.71%
With
Crowbar
0.90
A
1950
V
89.77%
190.80%
With
DVR+AN-
FIS
0.89
A
1151
V
1.03%
11.39%
Case 3: Asymmetric-dip
In this case, the asymmetric voltage-dip is applied from t= 0.44 sec to t=0.55 sec. Fig-
ure 21 shows the voltage that is applied to the DFIG.
Figure (22 a) explains how the non-balance in the stator voltage impacts on the ,
at t=0.44 sec the increased to 1285 V. It remains distorted even after the fault has
been removed after several cycles. Treating the fault using Crowbar protection also
proved ineffective in this case as the entire resulting wave was distorted, full of ripples
and exceeded the nominal values, figure (22 b). Figure (22 c) presented the resulted volt-
age in the DC-Link due to using the proposed technique
Figure 22. DC-Link Voltages under voltage dip a)
without any method
(b) Crowbar
(c) DVR+ANFIS
Figure 23. The rotor-currents under voltage dip a) with-
out any method
EJEEE 2025, Vol.3 60 of 64
(b) Crowbar
(c) DVR+ANFIS
As with other previous faults, the current increased to 3 A at t= 0.44 sec, and remained
unbalanced until the fault finished at 0.55 sec. The crowbar has not provided an acceptable
performance in this case. The current suffers from non-balances and decreased to a non-
acceptable value of 0.25 A, figure (23 b). The proposed technique provides in this case the
best performance to process this type of fault effectively.
As in previous cases, it was observed that the asymmetrical fault led to clear dis-
tortions in the torque, in addition to a change in the torque signal to positive, which clearly
indicates a defect in providing power to the grid and the machine’s tendency to consume
during this period until the fault is removed, and thus an abnormal operating condition
affecting the machine on the one hand and the power provided to the grid on the other
hand, figure (24 a). figure (24 b) shows that the crowbar performance was very poor for
this type of fault, as the distortions increased due to the distortions that occurred in the
rotor current, figure (24 b). By regulating the constant voltage, the DVR+ANFIS was able
to reduce torque oscillations and keep the machine operating within the power supply
mode to the grid (figure 24 c).
Figure 24. Electromagnetic-Torque for DFIG with a)
without any method
(b) Crowbar
EJEEE 2025, Vol.3 61 of 64
Table 5. Comparison between the two applied
methods in the asymmetric voltage-dip
Vdc
THD%
Tem Rip-
ples
Without
any Pro-
tection
Method
2.9 A
1285
V
76.80%
114.1%
With
Crowbar
0.25
A
1208.5
V
22.69%
140.42%
With
DVR+AN-
FIS
0.89
A
1152
V
1.02%
10.25%
(a) DVR with ANFIS-controllers
In table (6), there is a comparison is made between the results obtained in this study
using the proposed method and the results obtained by previous reference studies using
other methods. The comparison shows that each method addressed one type of fault with-
out focusing on the other types. These methods were limited and did not improve the
performance of the entire machine. The proposed method, however, was more flexible
and able to adapt to all the situations that the grid was exposed to.
(means that the fault type is non studied, ( means that the applied method the faults
results have not improved, ) means that the faults results have improved.
Table 6. Comparison between the results of the current study and the results of previous studies
Ref
Method
Vdc
Tem Ripples
S(50%)
S (>50%)
AS
S
S
S
[6]
Crowbar+ANFIS
[8]
Minithreshold
Crowbar
[9]
Crowbar
[10]
STATCOM
[11]
SVC
STATCOM
SDSC
[12]
DVR+FUzzy
[13]
DVR+PSO
[10]
ADRC+DVR
EJEEE 2025, Vol.3 62 of 64
This
study
DVR+ANFIS
the faults results have not improved.
the faults results have improved
: non studied
Practical considerations
Designing a DVR for the DFIG protection is expensive due to the need for complex
electronic devices. Also, implementing ANFIS controller increases the complexity of the
DVR and thus increases the cost of implementing the method. Despite its complexity and
cost, its flexibility and ability to maintain the required performance led to avoiding dam-
age to electronic circuit breakers in the molds, destruction of the insulation of the coils in
the machine, damage to the capacitor, and other potential damages resulting from grid
failures.
4. Conclusions
In this study, two topologies are implemented to investigate the DFIG transient per-
formance improvement. The space-vector PWM is applied to the RSC with the two stud-
ied topologies. The DFIG nominal power is 3 MW. Crowbar and DVR integrated with
ANFIS controllers are implemented and compared. Three fault conditions are applied to
the DFIG in order to test the DVR with ANFIS flexibility and ability to adapt to faults.
The variations of DFIG parameters and their effects on the transient performance
are discussed. The simulation outcomes are presented that the measured DFIG's parame-
ters are stable during the voltage-dip when the proposed solution is applied. Moreover,
the oscillations in the and the are also mitigated. Furthermore, the need to short
the RSC during the voltage-dip is also eliminated. The limitations of the methods dis-
cussed in previous studies, which were summarized and compared with the proposed
method (Table 6) in this study showed that designing a controller (ANFIS) and training it
on network data for various faults and integrating the resulting controller with (DVR)
effectively enhanced the performance of the (DFIG). In addition, the proposed method
improved the electromagnetic torque by reducing its ripples, which was not addressed by
previous studies.
The model designed in this study is the basis for building a robust and strong system
to protect the DFIG and develop its performance to ensure the reliability of the energy
supplied to the grid via the wind system. For future work, this study recommends com-
bining the DVR with other example algorithms such as the (PSO), or (genetic algorithm)
and comparing their performance with the ANFIS that is applied in this study.
Acknowledgments: The authors declare, there is no any grant that support this work
Conflicts of Interest: Declare conflicts of interest or state “The authors declare no conflict of inter-
est.”
EJEEE 2025, Vol.3 63 of 64
Appendix
1. If (input1 is in1mf1) and (input2 is in2mf1) then (output is out1mf1) (1)
2. If (input1 is in1mf1) and (input2 is in2mf2) then (output is out1mf2) (1)
3. If (input1 is in1mf1) and (input2 is in2mf3) then (output is out1mf3) (1)
4. If (input1 is in1mf1) and (input2 is in2mf4) then (output is out1mf4) (1)
5. If (input1 is in1mf1) and (input2 is in2mf5) then (output is out1mf5) (1)
6. If (input1 is in1mf2) and (input2 is in2mf1) then (output is out1mf6) (1)
7. If (input1 is in1mf2) and (input2 is in2mf2) then (output is out1mf7) (1)
8. If (input1 is in1mf2) and (input2 is in2mf3) then (output is out1mf8) (1)
9. If (input1 is in1mf2) and (input2 is in2mf4) then (output is out1mf9) (1)
10. If (input1 is in1mf2) and (input2 is in2mf5) then (output is out1mf10) (1)
11. If (input1 is in1mf3) and (input2 is in2mf1) then (output is out1mf11) (1)
12. If (input1 is in1mf3) and (input2 is in2mf2) then (output is out1mf12) (1)
13. If (input1 is in1mf3) and (input2 is in2mf3) then (output is out1mf13) (1)
14. If (input1 is in1mf3) and (input2 is in2mf4) then (output is out1mf14) (1)
15. If (input1 is in1mf3) and (input2 is in2mf5) then (output is out1mf15) (1)
16. If (input1 is in1mf4) and (input2 is in2mf1) then (output is out1mf16) (1)
17. If (input1 is in1mf4) and (input2 is in2mf2) then (output is out1mf17) (1)
18. If (input1 is in1mf4) and (input2 is in2mf3) then (output is out1mf18) (1)
19. If (input1 is in1mf4) and (input2 is in2mf4) then (output is out1mf19) (1)
20. If (input1 is in1mf4) and (input2 is in2mf5) then (output is out1mf20) (1)
21. If (input1 is in1mf5) and (input2 is in2mf1) then (output is out1mf21) (1)
22. If (input1 is in1mf5) and (input2 is in2mf2) then (output is out1mf22) (1)
23. If (input1 is in1mf5) and (input2 is in2mf3) then (output is out1mf23) (1)
24. If (input1 is in1mf5) and (input2 is in2mf4) then (output is out1mf24) (1)
25. If (input1 is in1mf5) and (input2 is in2mf5) then (output is out1mf25) (1)
EJEEE 2025, Vol.3 64 of 64
References
1. Ananth, D. and G.N. Kumar, Fault ride-through enhancement using an enhanced field oriented control technique for converters of
grid connected DFIG and STATCOM for different types of faults. ISA transactions, 2016. 62: p. 2-18.
2. Soomro, M., et al., Performance improvement of grid-integrated doubly fed induction generator under asymmetrical and symmetrical
faults. Energies, 2023. 16(8): p. 3350.
3. Qu, Y., et al., Crowbar resistance value-switching scheme conjoint analysis based on statistical sampling for LVRT of DFIG. Journal
of Modern Power Systems and Clean Energy, 2019. 7(3): p. 558-567.
4. Jotwani, A. and Z. Rather. An advanced investigation on lvrt requirement in wind integrated power systems. in 1st International
Conference on large-scale grid integration of renewable energy in India. 2017.
5. Mohseni, M. and S.M. Islam, Review of international grid codes for wind power integration: Diversity, technology and a case for
global standard. Renewable and Sustainable Energy Reviews, 2012. 16(6): p. 3876-3890.
6. Noureldeen, O. and I. Hamdan, A novel controllable crowbar based on fault type protection technique for DFIG wind energy
conversion system using adaptive neuro-fuzzy inference system. Protection and Control of Modern Power Systems, 2018. 3(4): p.
1-12.
7. Qin, B., et al., Low-voltage ride-through techniques in DFIG-based wind turbines: A review. Applied Sciences, 2020. 10(6): p. 2154.
8. Pannell, G., D.J. Atkinson, and B. Zahawi, Minimum-threshold crowbar for a fault-ride-through grid-code-compliant DFIG wind
turbine. IEEE Transactions on Energy Conversion, 2010. 25(3): p. 750-759.
9. Din, Z., et al., Impact of grid impedance on LVRT performance of DFIG system with rotor crowbar technology. IEEE Access, 2019. 7:
p. 127999-128008.
10. Döşoğlu, M.K., Enhancement of LVRT capability in DFIG-based wind turbines with STATCOM and supercapacitor. Sustainability,
2023. 15(3): p. 2529.
11. Ganthia, B.P., S.K. Barik, and B. Nayak, Shunt connected FACTS devices for LVRT capability enhancement in WECS. Engineering,
Technology & Applied Science Research, 2020. 10(3): p. 5819-5823.
12. Sarker, K., et al., Power quality investigation with multilevel inverter by photovoltaic-fed dynamic voltage restorer. International
Journal of Modelling and Simulation, 2024: p. 1-27.
13. Kumar, A., V.M. Mishra, and R. Ranjan, LVRT enhancement in grid connected DFIG based wind turbine using PSO optimized
DVR. Distributed Generation & Alternative Energy Journal, 2020: p. 249-264.
14. Baala, Y., et al. Enhanced LVRT capability of wind turbine based on DFIG using dynamic voltage restorer controlled by ADRC-based
feedback control. in E3S Web of Conferences. 2021. EDP Sciences.
15. Patra, R.R. and A. Banerjee. Indirect vector control of DFIG based wind energy system using fuzzy-PI controller. in 2018 2nd
International Conference on Power, Energy and Environment: Towards Smart Technology (ICEPE). 2018. IEEE.
16. habib BENBOUHENNI, H., Application of seven-level neural space vector PWM in direct vector control system of doubly fed
induction generator for wind turbine. International Journal of Smart Grid-ijSmartGrid, 2019. 3(3): p. 163-171.
17. Jerin, A.R.A., et al., FRT Capability in DFIG based wind turbines using DVR with Combined Feed-Forward and Feed-Back Control.
Energy Procedia, 2017. 138: p. 1184-1189.
18. Khatoon, N. and S. Shaik, Review of different fault ride through (FRT) control strategies for a DFIG wind turbine. International
Research Journal of Engineering and Technology (IRJET), 2017. 4(11): p. 1571-1587.
19. Rini Ann Jerin, A., et al., FRT Capability in DFIG based Wind Turbines using DVR with Combined Feed-Forward and Feed-Back
Control. Energy Procedia, 2017. 138: p. 1184-1189 DOI: https://doi.org/10.1016/j.egypro.2017.10.233.
20. Hagh, M.T., et al., Fuzzy-based controller for DVR in the presence of DG. Procedia computer science, 2017. 120: p. 684-690.
