You are on page 1of 12

Electrical Power and Energy Systems 141 (2022) 108220

Contents lists available at ScienceDirect

International Journal of Electrical Power and Energy Systems


journal homepage: www.elsevier.com/locate/ijepes

A new differential protection algorithm for transmission lines connected to


large-scale wind farms
A. Saber a, *, M.F. Shaaban b, H.H. Zeineldin a, c
a
Electric Power Engineering Department, Cairo University, Giza, Egypt
b
Electrical Engineering Department, Faculty of Engineering, American University of Sharjah, Sharjah, United Arab Emirates
c
AdvancedPower & Energy Center, Electrical Engineering and Computer Science Department, Khalifa University, Abu Dhabi, United Arab Emirates

A R T I C L E I N F O A B S T R A C T

Keywords: The fluctuated output power of the large-scale wind farms (WF) and their fault transient characteristics have an
Large-scale wind farm adverse effect on the current differential protection systems of the transmission lines. With increased trans­
Signed correlation mission line length and in the case of weak output power from the WF, the differential current between both line
Differential protection
ends increases in normal conditions since the capacitive current is comparable to the load current. This paper
Transmission line
proposes a new differential protection algorithm for transmission lines connected to large-scale wind farms. The
proposed current differential protection algorithm is developed based on the signs of the phase current samples
at both line ends, instead of utilizing directly the current magnitudes. The similarity between the signs of the
phase current samples at both line ends is evaluated utilizing the signed correlation criterion, and the fault
detection index is calculated to discriminate the internal faults from other events. In addition, a new online
technique is introduced to adjust the protection settings considering the different output power levels of the WF.
The conducted PSCAD/EMTDC simulation studies confirm the acceptable performance of the proposed protec­
tion algorithm for numerous normal and fault scenarios, including different fault resistances and inception angles
as well as all fault types. The effect of line length, type of wind turbine generator, and different values of WF
output power are also considered.

1. Introduction [8], and the ground faults were only considered [9,10]. In [11], an
additional adaptive unit has been integrated with the conventional relay
The integration of the large-scale wind farms (WFs) into the trans­ to comply with the different operating conditions of the WF. However, it
mission level has increased dramatically, and it has become a common is required to train a large amount of data to deem all likely operating
trend [1]. However, the different fault characteristics of the large-scale scenarios of the WF. To avoid the impact of the generated non-power
WFs reduce the sensitivity of the protection system of the transmission frequency components by the WTG during faults, Prony technique is
grid [2–5]. In addition, the output power of the WF fluctuates signifi­ utilized for phasor estimation [12]. However, this technique does not
cantly, and some of the existing protection algorithms fail to operate in investigate the impact of fault resistance and external faults on its per­
case of weak WF output power [6]. Moreover, the generated non-power formance. Only for symmetrical line faults, a communication-based
frequency components by WTG during faults have a negative impact on scheme has been introduced to share over-reach trip signals between
the accuracy of the estimated fundamental frequency phasors, which in both installed relays at both line terminals [13]. In [14,15], a
turn affect the protection system performance [6,7]. communication-based protection technique has been developed to share
Several schemes have been proposed to overcome the over- and the required information between both installed relays at both line
under-reach problems associated with the conventional distance relays terminals. The apparent impedance is calculated for double-line to
[8–19]. The relay settings [8] and tripping boundaries [9,10] of the ground faults, while two directional elements at both line terminals have
distance protection have been adapted to comply with the different been provided for double- and three-line faults. In [16], the settings of
operating conditions of the WF. However, the equivalent impedances of the distance relay have been adapted for ground faults. However, the
both grid and WF sides are required to estimate the new relay settings equivalent impedance of the WF is initially required. Based on knowing

* Corresponding author.
E-mail address: a_saber_86@yahoo.com (A. Saber).

https://doi.org/10.1016/j.ijepes.2022.108220
Received 31 December 2021; Received in revised form 28 February 2022; Accepted 7 April 2022
Available online 13 April 2022
0142-0615/© 2022 Elsevier Ltd. All rights reserved.
A. Saber et al. International Journal of Electrical Power and Energy Systems 141 (2022) 108220

the equivalent impedances of both grid and WF sides, the distance relay • The effects of measurement and synchronization errors are taken
settings have been adapted in time-domain based on the R-L differential into account.
equations of the transmission line [17]. In [18], based on the zero- • The proposed approach requires less computational burden, low
sequence impedance, an adaptive technique has been proposed, which sampling frequency, and less communication bandwidth compared
can distinguish internal faults correctly using only local data. However, to the conventional differential protection technique.
this technique requires a time delay of 0.3 sec for correct fault
identification. The power system is modeled on PSCAD/EMTDC software consid­
Based on the above discussion, it is clear that additional information ering numerous external and internal scenarios, different fault re­
is required to enhance the performance of the distance protection for sistances, inception angles, and fault types. In addition, the proposed
transmission lines connected to large-scale WFs. Therefore, conven­ algorithm takes into consideration the influence of measurement and
tional differential protection (CDP) is a good option in such a case due to synchronization errors. The paper is organized as follows. The problem
its selectivity and fast operation [19–22]. However, it is shown in [6] statement and the proposed current differential protection algorithm are
that the sensitivity of the CDP is reduced when a double-line fault occurs shown in Sections 2 and 3, respectively. The performance evaluation of
on the transmission line. This is because the angle difference between the proposed algorithm is introduced in section 4. Finally, Section 5
the faulted currents for one phase of the faulted phases at both line ends summarizes the paper.
is less than 90◦ , and accordingly, the restraining current is higher than
the differential current for this phase. In addition, the fault currents of 2. Problem statement
both type-3 and type-4 WT generators involve three components, i.e.,
power frequency, non-power frequency, and harmonic components [6]. Consider the one-line diagram in Fig. 1 for a 220 kV, 60 Hz, 150 km
Therefore, these fault current characteristics have a negative impact on transmission line (W-G) connected between the wind side bus-W and the
the estimation accuracy of the power frequency phasors, which in turn grid side bus-G. The 300 MW (60 × 5 MW) WF comprises a type-3
degrades the performance of the power frequency phasor-based pro­ doubly-fed induction generator (DFIG) and an AC-DC-AC voltage
tection methods. source converter, and it is connected to bus-W through a 33/220 kV, 500
Considering the above reasons, time-domain based current differ­ MVA transformer with a leakage reactance of 10%, and Yg/Yg solidly
ential protection algorithms have been developed for transmission lines grounded. The detailed data of both rotor and grid sides controllers are
connected to large-scale WFs [6,7,23]. New pilot protection schemes presented in [26], including the DC-link chopper and crowbar circuit
have been developed based on structural similarity and Pearson corre­ protection. The WTG data is shown in Appendix A.1, and the WT step-up
lation similarity [6,23], respectively. However, with increasing the transformer is 0.69/0.69/33 kV, 5.5 MVA, and Yg/Yg/Yg solidly
transmission line length and in the case of weak output power from the grounded connection. Further, the characteristics of current trans­
WF, the values of structural similarity and Pearson correlation indices formers (CTs) are considered, and its data is given in Appendix A.2. The
reduce during normal conditions, making them similar to the case of an parameters of the line (W-G) and the grid source connected to bus-G are
internal fault. Accordingly, this results in incorrect operation of the shown in Appendices A.3 and A.4, respectively.
differential protection scheme during normal conditions. Moreover, for
external faults at the WF side, the protection system may incorrectly 2.1. Phasor-based conventional differential protection
operate due to the effect of the transient capacitive current, which can
affect the calculated correlation indices, as indicated in [23]. Therefore, Fault currents of DFIGs exhibit different characteristics in compari­
increasing the transmission line length has a negative influence on these son to synchronous generators. When a fault occurs close to the DFIG
schemes [6,23]. The lengths of the transmission lines connected to large- terminals, its terminal voltage will decrease significantly. In addition,
scale WF are usually short or medium. However, there are also some over-voltage and over-current will appear at the rotor side, resulting in
long lines, such as the double-circuit transmission line with a length of activation of the crowbar resistance. In this case, the transient fault
280 km in Egypt [24]. In [7], the transient capacitive current is current includes mainly three components; stable PF component,
compensated utilizing the Bergeron computation technique. However, attenuated rotor speed frequency component (RSFC), and attenuated DC
both voltage and current data at both line terminals as well as the line component [26]. It is pointed out that the attenuated RSFC may domi­
capacitance are required to compensate for the transient capacitive nate these components [13,17]. As the rotor speed of the DFIG often
current. In [25], a new fault detection technique has been proposed deviates from synchronous speed, the fault current can be expressed as a
utilizing the signs of the half-cycle superimposed positive-sequence sine-wave with non-power frequency components. When a fault occurs
currents at both line ends. However, different factors are not consid­ far from the DFIG terminals, the drop in its terminal voltage may be
ered, such as the measurement and synchronization errors. In addition, relatively small, and the activation of the crowbar resistance will not
this technique may operate improperly during normal conditions with occur. In this case, the transient fault current includes a steady-state
increasing the transmission line length and in the case of weak output sine-wave of PF component and three attenuated sine-wave of non-
power from the WF due to the effect of the transient capacitive current. power frequency components [27]. Thus, the fault current can also be
In this paper, a new current differential protection algorithm is expressed as a sine-wave with non-power frequency components [27].
proposed for transmission lines connected to large-scale WFs. The main Accordingly, in the presence of large-scale wind farms, the generated
features and contributions of this paper are summarized, as follows: non-power frequency components by the WTG during faults have a
diverse effect on the accuracy of the estimated power frequency phasors,
• The proposed protection algorithm is not based directly on the which in turn negatively affects the protection system performance
magnitudes of the fault currents, where the signed correlation value [6,7]. The protection of the conventional differential protection can be
(SCV) is calculated utilizing the signs of the phase current samples at written as:
both line ends. ⃒ ⃒ ⃒ ⃒ ⃒ ⃒
• A fault detection index is developed to discriminate the internal
⃒Idiff ,j ⃒ > Idiff ,th &⃒Idiff ,j ⃒ > Q⃒Irest,j ⃒ (1)
faults from other events, where a half-cycle time window is only ⃒ ⃒
needed. Internal faults can be discriminated up to fault resistance of where ⃒Idiff,j ⃒ is the absolute differential current of phase i between both
⃒ ⃒
100 Ω. line ends, and is equal to ⃒IWG,j − IGW,j ⃒ considering the flow direction of
• Due to the fluctuations in WF output power, a new online technique the phasor currents (IWG and IGW ) at both ends of the line (W-G). “j”
⃒ ⃒
is developed to adjust the protection settings considering the refers to the phase a, b, or c. Idiff,th is the startup threshold value. ⃒Irest,j ⃒ is
different output power levels of the WF. the restraint phasor current of phase i between both line ends, and is

2
A. Saber et al. International Journal of Electrical Power and Energy Systems 141 (2022) 108220

Fig. 1. One-line diagram of the simulated power system.

⃒ ⃒ {⃒ ⃒/ }
equal to ⃒IWG,j + IGW,j ⃒. Q is the restraint coefficient and is equal to 0.5 WF (30 MW). As shown, the ratio ⃒Idiff ⃒ |Irest | exceeds the restraint
[28]. In the case of weak output power from the WF, the effect of the line coefficient (Q) in the case of weak output power from the WF. This is due
capacitance on the conventional differential protection is required to be to the capacitive current being comparable to the load current in the
compensated, especially with an increase in line length. As an example, case of weak output power from the WF. Accordingly, in this case, the
for the transmission line (W-G) shown in Fig. 1 with a line length of 150 restraint coefficient (Q) should be increased or the line capacitive cur­
{⃒ ⃒/ }
km, the ratio ⃒Idiff ⃒ |Irest | is shown in Fig. 2 for normal condition for rent needs to be compensated to avoid unnecessary tripping in such
two cases i.e. full capacity of WF (300 MW), and weak output power of cases. However, the reliability of the conventional differential

⃒ ⃒/
Fig. 2. ⃒Idiff ⃒ |Irest | during normal conditions: (a) Full capacity of WF (300 MW), (b) Weak output power of WF (30 MW).

3
A. Saber et al. International Journal of Electrical Power and Energy Systems 141 (2022) 108220

protection may be reduced due to increasing the restraint coefficient ∑m ( )( )


(Q), especially in the case of high fault resistance. In addition, the n=1 iWG,j (nT) − μIWG,j iGW,j (nT) − μIGW,j
voltage and current data at both line terminals as well as the line pa­ PCCWG,j = √̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅
m ( )2̅√̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅
m ( )2 (3)
∑ ∑
rameters are required to compensate for the transient capacitive current. iWG,j (nT) − μIWG,j iGW,j (nT) − μIGW,j
n=1 n=1
Fortunately, in the case of weak output power from the WF, the
contribution of the grid side fault current is much higher than that of the where the index “j“ refers to the phase a, b, or c. iWG,j and iGW,j represent
wind side since the fault current of the wind side is low. Therefore, if an the instantaneous currents for phase j at both line ends. “T” and “n”
internal fault occurs in the case of weak WF output power, the differ­ represent the sampling time and the sample number, respectively. “m” is
ential fault current is much higher compared with the differential fault the number of samples in the time window. μIWG,j and μIGW,j are the mean
current when the output power of the WF is equal to the rated output values of phase j for both phase currents (iWG,j and iGW,j ) at both line
power of the WF. Therefore, as the output power of the WF decreases,
ends, respectively. σ 2IWG,j and σ 2IGW,j are the variances of phase j for both
the differential fault current increases. Accordingly, it is suitable to
adjust the protection settings relative to the different output power iWG,j and iGW,j , respectively. σIWG,j − IGW,j is the covariance between iWG,j and
levels of the WF. iGW,j . “L” is the difference between the maximum and minimum values
of both iWG,j and iGW,j , respectively. The values of SSIMWG,j and PCCWG,j
2.2. Current similarity-based differential protection tend to be equal to 1 in normal conditions or in the case of external
faults, while they are close to zero during internal faults. In [6], two
Recently, current similarity-based differential protection schemes threshold values (SSIMth,h and SSIMth,l ) are set at 0.9 and 0.1, respec­
have been proposed for transmission lines connected to large-scale WFs tively. A trip signal is initiated when the value of SSIMWG,j is within [0.1,
[6,23]. These schemes utilize the structural similarity (SSIM) technique 0.9]. In [23], a threshold value (PPCth ) is set at 0.9, a trip signal is
[6] and Pearson correlation coefficient (PCC) technique [23] to measure initiated when the value of PCCWG,j is less than 0.9.
the similarity between both current waveforms at both line ends. For short lines, both protection schemes have a successful perfor­
Considering the transmission line (W-G) shown in Fig. 1, the formulas of mance for a wide range of output power of the WF. The scheme, in [6], is
the SSIM and the PCC (SSIMWG,i and PCCWG,i ) between the instantaneous tested considering only a short line of 40 km. Although the scheme in
phase currents (iWG and iGW ) at both line ends are shown in (2) and (3), [23] is tested considering a medium line of 150 km, it is not tested
respectively. considering weak output power of WF. Considering that the length of
( )( ) line (W-G) is equal to 150 km, Fig. 3 shows the PCCWG for two cases i.e.
2μIWG,j μIGW,j + (0.01L)2 2σ IWG,j − IGW,i + (0.03L)2 full capacity of WF (300 MW), and weak output power of WF (30 MW).
SSIM WG,j = ( )( ) (2) As shown, the values of PCCWG for the three phases are less than the
μ2IWG,j + μ2IGW,j + (0.01L)2 σ2IWG,j + σ 2IGW,j + (0.03L)2
threshold value of 0.9 in the case of weak output power from the WF.
Accordingly, the protection scheme will operate incorrectly in this case.
For the protection scheme in [6], the values of SSIMWG for the three
phases are shown in Fig. 4 for the two aforementioned cases. As shown,
the values of SSIMWG for the three phases fall inside the tripping zone in

Fig. 3. The values of the PCCWG during normal conditions: (a) Full capacity of WF (300 MW), (b) Weak output power of WF (30 MW).

4
A. Saber et al. International Journal of Electrical Power and Energy Systems 141 (2022) 108220

Fig. 4. The values of the SSIMWG during normal conditions: (a) Full capacity of WF (300 MW), (b) Weak output power of WF (30 MW).

the case of weak output power from the WF, and the protection scheme expected that the waveforms of the fault current (iWG ) on the wind side is
will also operate incorrectly in this case. As mentioned previously, the different from that of iGW on the other side of the line. In the case of the
reason for that is due to the adverse effect of the line capacitance in the external faults, both instantaneous currents (iWG and iGW ) on both line
case of weak output power from the WF, and one of the possible solu­ terminals are approximately similar based on Kirchhoff’s current law.
tions is to adjust online the protection settings considering the different The proposed differential protection algorithm is developed considering
output power levels of the WF. these characteristics. The signed phase currents at both line (W-G) ter­
{ }
minals i*WG (t) and i*GW (t) can be defined in discrete form:
3. Proposed differential protection algorithm {
1if iWG,j (nT) > 0
i*WG,j (nT) = (4)
− 1if iWG,j (nT) ≤ 0
As previously aforementioned, the fault current characteristics of the
DFIGs have a negative impact on the estimation accuracy of the power {
1if iGW,j (nT) > 0
frequency phasors, specifically the estimation accuracy of the phase i*GW,j (nT) = (5)
− 1if iGW,j (nT) ≤ 0
angle. This in turn degrades the performance of the power frequency
phasor-based protection methods, specifically the conventional phase If the current sample is higher than zero, the corresponding signed
comparison directional scheme. The proposed time-domain differential value will be equal to 1, otherwise, it will be equal to − 1. Both signed
protection algorithm is based on the same concept introduced in [6,23]. current samples of phase j at both line ends are utilized to derive the
However, the proposed protection algorithm relies on the signs of the equation of the signed correlation value (SCV) for phase j, as follows:
phase currents at both line ends, where the signed correlation method is ∑m { * }
*
utilized to measure the similarity between the signs of the phase currents n=1 iWG,j (nT) • iGW,j (nT)
at both line ends. The signed correlation method depends on the simi­ SCV j = (6)
m
larity of the shapes and polarities of the two current signals instead of
comparing the magnitudes of both signals [29,30]. Thus, it can be where “m” is the number of samples in the time window. The symbol ( •
adapted for protecting the transmission lines connected to large-scale ) refers to the product of the two samples. When a fault occurs, for
WF due to the different fault characteristics at both line ends. In addi­ example, at fault point F2, as shown in Fig. 1, iWG and iGW are approx­
tion, one of the merits of the signed correlation method is the low imately equal to each other or they are slightly different from each other
computational burden. Moreover, a new technique is introduced to due to the effect of line stray capacitance, which results in a slight
adjust online the settings of the proposed differential protection shifting of the phase current angles. Therefore, the value of the SCV for
algorithm. each phase will be approximately equal to 1. On the contrary, when a
fault occurs at fault point F1, as shown in Fig. 1, iWG and iGW will be out-
of-phase with each other. Therefore, the maximum expected value of the
3.1. Calculations of signed correlation values (SCVs)
SCV for the faulted phase will be − 1. Based on the above discussion, the
proposed fault detection index (FDI) for phase j is defined:
To explain the concept of the signed correlation method, consider the
flow direction of the phase currents (IWG and IGW ) at both terminals of FDI j = 1 − SCV j (7)
line (W-G) shown in Fig. 1. When a fault occurs at fault point F1, it is

5
A. Saber et al. International Journal of Electrical Power and Energy Systems 141 (2022) 108220

In normal condition or in the case of external faults, it is expected the calculated for each phase utilizing (6). Thus, the FDI value is calculated
FDI value is approximately equal to zero. On the other hand, in the case for each phase utilizing (7). A trip signal is initiated once the FDI value
of internal faults, the maximum expected FDI value is equal to 2. As the for any phase exceeds the threshold value (DFDIth ). In addition, it is clear
protection system must be secure and stable, especially during external from (4) up to (8) that the required computational burden for the pro­
faults and in the case of weak output power from the WF, a threshold FDI posed differential protection algorithm is negligible.
value (FDIth ) is required to be selected. The criterion for adapting the
protection settings online considering the different output powers of the 3.3. Communication issues
WF are explained in the next subsection.
The proposed protection algorithm can be implemented utilizing a
low sampling rate of 1.2 kHz. Therefore, a half-cycle time window of (10
3.2. Online adaptive protection settings samples) is utilized to calculate both SCV and FDI values for each phase.
Therefore, 10 binary samples (1 or − 1) are required to be shared be­
As the expected value of the FDI falls within [0, 2], the FDIth value is tween both sides of the line for each phase per half-cycle. In addition, the
set to be higher than zero. The equation, which calculates the FDIth value time delays of the communication links between both line terminals are
considering different output power of the WF, is written, as: one of the important factors to achieve better performance for any dif­
FDI th = kc + (1 − Ipu ) (8) ferential protection scheme. According to IEEE standard [32], the time
delay includes relay interface delay, fiber optic propagation delay, and
where kc is a constant, and Ipu is the pre-fault current in pre-unit of the substation multiplexer delay. The relay interface delay is typically in the
WF RMS rated current. The constant (kc ) can be determined based on range of 1–5 msec, the fiber optic propagation delay is typically 5 µsec/
three factors. The first factor considers the measurement and synchro­ km, and the substation multiplexer delay is typically 0.37 msec. Thus,
nization errors. The current measurement errors are typically within ± for different line lengths up to 400 km, the maximum time delay will not
5%, and the time or angle synchronization error is typically equal to ± exceed 7.37 msec. Thus, since a half-cycle time window is utilized for
26 µsec (±0.5616◦ ) for a 60 Hz system [31]. In addition, the second the proposed protection algorithm, the maximum tripping time is within
factor is the reliability or the marginal factor, which can be set at 10%. one cycle, which is typically acceptable for transmission lines
Further, the third factor is the security factor, which ensures that the protection.
protection algorithm will not operate in case of external faults. The se­ As for any differential protection technique, the proposed algorithm
curity factor can also be set at 10%. Therefore, the constant (kc ) is set at has some limitations. One of these limitations is related to the avail­
0.3 in this paper. Accordingly, the FDIth value falls in the range of [0.3, ability of the communication system. If the communication links fail or
1.3], where the FDIth value is equal to 0.3 when the WF output power is are not available for any reason, the proposed algorithm will not be
equal to the rated power, and it is equal to 1.3 when the WF output applicable for line protection. However, there are several solutions for
power is equal to zero. Based on the proposed online adaptive setting in this case. Practically, the transmission lines are protected using two
(8), it is worth noting that there is no need to compensate for the tran­ main protective relays (main-1 and main-2 relays) as well as overcurrent
sient capacitive current, unlike the scheme in [7], where both voltage protection functions. Main-1 and main-2 protective relays can be current
and current data at both line terminals as well as the line capacitance are differential relays or distance relays. Thus, one of the solutions is to
required to compensate for the transient capacitive current. Also, the utilize both main-1 and main-2 relays as differential relays. However, in
pre-fault current in pre-unit (Ipu ) can be calculated continuously for each this case, the communication links should be duplicated; one commu­
periodic cycle to update the threshold value accordingly. It can be noted nication link for each relay, which impose additional cost. Another so­
that the selected threshold value in this paper does not depend on the lution is to utilize main-1 relay as a differential relay and main-2 relay as
parameters of the power system connected to the transmission line. a distance relay. Further, the overcurrent protection schemes are prac­
The fault detection steps are summarized in Fig. 5. The current tically utilized as backup protection for the transmission lines in case of
samples iWG (nT) and iGW (nT) are measured, and the SCV value is communication failure. The second limitation of the proposed algorithm
is related to the high impedance ground faults. For ground faults with
very high fault impedance of more than 100 Ω, the ground fault currents
are limited to small values, which are comparable to the rated current of
the transmission lines. For such cases, the proposed algorithm cannot
detect this type of ground faults. However, in such cases, protection
schemes, which are mainly non-directional/directional earth fault pro­
tection schemes, can detect this type of ground faults since they rely on
measuring the ground fault currents.

4. Performance evaluation

The power system shown in Fig. 1 is simulated on PSCAD/EMTDC


program, and the necessary calculations are implemented utilizing
MATLAB program. To assess the performance of the proposed protection
algorithm, different internal and external faults are simulated, including
different fault distances, resistances, and inception angles. All types of
faults are simulated, including single-phase to ground (1-ph-g), double-
phase to ground (2-ph-g), double-phase (2-ph), and three-phase (3-ph)
faults. For all cases, the sampling frequency is 1.2 kHz, and the wind
speed is set at 11 m/sec. Also, CT measurement error of ± 5% and angle
synchronization error of ± 1◦ are considered for all test cases.

4.1. Normal conditions

Fig. 5. Steps of proposed fault detection. To assess the performance of the proposed algorithm for different

6
A. Saber et al. International Journal of Electrical Power and Energy Systems 141 (2022) 108220

output power levels of the WF, the two previous cases are repeated proposed protection algorithm is satisfactory against external faults.
considering the full capacity of WF (300 MW) and weak output power of
WF (30 MW). As mentioned earlier, the length of the line (W-G) is equal 4.3. Effect of fault location and different values of WF output power
to 150 km. The current waveforms for phase b at both wind and grid
sides are shown in Fig. 6 for both WF output power values of 300 MW All internal fault types are simulated at different fault distances (0.1,
and 30 MW. In the case of 300 MW, the slight difference between both 0.5, and 0.9 per-unit) and different values of WF output power (30, 100,
current waveforms at both line ends is due to the effect of measurement 200, 300 MW), as shown in Table 1. The fault inception angle (δF ) and
and synchronization errors. It can be noted that the line capacitance is resistance (RF ) are set at 0◦ and 1 Ω for these cases, respectively. As
not significant in this case, and this is reflected in the FDI values shown shown in all cases in Table 1, the FDI values exceed the adaptive
in Fig. 7a. The FDI value for each phase is less than the threshold value of threshold values for different values of WF output power. Thus, a trip
0.3. On the contrary, in the case of 30 MW, the effect of line capacitance signal is initiated in all cases. In addition, the minimum recorded FDI
is more significant, and this is reflected in the FDI values shown in value is equal to 1.4, which exceeds the threshold values of 0.3 and 0.97,
Fig. 7b. The FDI values for the three phases fall within [0.6, 0.8]. as shown for the cases of WF output power of 300 and 100 MW,
However, all FDI values are less than the adaptive threshold value of 1.2 respectively. As an example, Fig. 9 shows the FDI values for the three
in this case. Comparing with the conventional differential protection phases during a-c fault occurrence at 10% of line length and WF output
scheme and the schemes in [6,23], the proposed algorithm shows suc­ power of 30 MW. The FDIs values for phases a and c are equal to 2
cessful performance considering different values of WF output power. exceeding the threshold value of 1.2. Therefore, a trip signal is initiated
to isolate the transmission line. The obtained results assure the accept­
4.2. External bus faults able response of the proposed protection algorithm for all fault types and
different values of WF output power.
The most severe external faults are the faults with very low fault
resistance at the adjacent buses (bus-W and bus-G) to the line (W-G). A 4.4. Effect of fault resistance
three-phase solid fault is conducted on bus-G considering WF output
power values of 300 MW and 30 MW. The FDI values for both cases are Different fault cases are simulated considering three values of fault
shown in Fig. 8. In the case of 300 MW, the threshold value is adjusted to resistance (0.01, 10, and 100 Ω) and all fault types, as illustrated in
be 0.3, the FDI values in Fig. 8a are less than 0.3. Therefore, no trip Table 2. The fault inception angle (δF ) and fault distance are set at 90◦
signal is initiated. In the case of 30 MW, the threshold value is adjusted and the middle of line (W-G), respectively. In addition, different values
to be 1.2 according to (8) as the pre-fault current is equal to 0.1 per-unit of WF output power (20, 50, 150, 250 MW) are considered. As shown in
(30/300 = 0.1 per-unit), the maximum value of the FDI in Fig. 8b is Table 2, the FDI values exceed the adaptive threshold values for different
equal to 1, which is less than 1.2. Therefore, no trip signal is initiated values of WF output power. Thus, a trip signal is initiated in all cases.
also in this case. The two cases are repeated for the same fault on bus-W, Further, the minimum recorded FDI value is equal to 1, which exceeds
and similar results are obtained. It can be noted that the security of the the threshold value of 0.47, as shown for the case of WF output power of

Fig. 6. Current waveforms for phase b at both line terminals during normal conditions: (a) WF output power = 300 MW, (b) WF output power = 30 MW.

7
A. Saber et al. International Journal of Electrical Power and Energy Systems 141 (2022) 108220

Fig. 7. FDI values for the three phases during normal conditions: (a) WF output power = 300 MW, (b) WF output power = 30 MW.

Fig. 8. FDI values for the three phases during a solid three-phase fault on bus-G: (a) WF output power = 300 MW, (b) WF output power = 30 MW.

8
A. Saber et al. International Journal of Electrical Power and Energy Systems 141 (2022) 108220

Table 1
FDI values for all fault types and different WF output power. Table 2
FDI values for all fault types and different values of wf output power.
Output Power Fault type DFDIth Max. FDI
Fault location (p.u.) Output Power Fault type DFDIth Max. FDI
Fault resistance (Ω)
0.1 0.5 0.9
0.01 10 100
30 MW a-g 1.2 2.0 2.0 2.0
a-b-g 2.0 2.0 2.0 20 MW b-g 1.23 2.0 2.0 2.0
a-c 2.0 2.0 2.0 a-c-g 2.0 1.8 2.0
a-b-c 2.0 2.0 2.0 b-c 2.0 2.0 2.0
a-b-c 2.0 2.0 2.0

100 MW b-g 0.97 1.4 1.4 1.4


a-c-g 2.0 2.0 2.0 50 MW c-g 1.13 1.8 1.6 1.6
b-c 2.0 2.0 2.0 b-c-g 2.0 2.0 2.0
a-b-c 2.0 2.0 2.0 a-b 2.0 2.0 2.0
a-b-c 2.0 2.0 2.0

200 MW c-g 0.67 2.0 2.0 2.0


b-c-g 2.0 2.0 2.0 150 MW a-g 0.8 1.6 1.6 1.6
a-b 2.0 2.0 2.0 a-b-g 2.0 2.0 1.8
a-b-c 2.0 2.0 2.0 a-c 1.6 2.0 1.8
a-b-c 2.0 2.0 2.0

300 MW a-g 0.3 1.4 1.6 1.8


a-b-g 1.8 1.8 2.0 250 MW b-g 0.47 2.0 2.0 1.0
a-c 2.0 2.0 2.0 a-c-g 1.6 1.6 1.4
a-b-c 2.0 2.0 2.0 b-c 2.0 2.0 1.8
a-b-c 2.0 2.0 2.0

250 MW.
Table 3
4.5. Effect of different line lengths FDI values for different line lengths and WF output power.
Output Power Fault type DFDIth Max. FDI
In previous test cases, the line length is equal to 150 km. Considering Line length (km)
different line lengths (50 km, 100 km, and 200 km), all internal fault
50 100 200
types are simulated at different WF output power (30, 100, 200, 300
MW), as shown in Table 3. The inception angle (δF ) and distance are set 30 MW c-g 1.2 2.0 2.0 2.0
b-c-g 2.0 2.0 2.0
at 180◦ and 10% of line length, respectively. The FDI values exceed the
a-b 2.0 2.0 2.0
threshold values for different values of WF output power and line length. a-b-c 2.0 2.0 2.0
Accordingly, a trip signal is initiated in all cases. Moreover, the mini­
mum FDI value is 1, which exceeds the threshold values of 0.3 and 0.67,
100 MW a-g 0.97 2.0 2.0 2.0
as shown for WF output power values of 300 and 200 MW, respectively. a-b-g 2.0 2.0 1.8
The obtained results prove the applicability of the proposed algorithm a-c 2.0 2.0 2.0
for different line lengths without the need for compensating the tran­ a-b-c 2.0 2.0 2.0
sient capacitive current.
200 MW b-g 0.67 1.6 1.2 1.0
4.6. Effect of type-4 wind farm a-c-g 2.0 2.0 2.0
b-c 2.0 2.0 2.0
a-b-c 2.0 2.0 2.0
In previous test cases, the proposed protection algorithm considers
type-3 WT model. To check also the effect of type-4 WT model, large-
300 MW c-g 0.3 2.0 1.8 1.0
scale WF with rated power of 5 × 90 = 450 MW are considered. The b-c-g 2.0 1.8 1.8
detailed data of type-4 WT model are provided in [33], including the a-b 1.8 1.8 1.6
electrical and mechanical components. Different fault cases are simu­ a-b-c 2.0 2.0 2.0
lated considering all fault types, three values of fault resistance (0.01,

Fig. 9. FDI values for the three phases during a-c fault at 10% of line length and WF output power = 30 MW.

9
A. Saber et al. International Journal of Electrical Power and Energy Systems 141 (2022) 108220

10, and 100 Ω), and different values of WF output power (50, 150, 300,
450 MW). The results are recorded in Table 4, where the FDI values
exceed the adaptive threshold values for different values of WF output
power. Thus, a trip signal is initiated in all cases. Further, the minimum
recorded FDI value is equal to 0.6, which exceeds the threshold value of
0.3, as shown for the case of WF output power of 450 MW.

4.7. Effect of large load switching, capacitor switching, and noise

To show the effect of the large load switching, a large load is installed
at terminal G, where the capacity of this load is equal to the WF rating.
This load capacity is equal to 300 MW with a power factor of 0.8. Both Fig. 10. FDI values for the three phases during load disconnecting.
cases of connecting and disconnecting this load are simulated at instant
0.1 s considering a WF output power of 300 MW. In addition, Gaussian
noise is added to the current samples. To verify the robustness of the
proposed algorithm, a signal-to-noise ratio of 40 dB is considered [34].
For both cases, no trip signal is initiated as the values of the FDI are less
than the threshold value. For example, in the case of load disconnecting,
the FDI values for each phase are given in Fig. 10. As shown, the FDI
values are less than the threshold value of 0.3.
In addition, the proposed algorithm is tested for capacitor switching.
The line (W-G) is compensated with shunt compensation installed at
terminal G. The capacity of the shunt compensation is equal to 50 MVar.
Both cases of connecting and disconnecting the shunt compensation are
simulated at instant 0.1 s considering the WF output power of 300 MW Fig. 11. FDI values for the three phases during capacitor connection.
and a signal-to-noise ratio of 40 dB. For both cases, no trip signal is
initiated as the values of the FDI are less than the threshold value. For the case of weak WF output power; especially with increasing the
example, in the case of capacitor connection, the FDI values for each lengths of the transmission lines. The proposed protection algorithm
phase are given in Fig. 11. As shown, the FDI values are less than the solves this issue by online adaptive protection settings. In [7] and [19],
threshold value of 0.3. long lines with lengths of 200 km and 400 km are considered, respec­
tively. However, the effect of the weak output power from the WF during
normal conditions has not been considered in both schemes. In [20], the
4.8. Comparative evaluation
introduced differential protection scheme is based on the decision tree
technique. Accordingly, it cannot be applied for new configurations of
Due to the intermittency of wind speed, the voltage, frequency, and
transmission lines since training data is initially required. Therefore,
output power of the WTG continuously fluctuate. However, the fluctu­
training data-based schemes are not utilized in practice. In [22], the
ations of the voltage and frequency can be controlled using modern
total capacity of the WF is very small (9 MW), and a short line of 30 km is
power electronics-based technology [19]. In addition, the output power
utilized. Therefore, considering the effect of weak WF output power and
of the WF has a nonlinear relationship with the wind speed, resulting in
long line length, the maloperation problem under the effect of weak WF
limiting the output power capability of the WF. As aforementioned in
output power and long line length does not exist for this scheme.
Sections 2.1 and 2.2, previous protection schemes might maloperate in
In Table 5, the proposed time-domain differential protection algo­
rithm is compared with other time-domain differential protection al­
Table 4
gorithms in literature [6,7,23] as well as the conventional differential
FDI values for type-4 WF considering different fault resistances and WF output
protection scheme. The basic method of fault detection for each pro­
power.
tection is provided in the table. All schemes use the currents at both line
Output Power Fault type DFDIth Max. FDI
terminals except the scheme in [7], which requires voltage and current
Fault resistance (Ω)
data as well as line parameters. The considered line lengths are short for
0.01 10 100 the schemes in [6] and [23], and are considered long for the proposed
50 MW a-g 1.19 2.0 2.0 2.0 scheme and that in [7]. With increasing the line length and in the case of
a-b-g 2.0 2.0 2.0 weak output power from the WF, the structural similarity index for the
a-c 2.0 2.0 2.0
scheme in [6] and the Pearson correlation index for the scheme in [23]
a-b-c 2.0 2.0 2.0
reduce during normal conditions, making them similar to the case of an
internal fault. In addition, for external faults at the WF side, the transient
150 MW b-g 0.97 2.0 1.8 2.0
capacitive current may have an adverse effect on the calculated indices
a-c-g 2.0 1.8 2.0
b-c 1.6 1.8 2.0 for both schemes. For the conventional protection scheme, it operates
a-b-c 2.0 2.0 2.0 incorrectly in normal conditions since the capacitive current for long
line is comparable to the load current in the case of weak output power
300 MW c-g 0.63 2.0 2.0 1.4 from the WF. In [7], this problem was solved by compensating the
b-c-g 2.0 1.8 2.0 transient capacitive current of the line. However, the voltage and cur­
a-b 2.0 2.0 1.8 rent data as well as the line parameters are required to achieve this
a-b-c 2.0 2.0 1.4
compensation. On the other hand, as aforementioned, the proposed
protection algorithm solves this problem by adapting online the pro­
450 MW a-g 0.3 2.0 1.8 0.6 tection settings. Lastly, since the fault currents of WTGs involve non-
a-b-g 2.0 2.0 1.0
a-c 2.0 2.0 2.0
power frequency and harmonic components, the performance of the
a-b-c 2.0 2.0 1.8 conventional protection scheme is degraded due to the inaccurate

10
A. Saber et al. International Journal of Electrical Power and Energy Systems 141 (2022) 108220

Table 5
Comparative evaluation with previous algorithms.
Item Proposed Conventional [6] [7] [23]

Method Basis Signed Differential Structural Model Recognition and Pearson


Correlation Characteristic Similarity Unsymmetrical Correlation
Vector Reconstruction
Required Input Current Current Current Voltage, Current, Line Current
Parameters
Line Length 200 km – 40 km 200 km 40 km
Problem of Capacitive Current for Long Lines and No Yes Yes No Yes
Weak output power
Problem of Accurate Calculations of PF Phasors No Yes No No No

estimation of the power frequency (PF) phasors. Other schemes do not measurement and time synchronization errors are taken into
experience this problem since they are developed in time-domain. consideration.

5. Conclusions CRediT authorship contribution statement

This paper introduces a new current differential protection algorithm A. Saber: Conceptualization, Methodology, Investigation, Valida­
for transmission lines connected to large-scale WF. The proposed pro­ tion, Software, Writing – original draft. M.F. Shaaban: Conceptualiza­
tection algorithm is based on the signs of the phase current samples at tion, Writing – review & editing. H.H. Zeineldin: Investigation,
both line terminals. The SCV is developed to measure the similarity Methodology, Writing – review & editing.
between the signs of the current samples at both line ends, and the FDI is
calculated to differentiate the internal faults from other events. In Declaration of Competing Interest
addition, the protection settings are adjusted online considering the
different output power levels of the large-scale WF without the need for The authors declare that they have no known competing financial
compensating the capacitive current, especially with increasing the line interests or personal relationships that could have appeared to influence
length. PSCAD/EMTDC simulation studies assure the high efficacy of the the work reported in this paper.
proposed protection algorithm for different output power levels of the
large-scale WF as well as external and internal faults. In addition, the Acknowledgement
proposed algorithm offers acceptable performance considering all fault
types, different fault distances and resistances, different line lengths, as This work was supported by American University of Sharjah under
well as type-3 and type-4 WT models. Moreover, the effect of Award No. FRG20-L-E112.

Appendix

A.1. WT generator data

Parameter Value

Rated power (MW) 5


Rated stator voltage (V) 690
Rated frequency (Hz) 60
Rotor resistance (per-unit) 0.00607
Rotor leakage inductance (per-unit) 0.11
Stator resistance (per-unit) 0.0054
Stator leakage inductance (per-unit) 0.10
Magnetizing inductance (per-unit) 3.5
Mutual inductance (per-unit) 0.02

A.2. CT data

Parameter Value

Model JA model
Ratio 1000/1 A
Burden impedance 2 + 0.302i Ω

A.3. Transmission line (W-G) parameters

⎡ ⎤
0.11471 0.09728 0.094419
RW− G = ⎣ 0.09728 0.12064 0.09728 ⎦ (Ω/km)
0.094419 0.09728 0.11471

11
A. Saber et al. International Journal of Electrical Power and Energy Systems 141 (2022) 108220

⎡ ⎤
0.71573 0.36759 0.31676
XW− G = ⎣ 0.36759 0.71283 0.36759 ⎦ (Ω/km)
0.31676 0.36759 0.71573
⎡ ⎤
3.6747 − 0.8697 − 0.3625
YW− G = 10 − 6
× ⎣ − 0.8697 3.9331 − 0.8697 ⎦ (S/km)
− 0.3625 − 0.8697 3.6747

A.4. Grid parameters data

Sequence Positive Zero

Value 1.6 + 18.5i Ω 2.8 + 32.4i Ω

References [17] Chen Y, Wen M, Yin X, Cai Y, Zheng J. Distance protection for transmission lines of
DFIG-based wind power integration system. Int J Elect Power Energy Syst 2018;
100:438–48.
[1] Georgilakis PS. Technical challenges associated with the integration of wind power
[18] Fang Yu, Jia Ke, Yang Z, Li Y, Bi T. Impact of Inverter-Interfaced Renewable Energy
into power systems. Renew Sustain Energy Rev 2008;12(3):852–63.
Generators on Distance Protection and an Improved Scheme. IEEE Trans Ind Electr
[2] Telukunta V, Pradhan J, Agrawal A, Singh M, Srivani SG. Protection challenges
2019;66(9):7078–88.
under bulk penetration of renewable energy resources in power systems: a review.
[19] Tripathy LN, Jena MK, Samantaray SR. Differential relaying scheme for tapped
CSEE J. Power and Energy Syst. 2017;3(4):365–79.
transmission line connecting UPFC and wind farm. Int J Elect Power Energy Syst
[3] Morren J, de Haan SWH. Short-Circuit Current of Wind Turbines With Doubly Fed
2014;60:245–57.
Induction Generator. IEEE Trans Energy Convers 2007;22(1):174–80.
[20] Jena MK, Samantaray SR, Tripathy L. Decision tree-induced fuzzy rule-based
[4] Liu W, Wu Y, Lee C, Chen C. Effect of Low-Voltage-Ride-Through Technologies on
differential relaying for transmission line including unified power flow controller
the First Taiwan Offshore Wind Farm Planning. IEEE Trans Sustain Energy 2011;2:
and wind-farms. IET Gener Transm Distrib 2014;8(12):2144–52.
78–86.
[21] Ghorbani A, Mehrjerdi H, Al-Emadi NA. Distance-differential protection of
[5] Okedu KE, Muyeen SM, Takahashi R, Tamura J. Wind Farms Fault Ride Through
transmission lines connected to wind farms. Int J Elect Power Energy Syst 2017;89:
Using DFIG With New Protection Scheme. IEEE Trans Sustain Energy 2012;3(2):
11–8.
242–54.
[22] Prasad CD, Biswal M, Abdelaziz AY. Adaptive differential protection scheme for
[6] Zheng L, Jia Ke, Bi T, Yang Z, Fang Yu. A Novel Structural Similarity Based Pilot
wind farm integrated power network. Int J Elect Power Energy Syst 2020;187:
Protection for Renewable Power Transmission Line. IEEE Trans Power Del 2020;35
106452.
(6):2672–81.
[23] Jia K, Li Y, Fang Y, et al. Transient current similarity based protection for wind
[7] Ma K, Chen Z, Leth Bak C, Liu Z, Castillo M, Torres-Olguin RE, et al. Novel
farm transmission lines. Appl Energy 2018;225:42–51.
differential protection using model recognition and unsymmetrical vector
[24] Marcelino M, Steven S. Transmission Expansion for Renewable Energy Scale-Up:
reconstruction for the transmission line with wind farms connection. Int J Elect
Emerging Lessons and Recommendations. Washington, DC: World Bank Study;
Power Energy Syst 2020;123.
2012.
[8] Pradhan AK, Jos G. Adaptive distance relay setting for lines connecting wind farms.
[25] Biswas S, Nayak PK. A New Approach for Protecting TCSC Compensated
IEEE Trans Energy Convers 2007;22(1):206–13.
Transmission Lines Connected to DFIG-Based Wind Farm. IEEE Trans Ind
[9] Dubey R, Samantaray SR, Panigrahi BK. Simultaneous impact of unified power flow
Informatics 2021;17:5282–91.
controller and off-shore wind penetration on distance relay characteristics. IET
[26] Type 3 Wind Turbine Generator (WTG), Jan. 18 2018. [Online]. Available: https://
Gener Transm Distrib 2014;8(11):1869–80.
www.pscad.com/knowledge-base/article/495.
[10] Dubey R, Samantaray SR, Panigrahi BK, Venkoparaoc GV. Adaptive distance relay
[27] Sulla F, Svensson J, Samuelsson O. Symmetrical and unsymmetrical short-circuit
setting for parallel transmission network connecting wind farms and UPFC. Int J
current of squirrel-cage and doubly-fed induction generators. Elect Power Syst Res
Elect Power Energy Syst 2015;65:113–23.
J 2011;81:1610–8.
[11] Sadeghi H. A novel method for adaptive distance protection of transmission line
[28] El-Naggar A, Erlich I. Fault Current Contribution Analysis of Doubly Fed Induction
connected to wind farms. Int J Elect Power Energy Syst 2012;43(1):1376–82.
Generator-Based Wind Turbines. IEEE Trans Ener Conv 2015;30:874–82.
[12] Guajardo LAT, Enríquez AC, Leonowicz Z. Error compensation in distance relays
[29] Lian J, Garner G, Muessig D, Lang V. A simple method to quantify the
caused by wind power plants in the power grid. Elect Power Syst Res J 2014;106:
morphological similarity between signals. Signal Process 2010;90(2):684–8.
109–19.
[30] Saber A. A Backup Protection Algorithm for Bipolar Line-Commutated Converter
[13] Hooshyar A, Azzouz MA, El-Saadany EF. Distance protection of lines connected to
HVDC Lines. IEEE Systems J 2021;15:1172–8.
induction generator-based wind farms during balanced faults. IEEE Trans Sustain
[31] IEEE Measuring relays and protection equipment – Synchrophasor for power
Energy 2014;5(4):1193–203.
systems – Measurements, IEC/IEEE 60255-118-1-2018; 2018.
[14] Hooshyar A, Azzouz MA, El-Saadany EF. Distance protection of lines emanating
[32] IEEE Guide for Application of Digital Line Current Differential Relays Using Digital
from full-scale converter-interfaced renewable energy power plants—part I:
Communication, IEEE Std C37.245; 2015.
problem statement. IEEE Trans Power Del 2015;30(4):1770–80.
[33] HVDC VSC transmission linking an (offshore) islanded wind farm with (onshore)
[15] Hooshyar A, Azzouz MA, El-Saadany EF. Distance protection of lines emanating
AC grid, 2020. [Online]. Available: https://www.pscad.com/knowledge-base/
from full-scale converter-interfaced renewable energy power plants—part II:
article/223.
solution description and evaluation. IEEE Trans Power Del 2015;30(4):1781–91.
[34] Davoudi M, Sadeh J, Kamyab E. Transient-Based Fault Location on Three-Terminal
[16] Dubey R, Samantaray SR, Panigrahi BK. Adaptive distance protection scheme for
and Tapped Transmission Lines Not Requiring Line Parameters. IEEE Trans Power
shunt-FACTS compensated line connecting wind farm. IET Gener Transm Distrib
Deliv 2018;33:179–88.
2016;10(1):247–56.

12

You might also like