This paper is a post-print of a paper submitted to and accepted for publication in IEEE Transaction on Power Electronics and is subject to

Institution of Electrical and Electronic Engineering Copyright. The copy of record is available at IEEE Xplore Digital Library

DC Fault Protection Algorithms of MMC HVDC

Grids: Fault Analysis, Methodologies, Experimental
Validations and Future Trends
Wang Xiang, Member, IEEE, Saizhao Yang, Grain Philip Adam, Member, IEEE, Haobo Zhang,
Wenping Zuo, Jinyu Wen, Member, IEEE
USA, China and India, the multi-terminal MMC HVDC grid is


Abstract- To protect the converters and minimize the power

transmission interruption during DC line faults, it is necessary to the most cost-effective solution. It allows massive integration
detect the faults at ultra-high-speed for the MMC based HVDC of renewable energy sources and flexible infeed of local AC
grids. This paper reviews the state-of-the-art of DC fault grids. In 2020, the world’s first MMC HVDC grid project
protection methods of MMC HVDC grids, and summarizes the (Zhangbei project) is commissioned in Beijing -Tianjin-Hebei
underlying principles of each method. On this basis, the DC fault area of China, which transmits 4500 MW wind and solar power
characteristics analysis in terms of modal-domain, time-domain
to the Beijing load center [4][5].
and frequency-domain are analyzed, which direct the protection
design. Typical boundary protection and non-boundary When the MMC HVDC grid is subject to short-circuit faults
protection schemes are reviewed. The advantages and on the transmission lines, all the sub-module (SM) capacitors
disadvantages of existing fault protection methods are compared. discharge rapidly, leading to a high fault current. Due to the low
A two-terminal MMC-HVDC prototype is developed to test the inertia of the DC networks, the DC fault propagation speed is
effectiveness of three fault protection methods. Comprehensive much faster than the AC fault [6]. To protect the converters,
quantitative assessments of the protection methods discussed double thyristor switches can be connected in parallel with the
above are carried out in a four-terminal MMC HVDC grid. half-bridge sub-modules to isolate the fault lines by turning on
Finally, the future trends of the protection schemes are discussed the firing signals [7]. However, the fault clearance time is
and the findings are concluded.
partially determined by the current decaying time. And the AC
Index Terms- DC grid, DC fault, fault protection, HVDC grid protection will be interrupted under some extreme cases.
transmission, modular multilevel converter, renewable energy. To interrupt the DC faults, various DC fault-tolerant MMC
topologies have been proposed, which can be classified into DC
I. INTRODUCTION fault blocking [8]-[12] and DC fault ride-through (without
With the increasing energy consumption and growing blocking) approaches [13]-[23]. For the blocking approach,
concerns on environmental degradation, renewable energy references [16]-[19] proposed various improved SM topologies
integration is receiving increased attention in decarbonizing the to block the DC fault current by switching off the power
power sectors and slowing down the pace of climate change. electronic devices. For the latter approach, references [20]-[22]
Many countries set their goals to produce more renewable utilized the cell topologies which can output negative voltage to
energy and reduce carbon emissions. In the United Kingdom, ride through DC faults and maintain continuous operation
half of the UK’s electricity will come from renewable sources during DC faults. However, these solutions are non-selective or
by 2025. It also became the first major economy in the world to partially selective [24], which will interrupt the power
pass laws to end its contribution to global warming by 2050. transmission of healthy parts. Thus, they are only suitable for
Due to the intermittent nature of renewable energy resources, point-to-point HVDC transmission.
they impose new technical challenges for integration into To facilitate fully selective DC fault protection, the high
power systems. With the merits of self-commutation, voltage DC circuit breakers (DCCBs) should be applied to both
decoupled active and reactive power control and black start ends of each protection zone to disconnect the faulty zone [25].
capabilities, the modular multilevel converter (MMC) based Thus, the impact of the fault condition is minimized and the
DC grid technology is widely recognized as a suitable approach healthy parts of the system continue transmitting power, which
for long-distance power transmission [1]-[3]. To utilize the is similar to the traditional AC protection strategies. Due to the
diversity of renewable energy resources across different fast fault current rise rate in HVDC grids, the DCCBs are
regions, such as the North Sea in Europe, the inland areas in required to be tripped in a range of time shorter than 10ms [26]
[27], leading to an ultra-fast requirement for DC fault
This work is sponsored by the Joint Funds of the National Natural Science protection algorithms. Taking the Zhangbei project as an
Foundation of China (U1766211). (Corresponding author: Saizhao Yang) example, the DC fault protection algorithms need to detect
W. Xiang is with the Department of Electronic and Electrical Engineering,
University of Strathclyde, Glasgow G1 1XW, U.K. (e-mail:
different faults within 2-3ms [28].
[email protected]) One of the major challenges in the practical operation of
G. P. Adam is with the NEOM, Energy Department. (e-mail: DCCBs is the dependence on ultra-high-speed protection
[email protected]) algorithms. Local measurement-based algorithms are
S. Yang, H. Zhang, W. Zuo, J. Wen are with the State Key Laboratory of
Advanced Electromagnetic Engineering and Technology, Huazhong University
considered to accelerate the detection speed, such as the
of Science and Technology, Wuhan 430074, China. (e-mail: conventionally overcurrent, low-voltage, the rate of change of
[email protected], [email protected], radio.zuo@ current or voltage algorithms. To guarantee high selectivity and
foxmail. com, jinyu.wen@ hust.edu.cn).
This paper is a post-print of a paper submitted to and accepted for publication in IEEE Transaction on Power Electronics and is subject to
Institution of Electrical and Electronic Engineering Copyright. The copy of record is available at IEEE Xplore Digital Library

reliability under different fault types, fault resistances and noise sub-modules based MMC technology. The DC circuit breakers
disturbance, many improved algorithms have been proposed. are installed on both ends of the transmission lines to protect the
Aimed at investigating the most competitive and promising lines. In the meshed HVDC grid, multiple converters will feed
fault protection methods, this paper carries out the review of the current to the fault point during DC short circuit faults. Thus,
state-of-the-art of DC fault protection algorithms for the MMC large current limiting inductances are implemented at the
based HVDC grids, with a focus on their functions, technical terminals of each line.
advantages and disadvantages. Id c1 Idc12 CB12 Ld c Line12 Ldc CB21 Idc21 Idc2
MMC1 MMC2
The outline of this paper is organized as shown in Fig. 1. In CB23
CB14
Section II, the technical requirements of the MMC HVDC Vdcbus1 Vline12 Vline21
Ld c Vdcbus2
Ldc
protection are introduced, so as to provide a basis for AC1
1500MW 1500MW
AC2

performance evaluation of DC fault protection schemes. F12 F2

F1 F32
Sections III-V carry out the fault characteristic analysis of Line41
F14 DCCB Line23

MMC HVDC grids in modal-domain, time-domain and Overhead

Line Ldc
Ld c
frequency-domain, providing the theoretical foundation for the CB41 CB32 Idc3
Id c4
design of protection methodologies. Section VI summarizes the MMC4 CB43 Ldc Line34 Ld c CB34
MMC3
advantages and disadvantages of existing fault protection FB4 Idc4 3 Idc34
Vdc3
Vdcbu4 Vlin e43 Vlin e34 Vdcbus3
algorithms. And a comprehensive comparison is carried out for AC4 AC3
the mentioned schemes. The performance of three typical 1500MW 1500MW

protection algorithms is validated by experimental tests and Fig. 2. The structure of a four-terminal MMC-based DC grid with DCCB.
evaluated in a four-terminal MMC HVDC grid in Section VII
For the meshed HVDC grid shown in Fig. 2, the protection
and VIII, respectively. Section IX discusses the future trends of
zones can be configured as shown in Fig. 3, namely, the AC
the HVDC grid protection schemes. Section X concludes the
yard protection zone, the converter protection zone and the DC
founding in this paper.
DC Fault Protection Algorithms of MMC HVDC Grids
yard protection zone [29][30]. The DC transmission lines are
exposed to air and the probability of faults is high. To prevent
Speediness Selectivity Sensitivity Reliability converters from overcurrent, a comprehensive DC line fault
Requirement Fault resistance Noise disturbance protection algorithm is required.
Trade-off Fault Type Lightning Strike Protection
provide a basis for performance evaluation of protection schemes zone

Features DC yard DC line protection DCCB protection DC bus protection

Modal domain Asymmetric fault
Fault analysis Time domain TW propagation characteristic Valve overcurrent Valve overvoltage
protection protection
Frequency domain High-frequency characteristic Valve
Valve cooling SM protection
provide theoretical foundation for design of protection schemes system protection

Current-limiting inductance Transformer

AC yard AC line protection protection AC bus protection
Protection
design Boundary effects: Suppress the fault current: Protection configuration
Discriminate internal and external faults allow large detection margin
Fig. 3. The protection zones of meshed MMC HVDC grids.

Classifications Challenges Solutions B. Technical Requirements of DC Fault Protection Algorithm

Time-domain TW
Vulnerable to fault resistance WT and MG are used to
detect TW wave-front
1) Speediness
based method
Vulnerable to noise
High sampling frequency
WT based multi-resolution When a short-circuit fault occurs on the DC transmission
algorithms and MF
Cannot detect closed-in faults Introduce pilot protection
line, multiple converters will feed fault current, leading to a
Signal processed
based method Threshold determination lacking Improve sampling frequency
high amplitude of fault current. To protect the semiconductors
theoretical analysis
Intelligent Protection
from overcurrent, the fault is desired to be detected and
Inductance voltage Cannot work under weak boundary
based method conditions Non-boundary Protection interrupted within several milliseconds. Taking the
four-terminal Zhangbei HVDC grid project as an example, the
 Hardware and control systems
fault protection time is less than 3ms while the total fault
Coordination 

Protection against different fault conditions
Protection at different stages
clearing time is less than 6ms [31].
Future Impacts of renewable FRT, frequency stability, the intermittence and As pointed out in [32][33], the speed of fault protection
Trends energy access fluctuation characteristics depends on used signals and the delay introduced by the
Adaptive threshold
determination
Respond to various changes and improve the
reliability under different fault conditions measurement unit. Thus, single-end measurement is
Fig. 1. The overall framework of this paper.
recommended to avoid communication delay. Also, the
algorithms involved in the data processing should not be too
II. SPECIFICITY OF FAULT PROTECTION SCHEME FOR MMC complicated, otherwise, it will increase the calculation burden
BASED HVDC GRID and delay of the hardware processor. A simple and feasible data
processing algorithm should be selected to reduce the
A. Topology and Protection Configuration of a Typical MMC computation burden and time cost.
HVDC Grid
2) Selectivity
Fig. 2 shows the topology of a typical four-terminal meshed Selectivity implies that the protection relay should detect the
HVDC grid, where each converter employs the half-bridge
This paper is a post-print of a paper submitted to and accepted for publication in IEEE Transaction on Power Electronics and is subject to
Institution of Electrical and Electronic Engineering Copyright. The copy of record is available at IEEE Xplore Digital Library

faults in its coverage zone of the HVDC system and do not are two aspects: dependability and security [40][41]. The
operate out of its coverage zone. Taking protection CB12 dependability indicates that the protection algorithm should
shown in Fig. 2 as an example, F12 is an internal fault for CB12 operate correctly under internal faults. While the security
while F1, F2, F14 and F32 are deemed as external faults for requires that the protection algorithm should not operate falsely
CB12. The selectivity of the DC line protection algorithm of under external faults and unfaulty operating conditions.
CB12 means that CB12 should be tripped only during internal The main factors affecting reliability are noise disturbance
DC line faults (F12) while do not respond to external faults (F1, and lightning strikes. For the noise disturbance, the
F2, F14 and F32). signal-noise-ratio (SNR) is conventionally used to describe the
To ensure selectivity, it is possible to increase the protection relationship between the original signal and the noise signal,
thresholds to avoid tripping during external faults. However, which is defined as follows:
this will reduce the endurance to high-impedance faults. The P
pilot protection schemes are robust to high-impedance faults. SNR  10lg s (db) (1)
Pnoise
but they need to receive information from the opposite station,
where Ps and Pnoise represent the original and noise signal
which increases the protection time delay. Thus, they are often power, respectively. For the lightning strike, the IEC standard
used as backup protection [34]. recommends that the wave-front and half-peak time of the
3) Sensitivity lightning current are 1.2µs/50µs for a standard lightning current
The sensitivity requires the protection devices to accurately waveform [42]. It will cause a large transient in voltage and
detect any internal faults, regardless of fault types and fault current, leading to the malfunction of protection.
resistances [35], e.g., the pole-to-pole (PTP) and the pole-to-
Both noise disturbance and lightning strikes are
ground (PTG) faults. Considering the coupling of the faulted high-frequency signals [42], which possess few low-frequency
and healthy poles, some methods have been proposed to components. Compared with transient voltage and current
discriminate the faulted poles, e.g., the integration of the
during internal and external faults, noise and lightning
change of pole voltages [31], the zero-mode component based interference have obvious differences in the frequency
methods [36], the morphological gradient based methods [37]. spectrum. Thus, the difference in frequency-domain is often
To improve the robustness to high-impedance faults, signal
adopted to design a protective algorithm for preventing noise
processing algorithms such as wavelet transform (WT) [31], and lightning strike interference. For example, the criteria
morphological gradient (MG) [38] can be used to detect the based on high-frequency spectrum differences are proposed in
arrival of fault traveling waves or extract high-frequency
[43][44] to discriminate the lightning disturbance. For the noise
components of voltage/ current for protection design. disturbance, wavelet transform with a multi-resolution
The sampling frequency is another factor that affects
algorithm [45] or morphological filter (MF) [46] is adopted to
sensitivity. The higher the sampling frequency, the richer the
weaken the impacts of noise. Considering that the average
fault data information. Thus, for most fault protection value of the noise is zero, the integral process is usually
algorithms, a higher sampling frequency results in higher
employed to mitigate the impacts of noise disturbance [47].
sensitivity. With the development of analog-to-digital In summary, the technical requirements for the fault
converters and digital signal processors, the sampling step of protection schemes of MMC HVDC grids are concluded in
the current fault protection system can reach 2µs [39].
Table 1 as well as the technical difficulties and objectives
4) Reliability [48][49].
Regarding the reliable operation of protection systems, there
Table 1. Summary of technical requirements of MMC HVDC grid protection.
Technical requirements Technical difficulties Overall objectives
Ultra-fast, e.g. 3ms in Zhangbei MMC Short time window, which is difficult
Speediness
HVDC meshed grid project. to ensure selectivity.
1. Investigate fast single-end protection schemes
Operate under all internal faults while do
Selectivity High sampling frequency. 2. Ensure the ultra-fast detection speed while
not operate under all external faults.
remaining selectivity to different fault conditions.
1. Distinguish different fault types, e.g., 3. Reduce the signal transmission and processing
1. Effective faulted pole selection.
Sensitivity pole-to-ground and pole-to-pole faults.
2. Endurance to high-resistance faults. delay and enhance the anti-interference ability.
2. Endurance to high-impedance faults. 4. Coordinate the main protection and backup
Reliable operation against noise protection to improve reliability.
Lightning and noise interference can
Reliability disturbance, lightning strike and other
easily cause malfunction of protection.
unfaulty operating conditions.
in [50] and will not be repeated in this paper. Whereas, for the
III. DC FAULT CHARACTERISTICS ANALYSIS IN
PTG faults, the interactions and unbalances between the
MODAL-DOMAIN
negative and positive poles impose more challenges on the
A. Equivalent Circuit in Modal-Domain calculation of fault currents [51][52].
Fault current calculation can be conducted to theoretically For the line jk, the simplified equivalent circuit of the
analyze the inherent fault characteristics, thereby providing a positive and negative overhead lines in s-domain can be
theoretical basis for some protection schemes, e.g., overcurrent obtained in Fig. 4. Where the ground admittance and line
protection. In addition, it can also be adopted to determine the capacitance of the overhead line are neglected [53]. Ldc is the
fault detection time, thereby avoiding overcurrent. The current limiting inductance. Rjk, Ljk and Mjk are the line
calculation of DC fault current under PTP fault is well studied resistance, self-inductance and mutual inductance of the line jk,
This paper is a post-print of a paper submitted to and accepted for publication in IEEE Transaction on Power Electronics and is subject to
Institution of Electrical and Electronic Engineering Copyright. The copy of record is available at IEEE Xplore Digital Library

respectively. Denote Vjk_p(s) (Vjk_n(s)) and Vkj_p(s) (Vkj_n(s)) as Ifp(n) and Vfp(n) are the positive (negative) pole current and the
the positive (negative) pole voltage at node j and k. And Ijk_p(s) positive (negative) PTG voltage at the fault point, respectively.
and Ijk_n(s) are the positive and negative line currents. The line-mode equivalent circuit and the zero-mode
According to Fig. 4, the pole voltages and currents yield: equivalent circuit can be obtained, as shown in in Fig. 6 (b) and
V jk _ p ( s )   s (2Ldc +L jk ) sM jk   I jk _ p ( s )  (c). Where Leq14_1(0) =(L14_1(0)+Leq4+2Ldc)//Leq1, Ceq14 = Ceq4//Ceq1
    and Req14 =(R14 +Req4)//Req1. If1(0) and Vf1(0) are the line
V jk _ n ( s )   sM jk s (2Ldc +L jk )   I jk _ n ( s ) 
(zero)-mode current and line (zero)-mode voltage at the fault
(2)
 R jk 0   I jk _ p ( s )  Vkj _ p ( s )  point, respectively. The fault boundary conditions under a
    positive PTG fault are Vfp=0, Ifn=0. Using phase-modal
 0 R jk   I jk _ n ( s )  Vkj _ n ( s ) 
transformation, it yields:
V f 1  V f 0  0
j sLd c sLjk Rjk sLdc k+  (4)
+
 I f 1  I f 0
Vjk _p Ijk _p (s) Vkj _p
Mik
Equation (4) indicates that the line-mode fault equivalent
circuit and the zero-mode equivalent circuit should be in series
Vjk _n Ijk _n (s) Vkj _n connection. Thus, a composite equivalent network in
_ _
modal-domain can be obtained, as shown in Fig. 6 (d).
Fig. 4. Equivalent circuit of overhead line jk in s-domain.
When a PTG fault occurs, it can be seen from equation (2) L14 R14 Ldc nL12 nR12 (1-n)L12 Ldc L23 R23
that there will be strong couplings between the pole voltages. Leq4 Ifp (1-n)R12
Leq2
Leq1 Leq3
Thus, to eliminate the couplings and reduce the analytical
Req4 M14 Req1 nM12 Vfp (1-n) Req3
complexity, phase-modal transformation is commonly adopted, M12 Req2 M32
Vfn
which can transform equation (2) into the following form [54]: Vdc 4 Ceq1 Vdc 1
Ifn
Vdc 2 Ceq2 Vdc 3
Ceq4 Ceq3
V jk _ 1 ( s)  1 1 -1 V jk _ p ( s) 
     MMC4 MMC1 MMC2 MMC3
V jk _ 0 ( s)  2 1 1  V jk _ n ( s)  (a) equivalent circuit of four-terminal DC grid under a positive PTG fault
 s (2Ldc +L jk _ 1 ) 0   I jk _ 1 ( s)  sLdc nsL12_1 nR12 (1-n)R12 (1-n)sL12_1sLdc
   (3)
 0 s(2Ldc +L jk _ 0 )   I jk _ 0 ( s)  1/sCeq14 Vdc 14 If1 Vdc 23 1/sCeq23
 R jk 0   I jk _ 1 ( s )  Vkj _ 1 ( s )  sLeq14_1 sLeq23_1
Vf1
   
 0 R jk   I jk _ 0 ( s )  Vkj _ 0 ( s )  Req14 Req23
where Vjk_1 (s) and Vjk_0 (s) are respectively the line-mode and
(b) line-mode equivalent circuit
zero-mode voltages of Vjk_p(s) and Vjk_n(s). Ljk_1=Ljk-Mjk,
sLdc nsL12_0 nR12 (1-n)R12 (1-n)sL12_0 sLdc
Ljk_0=Ljk+Mjk. Equation (3) demonstrates that the couplings
between the positive and negative poles can be eliminated by 1/sCeq14 If0 1/sCeq23
the phase-modal transformation. And the line-mode and sLeq14_0 sLeq23_0
Vf0
zero-mode equivalent circuits can be obtained:
Ld c Ld c Req14 Req23
j Ljk _1 Rjk k
Ijk _1 (c) zero-mode equivalent circuit
Vjk _1 Vkj _1 sLdc nsL12_1 nR12 (1-n)R12 (1-n)sL12_1sLdc
1/sCeq14 If1(0)
Vdc 14 Vdc 23 1/sCeq23
(a) line-mode equivalent circuit sLeq14_1 1/sCeq14 1/sCeq23 sLeq23_1
j Ldc Ljk _0 Rjk Ldc s(Leq14_0+2Ldc s[Leq23_0+2Ldc +
k
Req14 +2nL12_0) 2(1-n)L12_0 ] Req23
Ijk _0 (Req14+2nR12 ) [Req23+2(1-n)R12 ]
Vjk _0 Vkj _0
(d) composite-mode equivalent circuit
Fig. 6. Equivalent circuits in modal-domain during a PTG fault.
(b) zero-mode equivalent circuit
Fig. 5. Equivalent circuits of line jk in modal-domain. From Fig. 6 (d), the fault current If1(0) can be obtained:
Vdc1Z Σ2 +Vdc2 Z Σ1
B. Fault Current Calculation under PTG Faults I f 1(0 )  (5)
Z Σ1Z Σ3  Z Σ1Z Σ2  Z Σ2 Z Σ3
As shown in Fig. 2, supposing a solid positive PTG fault
happened at n% of line 12 (0≤ n% ≤1). Considering that MMC3 where Z∑1=Req14+2nR12+1/(sCeq14) +s(2Ldc+Leq14_1+2nL12_1),
and MMC4 are far away from the fault point, the fault current Z∑2= Req23+2(1-n) R12+1/(sCeq23) +s [2Ldc+Leq23_1+ 2(1-n)L12_1],
passing line 34 can be ignored [31]. Thus, the equivalent circuit Z∑4=Req14+2nR12+1/(sCeq14)+s(2Ldc+Leq14_0+2nL12_0),
of the four-terminal DC grid under a positive PTG fault can be Z∑5=Req23+2(1-n)R12+1/(sCeq23)+s[2Ldc+Leq23_0+2(1-n)L12_0],
obtained, as shown in Fig. 6(a). Where Leqi, Reqi, Ceqi Z∑3= Z∑4 ||Z∑5.
respectively represent the equivalent inductance, equivalent Considering that Vdc14(0-) =Vdc23(0-) =Vdcn (Vdcn is the rated
resistance and equivalent capacitance of MMC i (i=1, 2, 3, 4). DC voltage), the fault current in s-domain can be expressed as:
This paper is a post-print of a paper submitted to and accepted for publication in IEEE Transaction on Power Electronics and is subject to
Institution of Electrical and Electronic Engineering Copyright. The copy of record is available at IEEE Xplore Digital Library

800
2Vdcn ( Z Σ2 +Z Σ1 ) ZC1
I fp ( s)  I f 1  I f 0  (6) 700 ZC0
Z Σ1 Z Σ3  Z Σ1 Z Σ2  Z Σ2 Z Σ3

ZC (Ω)
600
Equation (6) shows that the fault current is highly dependent
500
on the fault distance. It is difficult to obtain a simplified
expression of fault current for protection design. Thus, DC 400
Fault characteristics analysis in other domains should be 300
1 1.5 2 2.5 3 3.5 4 4.5 5 5.5 6
studied. lg (f)
When a short-circuit fault occurs on the transmission line, (a) Line-mode and zero-mode wave impedance.
the backward voltage traveling wave (TW) induced from the 3
105

fault point will cause the rapid drop of DC line voltage and a 2.8
drastic increase of DC line current. The time-domain together

vp (km/s)
2.6
with the frequency- domain methods can be used to analyze the
2.4
fault propagation characteristics and frequency information of
2.2 vp1
TWs, thereby providing a theoretical basis for the DC grid vp0
protection design, which will be detailed in the following 2
1 1.5 2 2.5 3 3.5 4 4.5 5 5.5 6
sections. lg (f)
(b) Line-mode and zero-mode TW propagation velocity.
Fig. 8. Frequency-magnitude characteristics of wave impedance and
IV. DC FAULT CHARACTERISTICS ANALYSIS IN TIME-DOMAIN propagation velocity.
A. Time-Domain Travelling Wave Analysis Due to the attenuation effect of the transmission line, the
Once a short-circuit fault occurs on the DC line, a negative magnitude of the fault TW will decrease when the fault TW
voltage source is superimposed on the fault point and the propagates along the transmission line. The attenuation
voltage TW propagates towards both ends of the line. Fig. 7 coefficient k can be expressed as:
shows the distributed parameter model of one pole of overhead k  Re (r  j L)( g  jC ) (8)
lines, where r, L, g and C represent the resistance, inductance, Fig. 9 shows the attenuation factor |e | of the TW versus
-k x

conductance and capacitance of one unit segment of the different fault distances. From Fig. 9, it can be concluded that:
overhead line. The wave impedance ZC and propagation 1) As the fault distance increases, the TW attenuation degree
velocity vp can be expressed as: will be severe, and the corresponding cut-off frequency will
 Z r  j L decrease. 2) Under the same fault distance, the cut-off
 ZC  = frequency of line-mode TW is larger than that of the zero-mode
 Y g  jC
 (7) TW. 3) Under the same fault distance and frequency, the
v    attenuation degree of zero-mode TW is more serious.
=
 p Im ZY Im (r  j L)( g  jC ) 1

0.8
In equation (7), Z and Y are the impedance and admittance of 0.707
0.6
one segment of the overhead line, respectively.
0.4
e-k x

i(x+ dx,t)
1

rdx jw Ldx i(x, t) x=200km

x=350km
0.2 x=500km
x=1000km
0
gdx jw Cdx 1 1.5 2 2.5 3 3.5 4 4.5 5 5.5 6
lg(f)
v(x+ dx ,t) v(x, t) (a) line-mode attenuation factor.
1
x=200km
dx 0.8
x=350km
x=500km
Fig. 7. The distributed parameter model of the overhead line. 0.707 x=1000km
0.6
For the positive and negative poles, applying the
e -k x
0

0.4
phase-modal transform to decouple the couplings between the
0.2
poles. Using the overhead line parameters listed in [53], Fig. 8
depicts the frequency-magnitude characteristics of the 0
1 1.5 2 2.5 3 3.5 4 4.5 5 5.5 6
lg(f)
line-mode and zero-mode wave impedances (ZC1 and ZC0) and (b) zero-mode attenuation factor.
propagation velocity (vp1 and vp0), respectively. Fig. 9. Attenuation characteristics of the line-mode and zero-mode TWs.
It can be seen from Fig. 8 that, compared with zero-mode B. Characteristics of the Initial Traveling Wave
wave impedances ZC0 and propagation velocity vp0, the
line-mode wave impedances ZC1 and propagation velocity vp1 The length of the transmission line is denoted as l and the
are more stable, indicating that ZC1 and vp1 are less affected by fault distance is x. Assuming that the direction from the MMC
the variation of frequency. Thus, ZC1 and vp1 are often adopted to the transmission line is positive. The initial voltage TW and
for fault protection design. current TW from the fault point are backward, denoting them as
Vfb(x, t) and Ifb (x, t), respectively. Then, the measured backward
voltage and current TWs (Vb (x, t) and Ib (x, t)) at the terminal of
the line can be expressed as:
This paper is a post-print of a paper submitted to and accepted for publication in IEEE Transaction on Power Electronics and is subject to
Institution of Electrical and Electronic Engineering Copyright. The copy of record is available at IEEE Xplore Digital Library

Vb (0, t )  V fb ( x, t )e- ( x,t ) x resistance is small, αC≈0, indicating that the reflected TW at
 - ( x ,t ) x
(9) point A will not refract into the BC segment. Similarly, the
 I b (0, t )  I fb ( x, t )e reflected TW at point B will not refract into the AC segment, as
where e-γx is the propagation coefficient of the TW, which is the shown in Fig. 12.
main cause of TW distortion and dispersion [55]. Its expression For the measured voltage Vdc12 at point A, it contains multiple
in s-domain is [56]: reflected waves from the fault point. The reflected waves V2r , V
r , Vr , …, Vr from the fault point are:
4 6 2n
e kx
e  ( s ) x  e sx / v (10)  2 4 2
(1  sTa )  Vr   A CV fb，Vr  (  A C ) V fb
 6 3 2n n
(13)
where e-sx/v is the phase shift factor, representing the 
 Vr  (  A  C ) V fb ,... ， Vr  (  A  C ) V fb
propagation delay of the fault TW. Ta is the dispersion time Considering that the polarity of the initial TW Vfb is negative,
constant, which represents the distortion effect of the the polarities of the reflected waves V2r , V4r , V6r , …, V2nr from the
transmission line on the TW. The farther the fault distance x, fault point are respectively positive, negative, positive, ... When
the larger the dispersion constant Ta. the reflected waves V2r , V4r , V6r , …, V2nr arrive at point A, it will
Fig. 10 shows the waveform of the initial backward TW (Vb) cause the increase or decrease of the DC line voltage, as
detected at the terminal of the transmission line under different illustrated in Fig. 13. The time interval that causes the change of
fault distances. the voltage Vdc_line is ∆T=2x/vp.
Vb
t0 t0+x1/v t0+x2/v t0+x3/v t A x C B
0 V fb
Arr ival time
V r1
x3>x2>x1 V r2
V r3
V r4
V r5
Steady-state
Fig. 10. Measured backward TW at the terminal of the transmission line versus V r6
different fault distances. Fig. 12. Illustration of traveling wave reflection and refraction.
In Fig. 10, the arrival time t in x-axis represents the time
when the initial TW arrives at the terminal of the transmission decrease increase decrease increase
line. x1-x3 denote the different fault distances (x1< x2< x3). It can
be seen from Fig. 10 that, the farther the fault distance, the t
longer the propagation delay and the longer the time of TW Vb t0 t0+2x/vp t0+4x/vp t0+6x/vp t0+8x/vp
to reach steady-state. Fig. 13. Illustration of change tendency of measured voltage Vdc_line.
C. Refraction and Reflection of Traveling Wave Based on the aforementioned analysis, the main frequency of
As shown in Fig. 11, the fault TW induced from the fault the DC line voltage can be obtained as [57][58]:
point C propagates towards the terminals of the OHL and is 1 vp
reflected at the terminals (points A and B). f   (14)
T 4 x
OHL
MMC1 MMC2 D. Characteristics of the Backward Traveling Wave
Ldc A ZC ZC B Ldc Fig. 14 shows two DC faults happening at the meshed grid.
C
For the protection of line 12, F12 is an internal fault while F23 is
Vdc12 an external fault.
Rf
MMC1 MMC2
Ldc OHL12 Ldc
Fig. 11. Schematic diagram of traveling wave propagation.
Denote the first reflected wave at point A as Vr1 . It propagates ZC F12
towards the fault point and is reflected at point C, denoting as Vdc12 Vdc21
F23
Vr2 . The reflection coefficient βA at point A is:
Z Z OHL14 OHL23
A  T1 C  0 (11) Fig. 14. Illustration of internal and external faults.
ZT 1  Z C
Based on the phase-modal transformation, the initial
where ZT1 is the equivalent impedance of the terminal close to line-mode backward TW Vfb1 superimposed on the fault point is
MMC1. [59] :
The refraction coefficients αC and βC at point C are:
2V Z
 2Z C //R f Rf V fb1   dcn C1 (15)
 C   R f +ZC1
 Z C  Z C //R f R f  ZC 2
 (12) For the internal faults (e.g. F12), the initial line-mode
 Z C //R f  Z C backward TW (Vb1) measured at the terminal of line 12 is:

C  0
 Z C //R f +Z C
where Rf is the fault resistance. As can be seen, when the fault
This paper is a post-print of a paper submitted to and accepted for publication in IEEE Transaction on Power Electronics and is subject to
Institution of Electrical and Electronic Engineering Copyright. The copy of record is available at IEEE Xplore Digital Library

200
2Vdcn ZC1e kx e s / v
Vb1 ( F12 )  e ( s )lV fb1   (16) 150
( R f +ZC1 )(1  sTa )

Magnitude (kV)
For the external faults (e.g. F23), according to Peterson 100

equivalent circuit shown in Fig. 15, the initial line-mode 50

backward TW (Vb1) measured at the terminal of line 12 is: 0
 Vb1 ( F23 )  21 2 e  ( s )lV fb1 1 1.5 2 2.5 3 3.5 4 4.5 5 5.5 6 6.5 7 7.5 8 8.5 9 9.5 10

 f(kHz)
 s 2 Leq2 Ceq2  sCeq2 Req2  1 Fig. 17. Frequency spectrum of DC line voltage Vdc12 under an internal fault.
 1  As can be seen in Fig. 17, the main frequency in the transient
 2( s 2 Leq2 Ceq2  sCeq2 Req2  1)  sLdc  Z C1 (17) voltage is around 1400 Hz, which complies with the analytical

 Z C1 results in equation (14).
 2  Neglecting the line capacitance, the equivalent fault circuit
 sLdc2  Z C1
of the meshed DC grid is shown in Fig. 18 (a). Where Zjk_1 is the
where Γ1 is the high-frequency oscillation term and Γ2 is the line-mode impedance of line jk. Zjk_1= jωLjk_1+Rjk.
exponential decay term. Γ2 exhibits that the sharp TW Since MMC3 is far away from the fault point, the
wave-front will be smoothed by the current limiting inductor. short-circuit current fed through line 34 is negligible [31].
Supposing PTP faults occur at the opposite terminal of line Therefore, the equivalent circuit can be simplified as shown in
12 (close to MMC2) and the terminal of line 23 (close to Fig. 18 (b). The ratio between the DC line voltage Vdc12 and the
MMC2) at 2s, respectively. Using the parameters listed in DC bus voltage Vdc1 at MMC1 side in frequency-domain yields,
Table 4 and Table 5. the initial line-mode backward TW (Vb1)
measured at the terminal of line 12 is shown in Fig. 16.
As can be seen from Fig. 16, due to the smoothing effect Req1 Ceq 1 Leq 1 Ld c 0.25Z12_1 0.75Z12_1 Ld c Leq2 Ceq2 Req2
provided by the current-limiting inductor, the wave-front of the
backward TW is flattered under external faults. Thus, the rate Vfb1
Vd c1 Vdc12
of change of the amplitude of the backward TW can be used to
distinguish internal and external faults. Z14_1 Z2 3_ 1

+ sLdc sLdc Req 4 Ceq 4 Leq4 Leq3 Ceq3 Req 3

sLeq2 ZC1
ZC1 +
1/sCeq2 2e   ( s ) l V fb1 _ Z34_1
_
12

Req2 (a) equivalent fault circuit under an internal PTP fault

Req1 Ceq 1 Leq 1 Ld c 0.25Z12_1 0.75Z12_1 Ld c Leq2 Ceq2 Req2
Fig. 15. Peterson equivalent circuit under an external fault F23.
100
Backward TW (kV)

0 Vd c1 Vdc12 Vfb1
-100
F23 Z14_1 Z2 3_ 1
-200
F12
-300
-400 Leq3 Ceq3 Req 3
Req 4 Ceq 4 Leq4
-500
2 2.0005 2.001 2.0015 2.002
Time(s)
Fig. 16. Line-mode of initial backward voltage TW under an internal fault F12
and an external fault F23. (b) simplified circuit under an internal PTP fault
Fig. 18. Equivalent fault circuit of the HVDC grid under an internal PTP fault.
V. DC FAULT CHARACTERISTICS ANALYSIS IN FREQUENCY-
DOMAIN Vdc12 2 j Ldc  Z eq1 //Z 
H ( j )   (18)
A. Frequency Characteristics of Transient Voltage Vdc1 Z eq1 //Z 
As analyzed in Section IV.C, due to multiple refractions, the Z∑ and ZMMC1 are:
transient voltage possesses plentiful high-frequency 1

components. As shown in Fig. 2, when a metallic PTP fault  Z eq1( 4 )  j ( Leq1( 4 )  )  Req1( 4 )
occurs at 25% of line 12, the FFT is performed on voltage Vdc12. Ceq1( 4 )

The frequency spectrum of voltage Vdc12 is shown in Fig. 17.  (19)
 Z   Z eq4  4j Ldc  2Z L14 _ 1
 2L 2R 6C
 Leq1( 4 )  arm1( 4 ) ; Req1( 4 )  arm1( 4 ) ; Ceq1( 4 )  arm1( 4 )
 3 3 N
where Larmi, Rarmi and Carmi are the arm inductance, resistance
and capacitance of MMC i (i=1,2,34) respectively. N is the
number of sub-module per arm. According to equation (18), the
frequency-amplitude characteristic of |H(jω)| can be obtained
This paper is a post-print of a paper submitted to and accepted for publication in IEEE Transaction on Power Electronics and is subject to
Institution of Electrical and Electronic Engineering Copyright. The copy of record is available at IEEE Xplore Digital Library

in Fig. 19. currents of MMC1 versus different current limiting inductors

5 are shown in Fig. 20. The rated arm current and submodule
4
3
voltage of MMC 1 are 2.2kA and 4kV respectively. Thus, the
2 4.5kV/3kA press-pack IGBT developed by ABB is selected
lg H(jω)

1 [60], of which the maximum allowable arm current is 6kA.

0 9
L=100mH
-1 7.5 L=150mH
L=200mH

Iarm_max (kA)
-2 6
0 1 2 3 4 5 6
lg f 4.5
Fig. 19. Frequency-amplitude characteristics of |H(jω) |. 3
Fig. 19 shows that, due to the attenuation effect provided by 1.5
the current-limiting inductor, the DC line voltage (Vdc12) 0
0 0.5 1 1.5 2 2.5 3 3.3 3.5 4 4.5 4.7 5 5.5 6
possesses more high-frequency components (more 10kHz) Time(ms)
compared to the DC bus voltage (Vdc1). Thus, the difference of Fig. 20. Maximum arm currents of MMC1 versus different fault current
high-frequency components in transient voltages can be used as limiting inductors.
an indicator to identify the internal and external faults. Fig. 20 shows that a larger current limiting inductor has a
stronger current limiting effect. As the fault distance increases,
B. Analysis of Current Frequency Characteristics the TW propagation delay increases and the rise of arm current
Based on the aforementioned analysis in Section III, when a slows down. Supposing the operation time of DCCB is 3ms, the
metallic PTP fault occurs at n% of line 12 the fault equivalent allowable fault detection time td can be obtained:
network is depicted in Fig. 6 (b). The oscillation frequencies of td ( Ldc  100mH)  (3.3-3)ms  x / v p (22)
currents i12 and i21 can be obtained:
where x/vp represents the TW propagation delay. Similarly,
 2nR12  Req14 2
 ω  1
( )
when the current-limiting inductors are 150mH and 200mH, the
1

 (Leq14_1  2 Ldc +2nL12_1 )Ceq14 Leq14_1  2 Ldc +2nL12_1
(20) allowable fault detection times td are respectively:

 2(n-1)R12  Req23 2  x
td ( Ldc  150mH)  (4.7-3)ms  v
1
ω2  [ ]
 [Leq23_1  2 Ldc +2(n-1)L12_1 ]Ceq23 Leq23_1  2 Ldc +2(n-1)L12_1  p
 (23)
Substituting the parameters listed in Table 4 and Table 5 into  t ( L  200mH)  (6-3)ms  x
(20), the oscillation frequencies of the current can be obtained,  d dc vp

which are within hundreds of hertz.
As can be seen from equations (22) and (23), a larger
The differential current expression of protection CB12 is:
current-limiting inductance can suppress the fault current rise
id  idc12  idc21 (21) rate better, thereby allowing more detection time for fault
When internal faults occur, id is equal to the fault current. Its identification and reducing the breaking capacity of DCCBs.
frequency is the superposition of oscillation frequencies ω1 and However, the size of the current-limiting inductance is
ω2. When external faults occur, the differential current id is constrained by its cost CL and weight ML, which satisfy [61][62]
equal to the current flowing through the line-to-ground CL , M L  Ldc (24)
capacitor. The line-to-ground capacitor current is proportional
Taking the 5mH hollow inductor as an example, its weight is
to the derivative of the line voltage. Thus, the frequency of the
about 2.5 tons [63]. Therefore, there is a tradeoff between fault
line-to-ground capacitor current is determined by the frequency
current limiting effect and manufacturing difficulty for the
of the voltage TW [58], which is above 10 kHz. Therefore, the
selection of current-limiting inductors.
high-frequency components of the differential current can also
be used to identify the internal and external faults. B. Analysis of Boundary Characteristics
As a boundary element, the current-limiting inductors
VI. DC LINE PROTECTION SCHEMES FOR MMC BASED HVDC smooth the wave-head of traveling waves and attenuate the
GRIDS high-frequency components. Therefore, the amplitudes of the
From the above fault characteristic analysis, it can be seen change rate of DC line voltage and the high-frequency
that the current limiting inductors act as a boundary element for components in transient voltage are much larger under internal
fault protection. Therefore, most existing single-end protection faults than those under external faults. References [64] and [31]
schemes rely on the current-limiting inductors to design the respectively proposed the protection algorithms based on the
fault protection criterion. Thus, in the following section, the rate of change of DC line voltage (ROCOV) and the transient
impacts of current limiting inductors on fault protection energy of DC line voltage (EWT).
algorithms are analyzed firstly. Then, the DC line protection As shown in Fig. 2, taking CB12 as an example, the bus fault
algorithms depending on and independent of the current limit F1 is a backward external fault while F2 is a forward external
inductors will be reviewed. fault. The backward external fault F1 can be sorted out by the
A. Analysis of the Impact of Current Limiting Inductor on directional element easily. Thus, the most severe external fault
Fault Protection is the solid PTP fault at DC bus 2 (F2). Considering an internal
To test the current limiting effect of different sizes of current fault with 300 Ω resistance happened at the end of line 12 (close
limiting inductors, a solid PTP DC fault is applied at the to MMC2, F12), the sensitivity coefficients ksen1 and ksen2 against
terminal of line 12 (close to MMC1). The maximum arm different current-limiting inductors are compared, as shown in
This paper is a post-print of a paper submitted to and accepted for publication in IEEE Transaction on Power Electronics and is subject to
Institution of Electrical and Electronic Engineering Copyright. The copy of record is available at IEEE Xplore Digital Library

Fig. 21. Where the sensitivity coefficients ksen1 and ksen2 are backward voltage TW exceeds the threshold, it is deemed to be
defined as: an internal fault. These methods have the merits of fast
 ROCOV ( F12 ) detection speed and no dead protection zone. However, these
k sen1  ROCOV ( F ) methods suffer from poor performances under high resistance
 2
faults and noise disturbance.
 (25)
 k EWT ( F12 ) To improve reliability, reference [72] proposed a TW

 sne2 EWT ( F2 ) protection scheme with adaptive protective thresholds. The
where ROCOV (F2) and ROCOV (F12) respectively represent propagation characteristics of the initial zero-mode TW is
the measured ROCOV under the external fault F2 and the analyzed to estimate the fault distance and the fault resistance,
internal fault F12. EWT (F2) and EWT (F12) respectively represent thereby adaptively adjusting the threshold to improve the
the measured transient energy under faults F2 and F12. endurance to fault resistances. In [73], the first peak time of the
From Fig. 21, it can be concluded that, line-mode voltage TW is adopted to identify the internal and
(1) With the increase of current-limiting inductances, the external faults. In [74], considering that distortion of the fault
sensitivity coefficient increases. When the current-limiting TW wave-front depends on the fault distance, a protection
inductor is selected as 50mH, both ksen1 and ksen2 are less than 1, scheme using the zero-mode current waveform correlation is
indicating that the external and internal fault cannot be adopted to detect internal and external faults. References
correctly detected. Therefore, protection schemes based on [75][76] calculated the parameters that can present the fault
boundary elements rely on large current-limiting inductors to distance by fitting the zero-mode current TW, so as to achieve
ensure selectivity. the identification of internal faults. In [77], the time interval
(2) Under the same current-limiting inductance, ksen1 is between the sudden change point and the first peak point of the
smaller than ksen2, indicating that the protection algorithm based initial TW is adopted to identify internal faults. These
on transient energy is more reliable than that based on ROCOV. aforementioned schemes improved the performance of
2.2 16 endurance to large fault resistance (300Ω). However, these
2 ksen1 14 schemes are easily affected by noise disturbance and high
1.8 ksen2 12 sampling frequencies (100kHz or 200kHz) are required. Due to
1.6 10
the absence of zero-mode components, the methods in [72][74]
Ksen2

1.4 8
Ksen1

1.2 6 -[76] cannot identify internal PTP faults. In addition, these

1 4
0.8 2 protection schemes cannot work well under close-in faults. To
0.6 0 detect close-in faults, distributed sensors are adopted in [78] to
50 100 150 200 250
Ldc(mH) protect the entire transmission line.
Fig. 21. Sensitivity coefficients versus different current limiting inductors. Based on the above literature review, for the TW protection
C. Single-end Protection Schemes Relying on Boundary schemes based on the voltage/current derivative, a smaller
Elements sampling frequency is required and there is no dead protection
zone. But they are vulnerable to high fault resistances and noise
Regarding the classification of single-end boundary
disturbance. The protection schemes based on waveform
protection, it can be divided into two categories, i.e., the
correlation calculation or TW propagation characteristic
phase-domain and the frequency-domain based methods) [65]
analysis can improve the performance under high-impedance
or three categories based on the fault protection principle. Since
faults, but higher sampling frequency is required and they are
phase-domain and modal-domain based methods have been
unable to detect close-in faults.
investigated and reviewed in [65], the latter classification is
2) Frequency- domain signal processing based protection
adopted to evaluate different fault protection schemes.
schemes
Considering that most fault protection algorithms are
Wavelet transform (WT) has the merits of time-frequency
applicable to both cable and overhead line based systems, the
localization and multi-scale decomposition ability [79], which
following reviewed publications contain the protection
has been widely used to extract high-frequency components of
schemes for both cable and overhead line applications.
signals or detect the wave-front of the initial TW. References
1) Time-domain TW based protection scheme
[31][80][81] used WT to extract the high-frequency
Traveling wave protection is widely used in traditional AC
components of the DC line voltage, so as to take advantage of
power systems and LCC-HVDC systems, which has a simple
the boundary effect provided by the current-limiting inductor to
principle and can be easily implemented. For instance, Siemens
discriminate internal and external faults. While references
proposed the TW protection scheme based on the DC voltage
[82][83] used the wavelet transform modulus maxima (WTMM)
derivative (dV/dt), the change of DC line voltage (∆V) and the
to detect the arrival of the voltage TWs in the high-frequency
DC current derivative (dI/dt) [66]. For the MMC HVDC
spectrum. These protection schemes can tolerate
systems, references [67]-[69] proposed ultra-fast TW
high-impedance faults and protect the whole line. But they are
protection schemes based on the rate of change of DC voltage
easily affected by noise disturbance.
(ROCOV) and the rate of change of DC current (ROCOC)
To improve the endurance to noise disturbance, the WTMM
respectively. A comprehensive protection scheme that contains
with multi-resolution algorithms are adopted to detect the
the low-voltage criterion, the DC voltage derivative criterion
line-mode TW in [38][84]. However, the multi-scale
and the directional overcurrent criterion is proposed in [70]. In
decomposition of WT requires the sampling frequency as high
[71], the line-mode backward voltage TW is adopted to identify
as 1MHz, which is difficult to be implemented in practical
the internal DC faults. When the amplitude of the line-mode
This paper is a post-print of a paper submitted to and accepted for publication in IEEE Transaction on Power Electronics and is subject to
Institution of Electrical and Electronic Engineering Copyright. The copy of record is available at IEEE Xplore Digital Library

applications. Reference [85] adopted the morphological (2) Fault tolerable converter based HVDC transmission
gradient (MG) to detect the initial TW wave-front and the systems. The fault tolerable converters, e.g., the full-bridge
morphological filtering (MF) to resist noise disturbance. But MMC, the hybrid MMC, can suppress the fault current via fault
MG also brings about a high computation burden and sampling current limit control [90]. The Kunliulong three-terminal
frequency. HVDC project constructed by hybrid MMC will use a current
To be concluded, the protection schemes based on WT and limiting reactor of 100mH for fault current limiting [91].
MG are robust to large fault resistances and fault types. In (3) Medium/low voltage and small capacity transmission
addition, WT based on multi-resolution algorithms and MF can systems. In MVDC distribution systems, the current limiting
be adopted to improve the endurance to noise disturbance. inductors are conventionally in the range of 20mH [92].
However, they also suffer from the problems of the high To realize reliable fault protection under weak boundary
sampling frequency, computation burden and difficulty in conditions, single-end protection schemes independent of
selecting the mother wavelet. boundary conditions need to be proposed. Reference [93] used
3) Current-limiting inductor voltage based protection the arrival time difference between line-mode and zero-mode
schemes TW to realize non-boundary protection. This scheme is robust
In addition to the boundary element of the current-limiting to high-resistance faults. However, it cannot identify remote
inductor, the voltage across the inductor can also be used as faults and PTP faults. To improve the applicability for different
indicators for fault protection [86]. Reference [87] proposed a fault types, TW frequency characteristics which are dependent
protection scheme based on the rate of change of the on the fault distance are adopted to identify the internal faults in
current-limiting inductor voltage. The change-rate of inductor [94]. The proposed method is affected by the noise and a high
voltage is the second derivative of the line current. Compared sampling frequency is required. These non-boundary protection
with the protection scheme based on the change rate of current, schemes cannot work well for remote faults. Reference [36]
this method is robust to large fault resistances. However, this introduced the distortion effects of the line-mode backward
scheme cannot identify PTG faults. In order to improve the TWs to identify faults independent of boundary components.
applicability for different fault types, reference [88] used the Except for [36][93][94], there are no other references
inductor voltage differences between negative and positive concerning single-end protection schemes that independent of
poles to discriminate faulted poles. Nevertheless, similar to the boundary elements for the MMC HVDC systems.
current change-rate based schemes, this scheme also fails to E. Intelligent Algorithm Based Protection Schemes
detect high-impedance faults. In [53], based on the modal With the development of artificial intelligence, intelligent
domain analysis, the line-mode inductor voltage is adopted for algorithms are gradually applied in HVDC protection. The
fault identification while the zero-mode inductor voltage is protection schemes based on intelligent algorithms avoid the
adopted for faulted pole selection. The maximum detectable complex threshold determination process and have good
resistance is 200Ω, which needs to be further improved. performances on the speediness and selectivity.
Besides, the above protection schemes are robust to noise Reference [95] used four neural networks to detect the faults,
disturbance. whose training workload is large and complicated. Reference
In summary, compared with time-domain TW schemes, the [96] realizes the faulted types selection by comparing the
inductor voltage based methods improve the robustness to noise differences in the frequency spectrum. Compared with [95],
disturbance, and there are no dead protection zones. Compared this scheme used three neural networks to sequentially
with the protection schemes based on WT and MG, the inductor implement the functions of fault identification, fault
voltage based methods have the advantages of low computation classification and fault location. The workload is reduced, but it
burden and low sampling frequency. However, the is still very complicated. Reference [97] used the wavelet
performances under high-impedance faults need to be transform and fuzzy C-means algorithm to extract fault features.
improved. But this method did not consider the impact of fault resistances
D. Protection Schemes Independent of Boundary Elements and noise disturbance. Reference [98] proposed a protection
scheme based on the fuzzy control algorithm, which used a
As pointed out in [31], for boundary protection schemes,
large CLRs (over 100mH) are adopted to enable high multi-layer wavelet transform to extract fault features and
selectivity and reliability. In addition, various simulation implemented the fault recognition function through training
weights. However, the maximum detectable fault resistance is
results also demonstrate that 200mH or larger CLRs are
required to guarantee the robustness to large fault resistance only 50Ω. Reference [99] proposed a machine learning-based
and noise disturbance. However, there are several scenarios protection scheme. But this method needed to install a large
where the configured current-limiting inductors are small and number of current sensors along the line, which will increase
cannot provide strong boundary conditions for fault protection the investment cost.
Compared with the non-intelligent protection algorithm,
algorithms.
(1) DC cable-based HVDC transmission system. The intelligent protection methods avoid the complicated tuning
cable-based transmission system has a low probability of faults and calculation process. However, it is still at the preliminary
stage of research. The application in practical engineering
and most faults are permanent. Once a fault occurs, the
converters will be blocked and there is no need to implement a needs to be investigated.
large current limiting inductor for fault current limiting. Taking In summary, from the above literature review, the
performances of typical boundary and intelligent protection
the INELFE MMC-HVDC project as an example, the current
limiting inductor of the system is only 50mH [89]. schemes for MMC HVDC grids have been concluded in Table
This paper is a post-print of a paper submitted to and accepted for publication in IEEE Transaction on Power Electronics and is subject to
Institution of Electrical and Electronic Engineering Copyright. The copy of record is available at IEEE Xplore Digital Library

2. Where ‘√’ indicates that the method can detect both PTG and wave-front to improve the robustness to fault resistance. More
PTP fault with tolerating fault resistance larger than 300Ω, at importantly, a higher sampling frequency (100kHz or 200kHz)
least 20db noise disturbance. ‘×’ indicates that the method is recommended. For the close-in faults, the pilot protection can
cannot meet the technical requirements. ‘/’ indicates that the be adopted as back-up protection [100]-[104]. Regarding the
technical requirements are not discussed in the literature. robustness to noise disturbance, the WT based on
As can be seen in Table 2, the existing boundary protection multi-resolution algorithms and MF can be adopted to improve
schemes suffer from some of the following problems: 1) the endurance to noise disturbance. As for different fault types,
vulnerable to fault resistances; 2) easily affected by noise it is recommended that the line-mode components are adopted
disturbance; 3) cannot identify different fault types; 4) cannot for fault identification while the zero-mode components are
work well for close-in faults. To overcome the aforementioned adopted for faulted pole selection.
shortcomings, WT and MG can be adopted to detect the TW
Table 2 The Performance evaluation of typical protection methods under different protection classifications
TW based protection scheme in time-domain Intelligent algorithm based protection schemes
Performance
[64][67][68][69] [70] [72] [74][75][76] [77] [78] [99] [98] [95] [96] [97]
Fault resistance (Ω) Weak √ 50 50 69 10
Fault types √ × √ √
Noise endurance (dB) × weak × √
Sampling frequency
96 in [68] 100 200 100 100 5 /
(kHz)
Advantages fast speed, simple principle / /
Need supplementary More Applicable only Difficult to
Need many Need multiple
Other disadvantages / protection to detect close-in sensors in two-terminal extract fault
sensors ANNs
faults employed systems features
Operability Medium Low Low
Signal processing based protection schemes Current-limiting inductor voltage based protection schemes
Performance
[80] [81] [82][83] [31] [38] [84] [85] [87] [88] [53]
Fault resistance (Ω) √ 200 √ Weak 200
Fault types √ × √
Noise endurance(dB) Weak √ √
Sampling frequency
10 25 20 20 1000 20 200 5 20
(kHz)
Advantages / fast speed, simple principle
Other disadvantages Some need high sampling frequency and computation burden More voltage sensors
Operability Low Low High

VII. EXPERIMENTAL VALIDATIONS represents the internal fault and Rf is the fault resistance. F2
As mentioned above, the solid bus fault at the opposite represents the DC bus fault on the MMC2 side.
terminal is regarded as the most severe external fault. Thus, a Three typical boundary protection schemes as presented in
prototype of a two-terminal MMC HVDC system can be Section VI.C will be studied, namely, the time-domain TW
adopted to verify the selectivity of the DC line protection based protection as reported in [64] (denoted as ROCOV
schemes, as shown in Fig. 22. protection), the signal processing based protection using MG as
Line current reported in [38] (denoted as MGV protection) and the inductor
measurement voltage based protection as reported in [88] (denoted as VL
protection).
Considering that the fault impedance is usually small in low
Oscilloscope voltage DC systems, a 10Ω fault impedance is applied at the
MMC2 MMC1 end of the transmission line (F12) to test the selectivity of these
protection schemes under internal faults. The measured DC line
voltage Vdc12, the DC line current Idc12, and the voltage across
Fault the current-limiting reactor VL are shown in Fig. 24.
impedance As can be seen in Fig. 24, under the internal fault, the DC line
Line voltage Vdc12 drops rapidly from the rated voltage (400V).
impedance Meanwhile, the DC line current Idc12 rises rapidly, leading to the
CLR voltage fast increase of the current-limiting reactor voltage VL.
measurement
+ VL _ Idc12

Line voltage ~ Vdc12 Rf F2

Current limiting F12
Switch S1 measurement reactor (CLR) AC S1 S1
Fig. 22. The prototype of a two-terminal MMC-HVDC system. MMC1 Z12 MMC2
The electrical circuit and the parameters of the test system Fig. 23. The electrical circuit of the test system.
are shown in Fig. 23 and Table 3, respectively. In Fig. 23, F12
This paper is a post-print of a paper submitted to and accepted for publication in IEEE Transaction on Power Electronics and is subject to
Institution of Electrical and Electronic Engineering Copyright. The copy of record is available at IEEE Xplore Digital Library

Table 3. Parameters of the test system PROTECTION SCHEMES

Parameter Value
The rated power 2500VA
To further evaluate the performance of the three fault
AC voltage 380V protection schemes against different fault types, fault
The ratio of the AC transformer 380/220 resistances and noise disturbance, the four-terminal MMC
DC voltage 400V HVDC grid shown in Fig. 2 is built in PSCAD/EMTDC. In the
Arm inductance of MMC 10mH HVDC grid, the MMC1 controls the DC voltage of the DC grid,
Sub-module capacitance of MMC 6600μF
while the other converters control the active and reactive power.
Sub-module number per arm 20
Current-limiting reactor (CLR) 30mH Assuming that the power flowing into the DC grid is the
Line impedance Z12 0.5Ω positive direction, the active power transmitted by
Fault impedance Rf 10Ω MMC2~MMC4 is -1.0 pu, -1.0 pu and 1.0 pu respectively. The
current-limiting inductor installed on the transmission line is
0.1H and the sampling frequency is 25kHz. The lengths of the
lines OHL12, OHL14, OHL32 and OHL43 are 240km, 120km,
100km and 200km respectively. The other system parameters
are listed in Table 4 and Table 5.
Table 4. Parameters of the test system
DC voltage/kV AC voltage/kV power/MW
MMC1 ±400 380 1500
MMC2 ±400 380 1500
MMC3 ±400 380 1500
MMC4 ±400 380 1500
Table 5. Parameters of each converter
Larm/mH Carm/mF Vsm/kV N
MMC1 40 7.5 4 200
Fig. 24. The measured DC line voltage Vdc12, the DC line current Idc12, and the MMC2 40 7.5 4 200
current-limiting reactor voltage VL under an internal fault F12. MMC3 40 7.5 4 200
MMC4 4 7.5 4 200
3000
F12
Normal state
2000 ROCOV(kV/ms) F2
1000 No
Vdc<0.5Vdcn? Fault start-up
0 Yes Faulted pole
-1000 selection
ΔVdcp-ΔVdcn>Th ? ΔVdcp-ΔVdcn>Th ? -Th <ΔVdcp-ΔVdcn<Th ?
600
Yes Yes Yes
400 No
MGV(kV) MGV>setp? MGV>setn? MGV>setd?
200 No No
Faulted line
identification
0 Yes Yes Yes
P-PTG N-PTG PTP
-200
600 The DCCBs on the line are tripped
400 Fig. 26. Flow chart of the protection scheme of MGV.
VL(kV)
200 To make a fair comparison, the differences of the DC voltage
0 change rate between the positive and negative poles (∆Vdcp
-200 -∆Vdcn) are adopted to select the faulted pole for the three
0 0.5 1 1.5 2 2.5 protection schemes. To improve the sensitivity under different
Time(ms)
fault types, the internal fault identification criteria are
Fig. 25. The comparison results between internal and external faults.
conducted after the faulted pole selection. Taking the MGV
To test the effectiveness of these methods, an external solid
protection [38] as an example, the flow chart of the protection
fault F2 is applied at the DC bus of MMC2 to make a
scheme is presented in Fig. 26. It contains the fault start-up
comparison between the internal and external faults. The
element, the faulted pole selection criterion and the internal
comparison results are shown in Fig. 25.
fault detection criterion.
As can be seen, due to the boundary effects provided by the
CLR, the DC line voltage under the internal fault drops rapidly. A. The Threshold Determination
Thus, the ROCOV and MGV under internal faults are greater For the fault start-up element, its purpose is to distinguish the
than those under external faults. In addition, the fault current normal and fault states. Thus, the designed threshold can leave
rises faster. Thus, the VL under internal faults is greater than that a relatively large margin, e.g. 0.5Vdcn in Fig. 26. The faulted
under external faults. To be concluded, experiment results show pole selection criteria are the same for the three aforementioned
that the aforementioned protection schemes can identify the protection schemes. For the faulted pole selection, the threshold
internal faults and external faults selectively and quickly. determination can refer to [31]. For the internal fault
identification, taking relay CB12 as an example, the threshold
VIII. SIMULATION COMPARISON OF TYPICAL DC FAULT determination process is presented as follows.
This paper is a post-print of a paper submitted to and accepted for publication in IEEE Transaction on Power Electronics and is subject to
Institution of Electrical and Electronic Engineering Copyright. The copy of record is available at IEEE Xplore Digital Library

According to the analysis in Section VI.B, the solid bus fault ROCOV 200Ω≤Rf_max≤300Ω
F2 is the most severe external fault. Considering the impact of MGV Rf_max≥300Ω
VL Rf_max≤200Ω
noise disturbance, the solid bus fault F2 with 20dB noise is
regarded as the worst external case to calculate the thresholds. Table 7 demonstrates that the MGV based method is robust to
Applying a solid PTP fault at bus 2 and adding 20dB noise fault resistance while the VL based method is easily affected by
into the measured signals, the values ROCOV (F2, 0.01Ω, the fault resistance.
20dB), MGV (F2, 0.01Ω, 20dB), VL (F2, 0.01Ω, 20dB) are C. Response to Change of Operating Conditions and AC
1085kV/ms, 365 kV and 258 kV respectively. Faults
When the reliability coefficient is selected as 1.3 [105], the The operation mode of the MMC HVDC grids is flexible, so
thresholds can be calculated as 1410kV/ms, 478kV and 335kV it is necessary to investigate the impact of power flow change
respectively, as listed in Table 6. on the reliability of protection schemes. In addition, the
Table 6. Threshold determination of the three protection schemes robustness to AC faults should also be evaluated. Table 8 lists
Protection
worst external reliability measured
thresholds the identification results against power reversal and AC faults
case coefficient values for the three schemes.
ROCOV 1085kV/ms Set1: 1410kV/ms The power reversal does not change the magnitude and
F2, 0.01Ω,
MGV 1.3 365 kV Set2: 478kV
VL
20dB
258 kV Set3: 335kV
polarity of the DC line voltage. The DC voltage has a small
fluctuation, which will not trigger the fault start-up element.
B. Response to Fault Resistances Thus, it can be seen from Table 8 that the three protection
Applying internal PTP faults with different fault resistances schemes will not be affected by power reversal.
(200Ω and 300Ω) at F12 and a solid external PTP fault at F2, the In the case of some severe AC faults, though the amplitude of
values of ROCOV, MGV and VL are measured, as shown in Fig. the DC voltage is smaller than 0.5Vdcn, the DC voltage drops
27, where “x%” means the internal fault F12 is applied at the x% slowly. Thus, the methods based on ROCOV and MGV will not
of the OHL 12. operate falsely. Meanwhile, the DC line current rises slowly.
As can be seen in Fig. 27, with the increase of fault resistance, Thus, the VL based method will not operate falsely, either. To be
the amplitudes of the measured ROCOV, MGV and VL decrease. concluded, these methods are robust to AC faults.
When the fault resistance exceeds 200Ω, the VL based Table 8. Identification results against power reversal and AC faults.
protection cannot identify the internal and external faults Protection schemes Power reversal AC faults
correctly. When the fault resistance increases to 300Ω, the ROCOV Not action Not action
MGV Not action Not action
ROCOV based method cannot work well. The comparison of
VL Not action Not action
the fault resistance endurance is presented in Table 7.
2500 200
D. Response to Noise Disturbance
2000
300
According to Section II.B, the signal-to-noise ratio (SNR)
ROCOV(kV/ms)

1500
set1
represents the noise intensity. Adding disturbance with 20db,
1000
18db, 15db, 13db and 10db noise on the measured signals to
500
evaluate the impact of noise on the three protection schemes.
The results are shown in Table 9. As can be seen, Once the
0
0% 25% 50% 75% 100% F2 noise is stronger than 15dB, the ROCOV and MGV based
OHL12
Most severe
external fault protection schemes will output the wrong identification result.
(a) ROCOV protection In contrast, the VL based method is robust to noise. For ROCOV
1000 200
300
based protection, to mitigate the impact of noise, the first-order
800 damping element can be adopted, which has been introduced in
MMG(kV)

600
set2
[106]. For MGV based protection, the morphological filter (MF)
400 can be employed to mitigate the impact.
200
Table 9. The impact of noise disturbance on protection schemes.
00 Protection schemes SNR(db) Identification result
0% 25% 50% 75% 100% F2
Most severe 20 Correct
OHL12 external fault 18 Correct
(b) MGV protection ROCOV 15 Correct
500 200 13 False
300
400 10 False
set3 20 Correct
VL(kV)

300
18 Correct
200 MGV 15 Correct
100 13 Correct
10 False
0 20 Correct
0% 25% 50% 75% 100% F2
Most severe 18 Correct
OHL12 external fault VL 15 Correct
(c) VL protection 13 Correct
Fig. 27. Response to different fault resistances under PTP faults. 10 Correct
Table 7. Comparison of maximum detectable fault resistance E. Detection time
Protection schemes Maximum detectable fault resistance Rf_max
This paper is a post-print of a paper submitted to and accepted for publication in IEEE Transaction on Power Electronics and is subject to
Institution of Electrical and Electronic Engineering Copyright. The copy of record is available at IEEE Xplore Digital Library

There are two factors determining the detection time. One is the DCCB device protection will be activated once the fault
the time delay of the fault start-up element. The other is the current flowing through the DCCB exceeds the threshold.
time-window for the fault characteristic capture. In the case of Considering that DCCB protection on the adjacent lines may
internal faults, the fault start-up element is activated within one also be triggered, the second-stage protection is partially
sampling interval. For the fault protection algorithm, once the selective. When the first and second-stage protection refuse, the
initial TW arrives, the measured ROCOV, MGV and VL will MMC protection is triggered to block the converter and acts as
exceed the corresponding thresholds within several sampling the third-stage protection. Since the blocking of MMC will
intervals. During the design of the faulted pole selection cause the shutdown of the entire grid, the third-stage protection
criterion, a time-window with several milliseconds is adopted is non-selective [32].
to capture the differences between the positive and negative (4) Coordination between the hardware and the protection
poles. Based on equation (22), when the current-limiting algorithms, the control systems should be considered. The
inductance is selected as 0.1H, the detection time is 0.3ms+x/vp. illustration of the relationship between hardware devices and
It demonstrates that the detection time should be no more than protection algorithms is shown in Fig. 28. As can be seen, the
0.3ms when the fault occurs at the terminal of the transmission limitations of hardware devices may constraint the use of some
line. Considering that the sampling step is 0.04ms, the excellent protection algorithms and control methods. For
time-window for the faulted pole selection is selected to be example, the signal processing based protection schemes [84]
0.24ms (=0.04ms×6), which is smaller than 3ms. Hence, for and the active injection control [109] have outstanding
any fault distance, the detection time for the three protection performance in fault identification and fault location. However,
schemes are: they require high sampling frequency, which is a challenge for
td =0.04ms ( fault start - up)  0.24ms (time window) current hardware devices (10~50kHz in practical engineering).
(26) Therefore, those coordination protection schemes maybe not
 x / v p (TW propagation)
operable at present. Whereas, with the development of
hardware devices, the coordination between hardware,
IX. FUTURE TRENDS protection algorithms and control system will be a feasible
Based on the aforementioned analysis, regarding the approach.
protection scheme design for the MMC based DC grids, the MMC1
Bus1 DCCB12 DCCB21 Bus2 MMC2
future trends can be summarized as shown in Fig. 29.
(1) Coordination between the control system and the TV
protection system should be considered and designed Other lines Sampling
Tripping signal
[107][108]. For low-impedance faults (LIF), the fault circuit
voltage and
Other lines
characteristics are significant, e.g., the DC voltage drops A/D
current Protection algorithm written in

rapidly and the fault current rises quickly, which can be easily trans verter
FPGA

detected. The main purpose of the control system is to limit the Signal processing
fault current to ensure the safe operation of power system. unit (DSP+FPGA)

Whereas, under high-impedance faults (HIF), the fault

Fault f eatures

characteristics are less obvious compared with LIF. Thus, the Fig. 28 Illustration of the relationship between hardware devices and protection
control systems are recommended to actively provide algorithms.
information to improve fault protection reliability. (5) The impacts of renewable energy integration should be
(2) Coordination between different protection schemes under evaluated. Firstly, due to the intermittence and fluctuation
different fault conditions should be considered and designed. characteristics of renewable energy, the power flow will change
For LIFs, due to the large fault current, a single-end protection under normal states, which may result in the false operation of
with ultra-fast speed is required to guarantee the safe operation some protection schemes [110]. Also, some protection
of power system. For HIFs, the fault current is smaller, and a thresholds should be adjusted correspondingly. To smooth the
longer detection time margin is allowed. Hence, a longer time fluctuations of renewable power, energy storage is adopted.
window can be adopted to improve reliability. Considering However, the impact of energy storage on fault detection needs
these fault conditions, the single-end protection schemes work to be investigated. Secondly, to guarantee the uninterrupted
as the main protection. While the pilot protection, which has a operation of the integration system, the fault ride-through (FRT)
slow speed but high selectivity, can be adopted as the controls are commonly implemented [111]. These FRT controls
supplementary protection of the main protection. Once the will affect the transient voltage and current, which will in turn
single-end and pilot protection both refuse, the protection of affect the fault protection schemes. Moreover, the coordination
adjacent lines should respond to the faults, which are regarded between FRT controls and protection systems should be
as back-up protection of the whole line. considered. Thirdly, as pointed out in [112], higher penetration
(3) Coordination between protection schemes at different of renewable generation will deteriorate the system inertia and
stages should be considered and designed [106]. Similar to the cause the frequency instability problem. Thus, it is required to
three-stage fault protection under AC faults, there is also incorporate the DC grid protection and the frequency stability
three-stage protection for the DC grids. The first stage is the to guarantee system reliability.
transmission line protection, which is adopted to isolate the (6) The protection thresholds should be determined adaptively
faulted lines selectively. The second stage is the DCCB device and adjusted online. The renewable energy integration, the FRT
protection. In the case of refusal of the first-stage protection, controls, the active current-limiting controls and flexible loads
This paper is a post-print of a paper submitted to and accepted for publication in IEEE Transaction on Power Electronics and is subject to
Institution of Electrical and Electronic Engineering Copyright. The copy of record is available at IEEE Xplore Digital Library

will result in the frequent change of system operating states. thresholds should be smaller to improve the robustness to fault
Thus, the protection thresholds should be adjusted online to resistances. To be concluded, the protection thresholds should
respond to these changes. In addition, for LIFs, the thresholds be determined adaptively based on different fault conditions.
should be larger, thereby avoiding the false protection
operation under external faults. Whereas, for HIFs, the
Trend 6: adaptive threshold determination Trend 5: the impacts of the renewable energy
access on the protection systems
Control mode change
The decreasing inertia Frequency stability
System parameter change Adaptive threshold
Power flow change adjustment online FRT Impacts of FRT on the fault detection
The intermittence and Power flow change + Energy storage+ response
Fault conditions (such as: Fault resistance fluctuation characteristics time

Renewable Trend 4: coordination between

implement hardware and protection schems
Trends energy
Sampling frequency Active control
Coordination

Trend 1: coordination between Trend 2: coordination between different protection Trend 3: coordination between
the control system and protection schemes under different fault conditions protection at different stages
system
LIF HIF stage 1
Fault characteristics Line protection
(selectivity)
If refuses
Single-end + fast speed Single-end + a longer time stage 2
Significant Non-significant
window DCCB device protection (partial
Control system The main protection
selectivity)
If refuses
Current Improve the Pilot Protection Protection of adjacent lines
limiting detection reliability stage 3
The back-up protection of the The back-up protection for the MMC protection
main protection (non-selectivity)
whole line
Fig. 29. Future trends of the MMC HVDC grid protection.

protection can be summarized into three categories, the

X. CONCLUSION time-domain TW based methods, the signal processing based
Focusing on the DC fault protection of MMC HVDC grids, methods and the current limiting inductor voltage based
this paper introduces the key technical issues and methods. To improve the performance under weak boundary
systematically analyzes the fault behaviors in modal-domain, conditions, single-end protection schemes independent of
time-domain and frequency-domain, respectively. Typical boundary effects are required, which requires further
protection schemes dedicated to MMC HVDC grids are investigation.
reviewed and compared. It can be concluded that, 4) Simulations demonstrate that ROCOV, MGV and VL
1) Unlike AC power systems and LCC-HVDC systems, protection schemes are fast and not affected by the change of
speediness is the top concern for MMC HVDC grid protection. operation mode and AC faults. The VL method is robust to noise
Thus, ultra-fast single-end protection schemes are mostly disturbance (>10dB) while vulnerable to fault resistance (≤
adopted. To overcome the contradiction between speediness 200Ω). The MGV method is robust to fault resistance (>300Ω).
and sensitivity, a high sampling frequency is required to The endurance to noise disturbance of the MGV based method
improve the single-end protection reliability. And WT or MG (>13dB) is stronger than that of the ROCOV method (<15dB).
are adopted to detect the wave-front of the TW. 5) In future studies, the protection coordination needs to be
2) The modal-domain analysis method can be adopted to considered and designed, including the coordination between
decouple the couplings between positive and negative poles, the hardware, the control systems and the protection algorithms,
providing a theoretical basis to analyze the asymmetric faults. the coordination between different protection schemes under
The zero-mode component is adopted to select the faulted pole different fault conditions, and the coordination between
while the line-mode component is adopted to identify the protection schemes at different stages. In addition, the impacts
internal faults. The time-domain analysis method is mostly of renewable energy integration, energy storage and FRT
used to analyze the TW propagation behaviors, revealing the controls on the protection schemes need to be investigated. To
differences in voltage/ current changes between external and improve the reliability under different fault conditions, the
internal faults. The frequency-domain analysis method is used thresholds should be adjusted online and determined
to analyze the difference in high-frequency components of adaptively.
transient current and voltage, thereby adopting the WT and MG
to extract the high-frequency components. REFERENCES
3) Most existing single-end protection schemes are boundary [1] Z. Li, Z. Wang, Yue Wang, et al., “Accurate Impedance Modeling and
protection that relies on large current-limiting inductances to Control Strategy for Improving the Stability of DC System in
Multiterminal MMC-Based DC Grid,” IEEE Trans. Power Electron., vol.
guarantee high selectivity and sensitivity. The boundary 35, no. 10, pp. 10026-10049, Oct. 2020.
This paper is a post-print of a paper submitted to and accepted for publication in IEEE Transaction on Power Electronics and is subject to
Institution of Electrical and Electronic Engineering Copyright. The copy of record is available at IEEE Xplore Digital Library

[2] W. Lin, J. Wen and S. Cheng, “Multiport DC-DC Autotransformer for [24] Leterme, Willem, D. V. Hertem, “Classification of Fault Clearing
Interconnecting Multiple High-Voltage DC Systems at Low Cost,” IEEE Strategies for HVDC Grids,” CIGRÉ Symposium, Lund 2015.
Trans. Power Electron., vol. 30, no. 12, pp. 6648-6660, Dec. 2015. [25] M. J. Pérez-Molina, D. M. Larruskain, P. E. López, et al. “Challenges for
[3] K. Huang, W. Xiang, L. Xu and Y. Wang, “Hybrid AC/DC hub for Protection of Future HVDC Grids”. Frontiers in Energy Research, vol. 8,
integrating onshore wind power and interconnecting onshore and offshore pp. 1-6., Feb. 2020.
DC networks,” IET Renew. Power Gener., vol. 14, no. 10, pp. 1738-1745, [26] B. Ni, W. Xiang, M. Zhou, et al., “A mechanical DCCB with re-closure
Jul. 2020. capability and its performance in MMC based DC grid,” Int. J. Elec.
[4] H. Pang and X. Wei, “Research on Key Technology and Equipment for Power Syst., vol. 121, 2020.
Zhangbei 500kV DC Grid,” 2018 International Power Electron. Conf. [27] S. Wang, C. Li, O. D. Adeuyi, et al., “Coordination of MMCs With
(IPEC-Niigata 2018 -ECCE Asia), Niigata, 2018, pp. 2343-2351. Hybrid DC Circuit Breakers for HVDC Grid Protection,” IEEE Trans
[5] S. Yang, W. Xiang, X. Lu, et al., “An Adaptive Reclosing Strategy for Power Del., vol. 34, no. 1, pp. 11-22, Feb. 2019.
MMC-HVDC Systems with hybrid DC Circuit Breakers,” IEEE Trans. [28] X. Han, W. Sima, M. Yang, et al., “Transient characteristics underground
Power Del., vol. 35, no. 3, pp. 1111-1123, Jun. 2020. and short-circuit faults in a ±500kV MMC-based HVDC system with
[6] G. P. Adam, T. K. Vrana, R. Li, et al., “Review of technologies for DC hybrid DC circuit breakers,” IEEE Trans Power Del., vol. 33, no. 3, pp.
grids–power conversion, flow control and protection,” IET Power 1378-1387, Jun. 2018.
Electron., vol. 12, no. 8, pp. 1851-1867, 2019. [29] S. Pirooz Azad, D. Van Hertem, “A Fast Local Bus Current-Based
[7] X. Li, Q. Song, W. Liu, et al., “Protection of Nonpermanent Faults on DC Primary Relaying Algorithm for HVDC Grids,” IEEE Trans. Power Del.,
Overhead Lines in MMC-Based HVDC Systems,” IEEE Trans. Power vol. 32, no. 1, pp. 193-202, Feb. 2017.
Del., vol. 28, no. 1, pp. 483-490, Jan. 2013. [30] S. Yang, W. Xiang, J. Wen, “Review of DC fault protection methods for
[8] G. P. Adam, K. H. Ahmed, S. J. Finney, et al., “New Breed of Network the MMC based DC grid,” Proceedings of the CSEE, vol. 39, no. 22, pp.
Fault-Tolerant Voltage-Source-Converter HVDC Transmission System,” 6600-6616, Nov. 2019.
IEEE Trans. Power Syst., vol. 28, no. 1, pp. 335-346, Feb. 2013. [31] W. Xiang, S. Yang, L. Xu, et al., “A Transient Voltage based DC Fault
[9] R. Li, J. E. Fletcher, L. Xu, et al., “A Hybrid Modular Multilevel Line Protection Scheme for MMC based DC Grid Embedding DC
Converter with Novel Three-Level Cells for DC Fault Blocking Breakers,” IEEE Trans. Power Del., vol. 34, no. 1, pp. 334-345, Feb.
Capability,” IEEE Trans. Power Del., vol. 30, no. 4, pp. 2017-2026, Aug. 2019.
2015. [32] W. Leterme, M. Barnes, D. Van Hertem, “Signal processing for fast fault
[10] A. Nami, J. Liang, F. Dijkhuizen, et al., “Modular Multilevel Converters detection in HVDC grids,” CSEE J. Power and Energy Syst., vol. 4, no. 4,
for HVDC Applications: Review on Converter Cells and Functionalities,” pp. 469-478, Dec. 2018.
IEEE Trans. Power Electron., vol. 30, no. 1, pp. 18-36, Jan. 2015. [33] Y. M. Yeap, N. Geddada, K. Satpathi, et al., “Time- and
[11] J. Qin, M. Saeedifard, A. Rockhill, et al., “Hybrid Design of Modular Frequency-Domain Fault Detection in a VSC-Interfaced Experimental
Multilevel Converters for HVDC Systems Based on Various Submodule DC Test System,” IEEE Trans. Ind. Informa., vol. 14, no. 10, pp. 4353-
Circuits,” IEEE Trans. Power Del., vol. 30, no. 1, pp. 385-394, Feb. 2015. 4364, Oct. 2018.
[12] W. Xiang, W. Lin, T. An, et al., “Equivalent Electromagnetic Transient [34] Q. Huang, G. Zou, S. Zhang, et al., “A Pilot Protection Scheme of DC
Simulation Model and Fast Recovery Control of Overhead VSC-HVDC Lines for Multi-Terminal HVDC Grid,” IEEE Trans Power Del., vol. 34,
Based on SB-MMC,” IEEE Trans. Power Del., vol. 32, no. 2, pp. 778-788, no. 5, pp. 1957-1966, Oct. 2019.
Apr. 2017. [35] W. Leterme, D. Van Hertem,. “DC fault phenomena and DC grid
[13] E. C. Mathew, M. B. Ghat and A. Shukla, “A Generalized protection. Chap. 17,” in HVDC Grids : For Offshore and Supergrid of the
Cross-Connected Submodule Structure for Hybrid Multilevel Future, New Jersey: IEEE Press Wiley, 345-370, 2016.
Converters,” IEEE Trans. Ind. Appli., vol. 52, no. 4, pp. 3159-3170, July. [36] S. Yang, W. Xiang and J. Wen, “An Improved DC Fault Protection
2016. Scheme Independent of Boundary Components for MMC based HVDC
[14] W. Xiang, W. Lin, L. Xu, et al., “Enhanced independent pole control of Grids,” IEEE Trans. Power Del., early access, to be published.
hybrid MMC-HVDC system,” IEEE Trans. Power Del., vol. 33, no. 2, pp. [37] W. Xiang, H. Zhang, S. Yang, et al, “A Differential Pilot Protection
861-872, Apr. 2018. Scheme for MMC Based DC Grid Resilient to Communication Failure,”
[15] G. P. Adam, I. E. Davidson, “Robust and generic control of full-bridge IEEE J. Emerg. Sel. Topics Power Electron, early access, to be published.
modular multilevel converter high-voltage DC transmission systems,” [38] S. Jamali, S. S. Mirhosseini, “Protection of transmission lines in
IEEE Trans. Power Del., vol. 30, no. 6, pp. 2468-2476, Dec. 2015. multi-terminal HVDC grids using travelling waves morphological
[16] A. A. Elserougi, A. M. Massoud and S. Ahmed, “A Switched-Capacitor gradient,” Int. J. Elec. Power Syst., vol. 108, pp. 125-134, 2019.
Submodule for Modular Multilevel HVDC Converters With DC-Fault [39] X. Dong, S. Luo, S. Shi, et al., “Implementation and Application of
Blocking Capability and a Reduced Number of Sensors,” IEEE Trans. Practical Traveling-Wave-Based Directional Protection in UHV
Power Del., vol. 31, no. 1, pp. 313-322, Feb. 2016. Transmission Lines,” IEEE Trans. Power Del., vol. 31, no. 1, pp. 294-302,
[17] X Yu, Y Wei, Q Jiang, “New submodule circuits for modular multilevel Feb. 2016.
current source converters with DC fault ride through capability,” 2016 [40] I. Jahn, N. Johannesson, S. Norrga, “Survey of methods for selective DC
IEEE Applied Power Electron. Conf.e and Exposition (APEC), Long fault detection in MTDC grids,” 13th IET International Conf.on AC and
Beach, USA, 2016, pp. 1468-1474. DC Power Transm. (ACDC 2017), Manchester, 2017, pp. 1-7.
[18] T. H. Nguyen, K. A. Hosani, M. S. E. Moursi, et al., “An Overview of [41] Protection System Reliability Fundamentals, Dec. 2010,
Modular Multilevel Converters in HVDC Transmission Systems With http://www.nerc.com, NERC Report.
STATCOM Operation During Pole-to-Pole DC Short Circuits,” IEEE [42] J. Ren, J. Duan, B. Zhang, P. Li, “Identification of Lightning Disturbance
Trans. Power Electron., vol. 34, no. 5, pp. 4137-4160, May 2019 in Ultra-High-Speed Transmission Line Protection,” 2005 IEEE/PES
[19] C. Chen, G. P. Adam, S. Finney, et al., “H-bridge modular multi-level Transm. & Distri. Conf. & Exposition: Asia and Pacific, Dalian, 2005, pp.
converter: control strategy for improved DC fault ride-through capability 1-5.
without converter blocking," IET Power Electron., vol. 8, no. 10, pp. [43] F. Kong, B. Zhang, “A novel disturbance identification method based on
1996-2008, Oct. 2015. empirical mode decomposition for HVDC transmission line protection,”
[20] A Nami, L Wang, F Dijkhuizen, et al., “Five level cross connected cell for 12th IET International Conf. on Developments in Power Syst. Protection
cascaded converters,” 15th European Conf. on Power Electron. and Appli., (DPSP 2014), Copenhagen, 2014, pp. 1-6.
Lille, France, 2013, pp. 1-9. [44] B. Zhang, J. Duan, S. Luo, et al., “Systematic Study of Ultra-High-Speed
[21] E. C. Mathew, M. B. Ghat, A. Shukla, “A Generalized Cross-Connected Transient Based Protection for EHV/UHV Transmission Line,” 2008 IET
Submodule Structure for Hybrid Multilevel Converters,” IEEE Trans. Ind. 9th International Conf. on Developments in Power Syst. Protection (DPSP
Appli., vol. 52, no. 4, pp. 3159-3170, July-Aug. 2016. 2008), Glasgow, 2008, pp. 239-245.
[22] G. P. Adam, F. Alsokhiry, I. Abdelsalam, et al., “Hybrid converter [45] S. Mallat, W.L. Hwang, Singularity detection and processing with
topologies for dc transmission systems,” IET Power Electron., vol. 12, no. wavelets, IEEE Trans. Inf. Theory, vol. 38, no. 2, pp. 617-643, Mar. 1992.
3, pp. 607-619, 2019. [46] P. Maragos, “Representations for morphological image operators and
[23] J. He, K. Chen, M. Li, et al., “Review of protection and fault handling for analogies with linear operators,” Adv. Imaging Electron Phys., vol. 177,
a flexible DC grid,” Prot. and Control of Modern Power Syst., early acess, pp. 45-187, 2013.
to be published. [47] J. Zhang, C. Y. Chung, L. Guan, “Noise Effect and Noise-Assisted
This paper is a post-print of a paper submitted to and accepted for publication in IEEE Transaction on Power Electronics and is subject to
Institution of Electrical and Electronic Engineering Copyright. The copy of record is available at IEEE Xplore Digital Library

Ensemble Regression in Power System Online Sensitivity Identification,” [71] W. Leterme, M. Wang, D. Van Hertem, “Fault Discrimination in HVDC
IEEE Trans. Ind. Informat., vol. 13, no. 5, pp. 2302-2310, Oct. 2017. Grids with Reduced Use of HVDC Circuit Breakers,” 2018 IEEE Power
[48] J. L. Blackburn, Protective Relaying Principles and Applications. & Energy Society General Meeting (PESGM), Portland, OR, 2018, pp.
NewYork: Marcel Dekker, 1998. 1-5.
[49] J. Wang, B. Berggren, K. Linden, et al., “Multi-Terminal DC System line [72] C. Zhang, G. Song, T. Wang, et al., “An Improved Non-Unit Traveling
Protection Requirement and High Speed Protection Solutions,” CIGRÉ Wave Protection Method With Adaptive Threshold Value and Its
Symposium, Lund, 2015. Application in HVDC Grids,” IEEE Trans. Power Del., vol. 35, no. 4, pp.
[50] C. Li, C. Zhao, J. Xu, et al., “A Pole-to-Pole Short-Circuit Fault Current 1800-1811, Aug. 2020.
Calculation Method for DC Grids,” IEEE Trans. Power Syst., vol. 32, no. [73] C. Zhang, G. Song, X. Dong, “Non-unit Ultra-high-speed DC Line
6, pp. 4943-4953, Nov. 2017. Protection Method for HVDC Grids using First Peak Time of Voltage,”
[51] E. D. Kimbark, “Transient Over voltages Caused by Monopolar Ground IEEE Trans. Power Del, early access, to be published.
Fault on Bipolar DC Line: Theory and Simulation,” IEEE Trans. Power [74] C. Zhang, G. Song, L. Yang, et al., “Non-unit travelling wave protection
Apparatus Sys., PAS-89, no. 4, pp. 584-592, Apr. 1970. method for dc transmission line using waveform correlation calculation,”
[52] Y. Zhang, N. Tai, B. Xu, “Fault Analysis and Traveling Wave Protection IET Gener., Transm. Distrib., vol. 14, no. 12, pp. 2263-2270, May 2020.
Scheme for Bipolar HVDC Lines,” IEEE Trans. Power Del., vol. 27, no. [75] C. Zhang, G. Song, X. Dong, “A Novel Traveling Wave Protection
3, pp. 1583-1591, Jul. 2012. Method for DC Transmission Lines Using Current Fitting,” IEEE Trans.
[53] S. Yang, W. Xiang, R. Li, et al., “An Improved DC fault Protection Power Del, early access, to be published.
Algorithm for MMC HVDC Grids based on Modal Domain Analysis,” [76] C. Zhang, G. Song, T. Wang, et al., “Non-unit Traveling Wave Protection
IEEE J. Emerg. Sel. Topics Power Electron., early access, to be of HVDC Grids Using Levenberg-Marquart Optimal Approximation,”
published. IEEE Trans. Power Del, early access, to be published.
[54] Z. He, J. Hu, L. Lin, et al., “Pole-to-Ground Fault Analysis for [77] Q. Huang, G. Zou, X. Wei, et al., “A non-unit line protection scheme for
MMC-HVDC Grid,” 2019 10th International Conference on Power MMC-based multi-terminal HVDC grid,” Int. J. Elect. Power Energy
Electronics and ECCE Asia (ICPE 2019 - ECCE Asia), Busan, Korea Syst., vol. 107, pp. 1-9, 2019.
(South), 2019, pp. 2984-2989. [78] W. Leterme, D. Van Hertem, “Cable Protection in HVDC Grids
[55] K. Saleh, A. Hooshyar, E. F. El-Saadany, “Fault detection and location in Employing Distributed Sensors and Proactive HVDC Breakers,” IEEE
medium-voltage DC microgrids using travelling-wave reflections,” IET Trans. Power Del, vol. 33, no. 4, pp. 1981-1990, Aug. 2018.
Renew. Power Gener., vol. 14, no. 4, pp. 571-579, 2020. [79] K. D. Kerf, K. Srivastava, M. Reza, et al., “Wavelet-based protection
[56] Y. Wang, X. Fan, B. Zhang, “The Analytical Analysis and Protection strategy for DC faults in multi-terminal VSC HVDC systems,” IET Gener.
Setting of Traveling Wave Protection in VSC-HVDC Grid,” Proceedings Transm. Distrib., vol.5, no. 4, pp. 496-503, Mar. 2011.
of the CSEE, vol. 39, no. 11, pp. 3201-3212, 2019. [80] B. Li, Y. Li, J. He, et al., “A Novel Single-Ended Transient-Voltage
[57] Z. He, K. Liao, X. Li, et al, “Natural Frequency-Based Line Fault Based Protection Strategy for Flexible DC Grid,” IEEE Trans. Power
Location in HVDC Lines,” IEEE Trans. Power Del. vol. 29, no. 2, pp. Del., vol. 34, no. 5, pp. 1925-1937, Oct. 2019.
851-859, Apr. 2014. [81] L. Sabug, A. Musa, F. Costa, et al., “Real-time boundary wavelet
[58] M. Li, Y. Luo, K. Jia, et al., “Frequency-based current differential transform-based DC fault protection system for MTDC grids,” Int. J.
protection for VSC-MVDC distribution lines,” Int. J. Elect. Power Elect. Power Energy Syst., vol. 115, 2020.
Energy Syst, vol. 117, 2020. [82] S. Zhang, G. Zou, C. Wang, et al., “A Non-unit Boundary Protection of
[59] L. Tang, X. Dong, S. Shi and B. Wang, “Analysis of the characteristics of DC Line for MMC-MTDC Grids,” Int. J. Elect. Power Energy Syst., vol.
fault-induced travelling waves in MMC-HVDC grid,” The J. of 116, 2020.
Engineering, vol. 2018, no. 15, pp. 1349-1353, Oct. 2018. [83] S. Zhang, G. Zou, Q. Huang, et al., “Single-ended Line Protection for
[60] ABB. 5SNA 3000K452300 datasheet, https://search.abb.com/library/ MMC-MTDC Grids,” IET Gener. Transm. Distrib., vol. 13, no. 19, pp.
Download.aspx? DocumentID=5SYA1450& Language Code=en& 4331-4338, Aug. 2019.
Document PartId= &Action= Launch. [84] M. Kong, X. Pei, H. Pang, et al, “A lifting wavelet-based protection
[61] E. Kontos, S. Rodrigues, R. T. Pinto, et al., “Optimization of limiting strategy against DC line faults for Zhangbei HVDC Grid in China,” 19th
reactors design for DC fault protection of multi-terminal HVDC European Conf. on Power Electron. and Applica. (EPE'17 ECCE
networks,” Proc. IEEE Energy Convers. Congr. Expo., 2014, pp. Europe), Warsaw, Poland, Sept. 2017, pp. 1-11.
5347-5354. [85] S. Jamali, S. S. Mirhosseini, “Protection of transmission lines in
[62] M. Abedrabbo, W. Leterme, D. Van Hertem, “Systematic Approach to multi-terminal HVDC grids using travelling waves morphological
HVDC Circuit Breaker Sizing,” IEEE Trans. Power Del. vol. 35, no. 1, gradient,” Int J Electr Power Energy Syst. vol. 108, pp. 125-134, Jan.
pp. 288-300, Feb. 2020. 2019.
[63] D. Jovcic, W. Lin, S. Nguefeu, et al., “Low-Energy Protection System for [86] J. Descloux, B. Raison, J. Curis, “Protection algorithm based on
DC Grids Based on Full-Bridge MMC Converters,” IEEE Trans. Power differential voltage measurement for MTDC grids,” 12th IET
Del., vol. 33, no. 4, pp. 1934-1943, Aug. 2018. International Conf. on Developments in Power Syst. Protection (DPSP
[64] J. Sneath, A. Rajapakse, “Fault detection and Interruption in an earthed 2014), Copenhagen, 2014, pp. 1-5.
HVDC grid using ROCOV and hybrid DC breakers,” IEEE Trans. Power [87] R. Li, L. Xu, L. Yao, “DC fault detection and location in meshed
Del., vol.31, no.3, pp. 973-981, Jun. 2016. multiterminal HVDC systems based on DC reactor voltage change rate,”
[65] W. Leterme, S. Pirooz Azad, D. Van Hertem, “HVDC grid protection IEEE Trans. Power Del., vol.32, no.3, pp. 1516-1626, Jun. 2017.
algorithm design in phase and modal domains,” IET Renewable Power [88] C. Li, A. M. Gole, C. Zhao, “A fast DC fault detection method using DC
Generation, vol. 12, no. 13, pp. 1538-1546, Jun. 2018. reactor voltages in HVdc Grids,” IEEE Trans. Power Del., vol. 33, no. 5,
[66] F. Kong, Z. Hao, S. Zhang, et al., “Development of a Novel Protection pp. 2254-2264, Oct. 2018.
Device for Bipolar HVDC Transmission Lines,” IEEE Trans. Power Del., [89] M. A. Abdel-Moamen, S. A. Shaaban, et al., “France-Spain HVDC
vol. 29, no. 5, pp. 2270-2278, Oct. 2014. Transmission System with Hybrid Modular Multilevel Converter and
[67] N. Geddada, Y. M. Yeap, A. Ukil, “Experimental Validation of Fault Alternate-arm Converter,” 2017 Innovations in Power and Advanced
Identification in VSC-Based DC Grid System,” IEEE Trans. Ind. Computing Technologies (i-PACT), Vellore, 2017, pp. 1-6.
Electron., vol. 65, no. 6, pp. 4799-4809, Jun. 2018. [90] W. Lin, D. Jovcic, S. Nguefeu, et al., “Full-Bridge MMC Converter
[68] M. Ikhide, S. Tennakoon, A. Griffiths, S. Subramanian, et al., “Fault Optimal Design to HVDC Operational Requirements,” IEEE Trans.
detection in Multi-Terminal Modular Multilevel Converter (MMC) based Power Del., vol. 31, no. 3, pp. 1342-1350, Jun. 2016.
High Voltage DC (HVDC) transmission system,” 2015 50th International [91] X. Chen, H. Li, Y. Liang, et al., “A protection scheme for hybrid
Universities Power Engineering Conf. (UPEC), 2015, pp: 1-6. multi-terminal HVDC networks utilizing a time-domain transient voltage
[69] J. Sneath,A. D. Rajapakse, “DC Fault Protection of a Nine-Terminal based on fault-blocking converters,” Int J Electr Power Energy Syst. vol.
MMC HVDC Grid,” 11th IET International Conference on AC and DC 118, 2020.
Power Transmission, Birmingham，2015，pp：1-8． [92] Z. Dai, H. Zhu, H. Su, et al., “DC Line Protection for Flexible MVDC
[70] W. Leterme, J. Beerten, D. V. Hertem, “Nonunit protection of HVDC Distribution Grid With Renewable Energy Resources,” 2018 IEEE Power
grids with inductive DC cable termination”, IEEE Trans. Power Del., vol. & Energy Society General Meeting (PESGM), Portland, OR, 2018, pp.
31, pp. 820-828, Apr. 2016. 1-6.
This paper is a post-print of a paper submitted to and accepted for publication in IEEE Transaction on Power Electronics and is subject to
Institution of Electrical and Electronic Engineering Copyright. The copy of record is available at IEEE Xplore Digital Library

[93] N. Tong, X. Lin, Y. Li, et al., “Local Measurement-Based a visiting student at the University of Aberdeen and the University of
Ultra-High-Speed Main Protection for Long Distance VSC-MTDC,” Strathclyde in 2014 and 2016 respectively. Currently, he is a research associate
IEEE Trans. Power Del., vol. 34, no. 1, pp. 353-364, Feb. 2019. with the University of Strathclyde since 2018. His main research interests
[94] K. A. Saleh, A. Hooshyar, E. F. El-Saadany, et al., “Protection of include MMC-HVDC, high power dc/dc converters and dc grids.
High-Voltage DC Grids Using Traveling-Wave Frequency
Characteristics,” IEEE Systems Journal, early access, to be published. Saizhao Yang obtained his B.E. degree in electrical
[95] R. C. Santos, S. L. Blond, D. V. Coury, et al., “A novel and engineering from Huazhong University of Science
comprehensive single terminal ANN based decision support for relaying and Technology (HUST), China, in 2018. He is
of VSC based HVDC links,” Electr. Power Syst. Res., vol. 141, pp. currently pursuing his PhD degree at HUST. His
333-343, Dec. 2016. research interests include dc fault protection of
[96] Q. Yang, S. L. Blond, R. Aggarwal, et al., “New ANN method for MMC-HVDC and dc grids.
multi-terminal HVDC protection relaying,” Electr. Power Syst. Res., vol.
148, pp. 192-201, Jul. 2017.
[97] A Hossam-Eldin, A Lotfy, M Elgamal, et al, “Artificial intelligence-
based short-circuit fault identifier for MT-HVDC systems,” IET Gener.
Transm. Distrib., vol. 12, no.10, pp. 2436-2443, May 2018.
[98] R. Bertho, V. A. Lacerda, R. M. Monaro, J. C. M. Vieira and D. V. Coury, G.P. Adam (M’12) received a PhD in Electrical
“Selective Nonunit Protection Technique for Multiterminal VSC-HVDC Engineering from University of Strathclyde, UK, in
Grids,” IEEE Trans. Power Del., vol. 33, no. 5, pp. 2106-2114, Oct. 2018. 2007. Dr Adam is presently a lead engineer with
[99] D. Tzelepis, A. Dysko, G. Fusiek, et al., “Advanced fault location in NEOM, Energy Division, in the Kingdom of Saudi
MTDC networks utilising optically-multiplexed current measurements Arabia. Previously he was with Institute of Energy
and machine learning approach,” Int. J. Elect. Power Energy Syst., vol. 97, and Environment, University of Strathclyde,
pp. 319-333, Apr. 2018. Glasgow, UK. His research interests are: fault tolerant
[100] N. Tong, X. Lin, C Li, et al., “Permissive pilot protection adaptive to DC voltage source converters for medium and
fault interruption for VSC-MTDC,” Int. J. Elect. Power Energy Syst., vol. high-voltage applications; modelling and control of HVDC transmission
123, 2020. systems and multi-terminal HVDC networks; and emerging challenges of
[101] S. Gao, G. Song, Z. Ma, et al., “Novel Pilot Protection Principle for power electronics dominated power systems. Dr Adam has authored and
High-voltage Direct Current Transmission Lines based on Fault co-authored over 60 technical reports as part of his works in several industrial
Component Current Characteristics,” IET Gener. Transm. Distrib., vol. 9, and research projects, and over 100 journal and conference articles in
no. 5, pp. 468-474, Feb. 2015. fundamental and applications of power electronics in power systems. Also, he
[102] J. Liu, N. Tai, C. Fan, “Transient-Voltage-Based Protection Scheme for has authored and co-authored three books in fundamental and applications of
DC Line Faults in the Multiterminal VSC-HVDC System,” IEEE Trans. power electronics in power systems and renewable energy generation and
Power Del., vol. 32, no. 3, pp. 1483-1494, Jun. 2017. integration. Dr Adam is a reviewer for several IEEE and IET Transactions and
[103] L. Chen, X. Lin, Z. Li, et al., “Similarity Comparison Based High-Speed Journals, served in program and review committees of several conferences,
Pilot Protection for Transmission Line,” IEEE Trans. Power Del., vol. 33, member of IEEE and IEEE Power Electronics Society, and presently serving in
no. 2, pp. 938-948, Apr. 2018. editorial board of JESTPE as associate editor. Dr Adam has received prizes of
[104] B. Li, Y. Li, J. He, et al., “An Improved Transient Traveling-wave Based the best paper award in ICRERA conference 2018 and 2020, PES 2018
Direction Criterion for Multi-terminal HVDC Grid,” IEEE Trans. Power reviewer of the year, and star reviewer of the JESTPE 2019.
Del., early access, to be published.
[105] M. J. Thompson, D. L. Heidfeld, “Transmission line setting calculations -
beyond the cookbook,” 2015 68th Annual Conference for Protective Relay Haobo Zhang obtained his B.E. degree in electrical
Engineers, College Station, TX, 2015, pp. 850-865. engineering from Huazhong University of Science and
[106] W. Lin, D. Jovcic, S. Nguefeu, et al., “Coordination of MMC converter Technology (HUST), China, in 2020. He is currently
protection and DC line protection in DC grids,” 2016 IEEE Power and pursuing his master degree at HUST. His research
Energy Society General Meeting (PESGM), Boston, MA, 2016, pp. 1-5. interests include dc fault protection of MMC-HVDC
[107] B. Ni, W. Xiang, M. Zhou, et al., “An Adaptive Fault Current Limiting and dc grids.
Control for MMC and Its Application in DC Grid,” IEEE Trans. Power
Del., early access, to be published.
[108] W. K. A. Najy, H. H. Zeineldin, W. L. Woon, “Optimal Protection
Coordination for Microgrids with Grid-Connected and Islanded
Capability,” IEEE Trans. Ind. Electron., vol. 60, no. 4, pp. 1668-1677,
Apr. 2013.
[109] G. Song, J. Hou, B. Guo, et. al., “Active Injection for Single-ended Wenping Zuo received the B.S. degree and Ph.D
Protection in DC Grid Using Hybrid MMC,” IEEE Trans. Power Del, degree in electrical engineering from Huazhong
early access, to be published. University of Science and Technology (HUST), Wuhan,
[110] S. Cao, W. Xiang, X. Lu, et al., “Energy dissipation of MMC-HVDC China, in 2009 and 2017, respectively. Currently he is a
based onshore wind power integration system with FB-DBS and DCCB,” Post-Doctoral Research Fellow with HUST. His
IET Renewable Power Gener., vol. 14, no. 2, pp. 222-230, Feb. 2020. research interests include DC grid key equipment,
[111] C. Feltes, H. Wrede, F.W. Koch, et al., “Enhanced fault ride-through energy storage, and renewable energy integration.
method for wind farms connected to the grid through VSC-based HVDC
transmission,” IEEE Trans. Power Syst., vol. 24, no. 3, pp. 1537-1546,
Aug. 2009.
[112] J. Dave, H. Ergun, D. Van Hertem, “Incorporating DC Grid Protection, Jinyu Wen (M’10) received his B.Eng. and Ph.D.
Frequency Stability and Reliability Into Offshore DC Grid Planning,” degrees all in electrical engineering from Huazhong
IEEE Trans. Power Del., early access, to be published. University of Science and Technology (HUST), Wuhan,
China, in 1992 and 1998, respectively. He was a visiting
student from 1996 to 1997 and a research fellow from
2002 to 2003 all at the University of Liverpool, UK, and
BIOGRAPHIES a senior visiting researcher at the University of Texas at
Arlington, USA in 2010. From 1998 to 2002 he was a
Wang Xiang (S’16-M’17) received his B.Eng. and director engineer in XJ Electric Co. Ltd. in China. In
PhD degrees both in electrical engineering from 2003 he joined the HUST and now is a Professor at
Huazhong University of Science and Technology HUST. His current research interests include renewable energy integration,
(HUST), China in 2012 and 2017, respectively. He was energy storage application, DC grid, and power system operation and control.
This paper is a post-print of a paper submitted to and accepted for publication in IEEE Transaction on Power Electronics and is subject to
Institution of Electrical and Electronic Engineering Copyright. The copy of record is available at IEEE Xplore Digital Library