Review of DC-DC Converters for Multi-terminal HVDC Transmission
Networks
G.P. Adam, I. A. Gowaid, S.J. Finney, D. Holliday and B.W. Williams
Abstract: This paper presents a comprehensive review of high-power dc-dc converters for high-voltage direct current (HVDC)
transmission systems, with emphasis on the most promising topologies from established and emerging dc-dc converters.
Additionally, it highlights the key challenges of dc-dc converter scalability to HVDC applications, and narrows down the
desired features for high-voltage dc-dc converters, considering both device and system perspectives. Attributes and limitations
of each dc-dc converter considered in this study are explained in detail and supported by time-domain simulations. It is found
that the front-to-front quasi two-level operated modular multilevel converter, transition arm modular converter and controlled
transition bridge converter offer the best solutions for high-voltage dc-dc converters that do not compromise galvanic isolation
and prevention of dc fault propagation within the dc network. Apart from dc fault response, the MMC dc auto transformer and
the transformerless hybrid cascaded two-level converter offer the most efficient solutions for tapping and dc voltage matching
of multi-terminal HVDC networks.
Key words: High-voltage dc-dc converter; modular multilevel converter; multi-terminal high-voltage dc networks; and
prevention of dc faults propagation.
I. INTRODUCTION
For many decades researchers have recognised the possibilities multi-terminal high-voltage dc (HVDC) transmission networks
can offer when compared to well-established high-voltage ac (HVAC) systems[1-15]. For the majority of this period, the
difficulty of power reversal in complex multi-terminal networks based on line commutating converter high-voltage dc (LCC-
HVDC) technology has prevented development of generic dc grids with seamless control over the power flow in any of its
branches [16]. Also, the increased dependency of the LCC terminal on the ac network strength has caused significant concerns
regarding ac voltage stability of relatively weak ac networks, especially when large power is being exchanged [8, 16-26]. The
emergence of voltage source converter high-voltage dc (VSC-HVDC) in the early 1990s [10, 12, 27-46] that can reverse dc
power without any difficulty (power reversal is achieved without the need to reverse the dc voltage) has reinvigorated research
into generic multi-terminal HVDC networks. Recent consideration of offshore wind farms by many European countries with
real possibilities of building offshore multi-terminal HVDC networks and inter-regional dc grids, has encouraged serious
research and development effort from major HVDC manufacturers and academia [1, 8, 12, 43-45, 47-59]. These efforts include
all technology chains which are necessary for practical realisation of multi-terminal HVDC networks, such as dc switchgear
(dc circuit breakers and fast disconnectors) and high-voltage dc-dc converters [44, 47, 60-76]. Besides dc voltage matching,
high-voltage dc-dc converters are expected to play a pivotal role of splitting a large dc grid into several protection zones
(smaller dc networks, each capable of sustaining itself as an individual network or operating as part of a large network), thus
preventing entire system collapse during a severe dc network fault. These dc-dc converters (also known as dc-transformers) are
expected to provide galvanic isolation. The majority of the most promising dc-dc converters for HVDC applications known
today are developed around the front-to-front (F2F) dual active bridge originally developed by De Doncker for low and
medium-voltage applications [65, 77]. The topology is best suited for low and medium transformation ratios. These dc-dc
converters deliberately use an intermediate ac link to make a dc fault on one side appears as a controllable ac overcurrent at the
healthy side converter. Blocking the healthy converter will therefore be sufficient to isolate the faulty part.
These converters must therefore be seen as an enabling technology for a multi-terminal HVDC network as they minimise the
number of dc circuit breakers, and allow dc grids to behave in a similar fashion as ac grids during dc network faults. Because it
is not possible to cover all the dc-dc converters proposed in the literature for HVDC applications, this paper focuses on the
established and emerging dc-dc converters, which show promise for HVDC applications. A brief discussion of each dc-dc
converter will be presented, with broader emphasis on aspects related to power electronic systems. The discussions will be
supported by a number of illustrative simulations.
II. CHALLENGES AND REQUIREMENTS FOR HIGH-VOLTAGE DC TRANSFORMERS
To focus investigation, this paper summarises some of the technical issues related to scalability of dc-dc converters for multi-
terminal HVDC networks as follows:
a) F2F two-level and neutral-point clamped dc transformers require robust methods for static and dynamic voltage sharing of
the series device connection in order to be able to operate with dc voltage suitable for HVDC transmission systems.
However experiences from early generations of HVDC links show that the use of series connected IGBTs is limited to
±200 kV [47, 78].
b) Although adoption of a high fundamental frequency (1 kHz to 2 kHz) is attractive for reduced size and weight of magnetic
components, switching of large voltage steps (400 kV or higher) at such frequencies impresses extremely high dv/dt upon
interfacing transformers. This will make transformer design more challenging (that is, insulation and ability to transmit
powers associated with dominant low-order harmonics such as the 3
rd
, 5
th
and 7
th
).
c) Benefits of multi-level techniques in dc-dc applications, where fundamental frequencies are much higher than 50Hz may
not be significant. This is because the reduction in transformer size will be limited by the minimum clearance between
terminals and phases, bushing creepage, and transformer body structure must be mechanically strong to be able to support
the weight of long high-voltage bushings, including magnetic forces during normal and abnormal operation. Thus,
insulation and isolation within high voltage dc-dc converters will present major volume constraints.
Considering the importance of high-voltage dc-transformers for practical realisation of multi-terminal HVDC network with dc
operating voltage of up to 800 kV (pole-to-pole), the most desirable features for high-voltage dc-dc converters are:
1) Scalable to high-voltage, and are likely required to use an interfacing transformer for galvanic isolation [65, 79], and
better utilisation of switching devices at the converter with higher dc link voltage. In this manner, circulating reactive
power in the ac link is minimized, without de-rating of the switching devices. Besides voltage matching, a dc-dc
converter must be able to act as a dc power or dc voltage controller and prevent dc fault propagation within the dc
network, without exposing its switching devices to risk of damage.
2) Voltage stresses (dv/dt) presented at the primary and secondary windings of the interfacing transformer, and voltage
stresses across dc-dc converter switching devices and passive components must be fully controlled.
3) Since the size and weight reduction of the interfacing transformer in a high-voltage dc-dc converter is limited by the
level of switching losses and other high-voltage and mechanical considerations as stated previously, the fundamental
frequency in the ac link must be constrained to less than 1 kHz [65]. In multi-module dc-dc converters, where each
sub-module operates at relatively low dc voltage and contributes a small fraction of the total output power, higher
fundamental frequency can be achieved at the ac link [79, 80].
4) The dc-dc converter must be able to perform black-start at the ac link and controlled recharge of the dc link following
dc fault isolation on any one of its dc terminals. Thus, modulation index control over a wide (0-1) range is required.
III. REVIEW OF ESTABLISHED AND EMERGING DC-DC CONVERTER TOPOLOGIES
A) Two-level converter dual active bridge (DAB)
Fig. 1 shows an example of a dc-dc converter that uses a typical two-level dual active bridge with series connected IGBTs
(insulated gate bi-polar transistors) to enable operation at high dc voltage. With the use of a fundamental frequency ranging
from 250 Hz to 1 kHz in the ac link, the overall size and weight of the dc-dc converter can be reduced, without significant
efficiency sacrifice. Traditionally, such two-level dual active bridges are operated in a square wave mode at the fundamental
frequency, where each arm conducts for 180
o
(half a fundamental period), with the load angle between v
ao1
and v
ao2
(Fig. 1) is
traditionally used for power flow control between VSC
1
and VSC
2
[65, 77, 81-83]. In this operating mode, self-commutated
semiconductor devices in the DAB tend to turn on and off at zero currents for much of the operating range (while anti-parallel
diodes are in conduction); thus, low switching loss is achieved [65]. The use of pulse width modulation can provide an
additional degree of freedom, which can be exploited to minimise the circulating reactive power in the ac link between VSC
1
and VSC
2
. However, with fundamental frequency range stated, the use of high-frequency pulse width modulation (PWM) must
be precluded, because the increase in switching losses is expected to outweigh the gain that will be achieved by increased
control flexibility [65, 76, 84, 85]. To avoid this shortcoming of the high frequency PWM, low-frequency modulation schemes
such as selective harmonic elimination (SHE) with one notch per quarter cycle can be used to achieve the necessary control
flexibility, specifically ac voltage and reactive power control in the ac link between VSC
1
and VSC
2
, especially during a dc
fault. However, with SHE the range at which inherent soft switching during DAB turn-on and off is achieved will be reduced;
hence, switching losses are expected to increase.
Fig. 2 (a) shows switched ac voltages VSC
1
and VSC
2
the two-level DAB in Fig. 1 impress on the primary and secondary
windings of the medium-frequency transformer in the ac link. In this illustration, the fundamental frequency is 500 Hz, both
VSC
1
and VSC
2
are operated using fundamental frequency switching with 180
o
conduction, input and output dc link voltages of
VSC
1
and VSC
2
are ±400 kV and ±350 kV respectively, and the power flow direction is from VSC
1
to VSC
2
. Fig. 2 (c) shows
sampled dc power measured at the dc link of VSC
2
(when VSC
2
ramps the power flow from VSC
1
and VSC
2
, from 0 to 800
MW). The switched voltage waveform at the terminal of VSC
1
(v
ao1
for phase ‘a’) leads that of VSC
2
(v
ao2
) as expected when
the power flow direction is from VSC
1
to VSC
2
. Theoretically, the switch voltages v
ao1
and v
ao2
of the VSC
1
and VSC
2
can be
expressed as:
, , .
, , .
sin
()
sin ( )
()
dc
ao
n
dc
ao
n
V
nt
vt
n
V
n t n
vt
n
1
1
1 3 5
2
2
1 3 5
2
2
(1)
where δ is the angle of v
ao2
relative to v
ao1
and n=2j+1,
j
. If the medium-frequency transformer is assumed to be lossless
and with L
T
transformer leakage inductance referred to primary, the primary instantaneous current can be expressed as:
( ) ( ) sin ( ) sin ( )
d c d c
a o a o
nn
T
VV
i t i n t k n t n
L n n
12
11
11
22
22
11
22
1
0
(2)
With the instantaneous power at the terminals of the lead converter VSC
1
expressed as p
1
(t)=3v
ao1
i
ao1
(t), the average power
VSC
1
exchanges with VSC
2
is:
sin
()
T
d c d c
n
T
VV
n
P p t d t k
T L n
12
11
23
1
0
6
1
(3)
where k is the turn ratio (primary to secondary).
Equation (3) indicates that the medium-frequency transformer must be designed to transmit all the powers associated with
significant low-order harmonics, including that of the fundamental component. Here, the load angle is the only available
degree of freedom that can be used to control power flow. When SHE is used as depicted in Fig. 2 (b), expressions for v
ao1
and
v
ao2
become [86]:
, , .
, , .
co s
( ) sin
( ) co s sin ( )
dc
ao
n
dc
ao
n
n
V
v t n t
n
V
v t n n t n
n
1
1
1
1 3 5
2
22
1 3 5
21
2
2
1
21
(4)
The average power VSC
1
exchanges with VSC2 is:
sin
( ) c o s c o s
T
d c d c
n
T
VV
n
P p t d t k n n
T L n
12
1 1 1 2
23
1
0
6
1
2 1 2 1
(5)
Equations (4) and (5) show that with SHE the power flow in the ac link can be controlled using voltage magnitudes at the
terminals of VSC
1
and VSC
2
, and the phase shift between these voltages (load angle δ). Fig. 2(b) and (d) show ac voltages
VSC
1
and VSC
2
present at the primary and secondary windings of the coupling transformer, and dc power measured at the dc
link of VSC
2
when power flows from VSC
1
and VSC
2
, and VSC
2
ramps its power command from 0 to 800 MW (both VSC
1
and VSC
2
are controlled using SHE). From Fig. 2 (b) and (d), although SHE introduces an additional degree of freedom, the
principle of controlling the DAB remains the same.
With continuously increasing dc operating voltage of voltage source converter based high-voltage direct current (VSC-HVDC)
transmission systems, the rate of change of voltage dv/dt that the two-level DAB in Fig. 1 impresses upon the medium-
frequency transformer between VSC
1
and VSC
2
becomes intolerable, and restricts its applications to relatively low power and
dc voltages of up to ±200 kV dc, as in early generation two-level and neutral-point clamped based VSC-HVDC links.
Fig. 1: IGBT based two-level converter dual active bridge (V
dc1
=800 kV, V
dc2
=700 kV, 500Hz fundamental frequency and medium frequency transformer rated
at 1000 MVA, 500 kV/450 kV with 10% per unit reactance)
(a) Voltage waveforms at the terminals of VSC
1
and VSC
2
(v
ao1
and
v
ao2
) when power flow is from VSC
1
to VSC
2
(fundamental
frequency switching)
(b) Voltage waveforms at the terminals of VSC
1
and VSC
2
(v
ao1
and
v
ao2
) when power flow is from VSC
1
to VSC
2
(SHE with active
power and ac voltage control in the high frequency ac link)
(c) Power flow in the dc link of VSC
2
(fundamental frequency
switching, VSC
2
ramps its power command from 0 to 800 MW)
(d) Power flow in the dc link of VSC2 (SHE, VSC
2
ramps its power
command from 0 to 800 MW)
Fig. 2: Waveforms illustrating the basic operation of a two-level converter DAB (fundamental frequency=500 Hz)
B) Modular multilevel converter dual active bridge
As power handling and the dc operating voltage of the VSC-HVDC links continue to increase, the modular multilevel
converter dual active bridge (MMC-DAB) with a medium-frequency transformer in the ac link in Fig. 3 is more likely to be
adopted. The use of a medium-frequency ac link is not only beneficial for compact transformer design, but also leads to overall
reduction in the size of the MMC-DAB passive elements such as cell capacitances and arm reactors. Practically, the ac link of
an MMC-DAB can be operated using full multilevel modulation with sinusoidal voltage or in a quasi two-level mode with
trapezoidal voltage as suggested in [53]. These two possibilities will be explored in the subsequent parts.
(a)
(b)
Fig. 3: Modular multilevel converter dual active bridge
a) Multilevel operation of MMC-DAB with sinusoidal voltage and currents in the ac link
This mode operates the MMC-DAB in Fig. 3 using multilevel modulation similar to that used in the converter terminal of the
typical VSC-HVDC link, where the MMC is connected to the ac grid [50]. Basic operating principle of the MMC can be
![Fig. 10: Generic schematic diagram of dc-dc modular multilevel converter [112]](/figures/fig-10-generic-schematic-diagram-of-dc-dc-modular-multilevel-1h3rzivw.webp)
![Fig. 8: Alternative arm modular multilevel converter dual active bridge (A2M2C-DAB)]](/figures/fig-8-alternative-arm-modular-multilevel-converter-dual-2oa5ftur.webp)
