This paper is a preprint of a paper submitted to the Proceedings of Power Electronics, Machines and Drives 2024 (PEMD

2024) and is subject to Institution of Engineering and Technology Copyright. If accepted, the copy of record will be available at
IET Digital Library.
arXiv:2402.03978v2 [eess.SY] 5 Jun 2024

1
 RECONFIGURABLE POWER CONVERTERS WITH
INCREASED UTILIZATION FOR UNBALANCED
POWER DISTRIBUTION SYSTEM APPLICATIONS
Matthew Deakin* , Xu Deng

School of Engineering, Newcastle University, Newcastle-upon-Tyne, UK

*E-mail: matthew.deakin@newcastle.ac.uk

Keywords: MULTIPORT CONVERTER, CONVERTER MULTIPLEXING, POWER DISTRIBUTION,

UNBALANCE MITIGATION

Abstract
A low-cost reconfiguration stage connected at the output of balanced three-phase, multi-terminal ac/dc/ac converters can increase
the feasible set of power injections substantially, increasing converter utilization and therefore achieving a lower system cost.
However, the approach has yet to be explored for phase unbalance mitigation in power distribution networks, an important
application for future energy systems. This study addresses this by considering power converter reconfiguration’s potential for
increasing the feasible set of power transfers of four-wire power converters. Reconfigurable topologies are compared against both
conventional four-wire designs and an idealised, fully reconfigurable converter. Results show that conventional converters need
up to 75.3% greater capacity to yield a capability chart of equivalent size to an idealised reconfigurable converter. The number
and capacity of legs impact the capability chart’s size, as do constraints on dc-side power injections. The proposed approach
shows significant promise for maximizing the utilization of power electronics used to mitigate impacts of phase unbalance.

1 Introduction 11 : 0.4 kV

The electrification of demand and uptake of distributed gener-

ation in domestic properties will lead to a significant increase
in load on low voltage power (LV) distribution networks [1]. Goal: max.
Problem:
utilization
In many regions, domestic customers are connected to a sin- unbalanced
via balanced 4W VSC load,
gle phase supply, or have a three-phase supply feeding single load system
phase loads [2]. In the absence of active phase balancing, it
can be expected the vast majority of assets that would have to Fig. 1: A four-wire (4W) voltage source converter (VSC)
be replaced due to thermal constraints would be under-utilized system can be used to balance phase currents by injecting
prior to their replacement. This is because it is the phase with unbalanced active powers. Balanced phase currents ensures
the highest loading, rather than the average loading, which lim- that network assets can be fully utilized.
its operation of the distribution system components such as
substation transformers [3]. transformer), then dynamically injecting unbalanced power to
As a result, there has been rapidly growing interest in reduce loading on heavily loaded phases.
approaches to increase utilization of congested branches prior If the phase unbalance mitigation provided by a 4W VSC
to their reinforcement. Proposed approaches include static real- system changes in time, then the loading on the individual legs
location of single-phase loads [4], dynamic customer phase of the interfacing ac/dc converter will also vary. As a result,
reallocation via static transfer switch-based ‘phase shifting a 4W VSC system appears to be an attractive proposition for
devices’ [5–7], the control of single-phase customer genera- the use of the multiplexing approach described in [11, 12]. In
tors [8], or the use of distributed energy resources solutions particular, it is proposed that 4W VSC systems are constructed
interfaced through a three-phase four-wire (4W) power con- of an arbitrary number of legs, whose current-carrying capacity
verter. The latter approach is shown in Fig. 1, with works is not necessarily uniform, and whose output can be connected
published considering such an approach including standalone to any of the four wires of the power distribution system. By
Static Compensator (STATCOM)-based approaches [9], bat- allowing the capacity connected to each of the four wires to
tery energy storage [10], or back-to-back ‘soft open points’ change dynamically, the feasible set of power transfers can be
(SOPs) [7]. However, the approach is not limited to these and augmented. The purpose of this work is to explore this feasible
could also include any other dc resource–e.g., an LVDC cable, set (i.e., the capability chart), and explore how different sizing
a three-phase electric vehicle fast charger, or industrial motor strategies can increase the size of this set.
drive. These four wire (4W) voltage source converter (VSC)- There are a few works that have explored the use of power
based systems can support increased utilization by monitor- converter reconfiguration for mitigating phase unbalance. In
ing phase currents at a congested branch (e.g., a substation

2
[13], the authors describe a ‘phase changing SOP’, assumed to
be constructed using a multi-terminal ‘phase shifting device’ DC Link,
in recent works [7]. However, we show in this work that a Four- Source, Four-
wire Load or wire
three-leg, three-phase inverter with equally sized legs cannot VSC Energy VSC
benefit from reconfiguration on the output. Alternatively, works Storage
on balanced operation of SOPs and their capability charts have (a) Standalone 4W VSC System (b) Interconnected 4W VSC System
been considered in detail in [11, 12]. However, to the authors’
knowledge, there are no works that consider capability charts Fig. 2: In this work, we classify Four Wire VSC Systems as
of reconfigurable power converters for arbitrary unbalance mit- either (a) Standalone, which cannot transfer power to or from
igation, including the critical effect of neutral currents. This the dc subsystem (but can transfer power between phases), or
is an important and timely gap, as the proposed reconfigura- (b) Interconnected, with some resource connected to the dc
tion approach promising to lower the cost of power electronics subsystem to enable net power transfer across the VSC.
by increasing utilization substantially [12], and thereby provid-
ing a route for network operators to install more cost-effective 2.1 Proposed Four-Wire Reconfigurable Power Converter
active solutions to address distribution system congestion.
The major contribution of this work is to address this gap The essence of the proposed approach has been described in
by proposing and calculating capability charts for the pro- detail in [12] for a balanced, multi-terminal back-to-back VSC
posed four-wire reconfigurable power converters. This includes system. In a conventional VSC system (Fig. 3a), the output
numerical approaches to determine the size of these capability of each of the equally-sized half-bridge (HB) legs are hard-
charts, and corresponding visualizations of these sets. The aim wired to the four output wires; the maximum current that can
is provide metrics that can demonstrate the improved flexibility be injected into any one of the phases is static. In contrast, the
of the reconfiguration approach. proposed approach can have any number of legs, m, whose pu
This work is structured as follows. In Section 2, we describe size α can be varied (at the design stage), and which can be
the proposed topology and define corresponding capability connected to any of the outputs to vary the HB capacity con-
charts, then describe how the area (or volume) of these charts nected to a given phase on-the-fly. It is worth noting that if the
can be calculated numerically to demonstrate performance mission profile for the 4W VSC does not vary with time, then
improvement for a given 4W VSC design. In Section 3, a range the proposed approach will not provide a benefit.
of case studies are presented to highlight the improvements that There are two major differences between the balanced and
can be expected as a result of the proposed reconfiguration. unbalanced cases. Firstly, neutral currents can be neglected
Finally, salient conclusions are drawn in Section 4. in the balanced case. In contrast, as we demonstrate in Sec-
tion 3.1, neutral currents are often large and so must be cal-
2 Methodology culated for an accurate determination of the capability chart.
Secondly, individual legs of the VSC are considered, rather
In this work, two basic structures are considered for the 4W than three-phase VSC units as in, e.g., [12].
VSC system, as shown in Fig. 2. A Standalone 4W VSC, as For the purposes of this work, but without loss of generality,
shown in Fig. 2a, consists of only a four-wire VSC and dc we consider the main purpose of such a system is to inject only
capacitor, and so the power balance restricts the currents that active power P (and not reactive power), for the following rea-
can be injected (i.e., the sum of active powers injected into sons. Firstly, analysis of smart meter data suggests that modern
the network must have value zero). The standalone device is domestic customers have very good power factor–for example,
similar to a STATCOM, which has been used for providing it is suggested in [14] that a power factor of unity ought to be
dynamic reactive power compensation in power systems for assumed for simulation tasks. If this is the case, then injecting
several decades, and which can transfer power between phases reactive power will increase (rather than decrease) loading on
to mitigate high currents. In contrast, the Interconnected 4W each phase, and so is not useful for addressing thermal conges-
VSC systems, as shown in Fig. 2b, also has a dc-side compo- tion. Secondly, so long as the converter has sufficient capacity,
nent connected to enable a non-zero power injection from the issues with reactive power can be addressed with single phase
VSC into the ac network. assets injecting reactive power (where active power phase
In this Section, we outline how the capability chart of a rebalancing is not possible without changing the load point).
both of these classes of 4W VSC systems can be increased Finally, reactive power can be mitigated dynamically by other
by adding the proposed reconfiguration output stage. Firstly, low-cost means (e.g., switched capacitor banks). Nevertheless:
in Section 2.1 the mechanism by which this reconfiguration reactive power could be also considered in future applications,
increases the capability chart is revisited, highlighting with considering applications such as voltage unbalance mitigation.
a simple example how the proposed approach can be benefi-
cial under unbalanced operation. Subsequently, in Section 2.2, 2.1.1 Example Reconfigurable Converter: An example of a
capability charts are defined as a set, with it highlighted how reconfigurable converter, constructed of two HB triplet pack-
the size of these sets change as a function of total VSC capac- ages of 45 A and 15 A respectively are shown in Fig. 4. Such a
ity. Finally, in Section 2.3, numerical routines are described for converter can be connected in various operating modes.
calculating the size of these sets.

3
 Package: 3
half bridges
Source or
load Rating: 45 A
per leg

800V Selector
(a) Conventional design Battery switches

Package: 3
half bridges
Rating: 15 A
Source or per leg
...

load

legs selector Fig. 4: An example reconfigurable 4W VSC system with bat-

switch matrix
tery energy storage system and two HB packages.
(b) Proposed design with reconfiguation and leg resizing

Fig. 3: Conventional and proposed four-wire VSC system However, unbalanced current injections by a converter neces-
designs. The conventional design (a) has four legs hard-wired sitates consideration of neutral currents, and so in this section
to the ac output wires, where the proposed reconfigurable we revisit capability chart for the four-wire configuration.
design (b) has any number of legs m with arbitrary current For a 4W power converter operating with a fixed output
ratings α whose output can be reconfigured to redistributed cur- voltage and no dc-side constraints (e.g., assuming a large dc
rent carrying capacity as necessary. link capacitor) the power converter’s operating range is deter-
mined by the current rating of the capacity connected to each
wire Iϕmax ∈ R4+ (in Amps or pu). The phasors representing
• The converter can operate under balanced operation with the phase currents in each leg of the converter Iϕ ∈ C4 must
maximum power output. In this case, 60A is connected to therefore satisfy
phase 1–3; the neutral has no capacity allocated to it. In this
mode, the converter is acting like a conventional three-wire, |Iϕ [i]| ≤ Iϕmax [i] ∀ i ∈ {1, 2, 3, 4} . (1)
three-phase converter.
• The converter could be used to inject 90 A of active power
For such a system, consider the m per-unit leg capacities α ∈
into any of one of phases 1–3 to mitigate a temporary over- Rm , such that
load, with 90 A connected to the neutral (for the return Xm
current). In this mode, the converter is acting like a single- α[i] = 1 . (2)
phase converter. 50% extra current can be injected into the i=1
phase which the HBs are connected to as compared to the A variable binary matrix B ∈ {0, 1}4×m has m columns that
balanced converter. each represent the state of the m one-to-four selector switches,
• The battery is disconnected for maintenance. 45 A of capac- with the sum of each column of B having value 1 (to ensure
ity is connected to two of phases 1–3; 30 A to the other each leg is connected to just one output wire). Then, the capac-
phase; and 60 A to the neutral to enable phase rebalance ity connected to each output wire Iϕmax is simply the sum of the
currents (as shown in Section 3.1, the neutral current is typi- capacities of the legs connected to that wire, i.e.,
cally larger than the phase currents). The converter is acting
as a four wire standalone VSC system; depending on the Iϕmax = Ibase Bα , (3)
optimal operation, the 45 A and 30 A converters can be
reallocated. where the current Ibase has value of the sum of the capac-
In summary, the selector switches can enable a wide range ity of all legs (in Amps). Note that, if a converter is not
of unbalanced operation, with a higher utilization of the HBs reconfigurable, then B is replaced with an identity matrix.
as compared to a static design. In the next Section, we pro- For a four-wire converter injecting in-phase currents (i.e.,
pose the use of capability charts to systematically study the full active powers) into a stiff positive sequence voltage with fixed
range of operation (as compared to the specific operating points phase voltage magnitudes |V0 |, the per-phase powers P ∈ R3
highlighted in this example). are linked to those phase currents Iϕ according to the linear
relation
2.2 Defining Capability Charts P [i]
Iϕ (P )[i] = , i ∈ {1, 2, 3} , (4a)
A capability chart can be defined as the feasible operating |V0 |a(1−i)
region of a given power converter, i.e., the feasible currents 3
X
and/or voltages under which it can operate. This has been Iϕ (P )[4] = − Iϕ (P )[i] . (4b)
defined for balanced reconfigurable converters in [11, 12]. i=1

4
where a = e2πȷ/3 is the phase rotation operator. Note that the U Fix (3), U Fix (4) respectively. Note that U Fix (3) does not
neutral current INtrl. flows through the fourth wire in (4), have a neutral for return current.
• Designs with leg capacities chosen to maximise the Stan-
INtrl. = Iϕ [4] , (5) dalone system CCA, S4Opt. ; or the Interconnected system
CCV, I4Opt. . The sizes of each leg was chosen by evaluat-
and that the neutral is at the reference voltage at 0 V. As a result, ing the value of α that maximises the CCA or CCV using
the neutral cannot inject either active or reactive power (which a meshgrid of all α at a resolution of 0.01 pu. CCA and
is why there are only three components in the powers P ). CCV values for each α are determined using the method of
The capability chart can therefore defined as the feasible set Section 2.3.1.
of power injections P that satisfy the power converter operating • An idealised converter, denoted Ω–i.e., a design for which
constraints, i.e., leg capacity can be allocated continuously, providing an
upper bound on CCA and CCV values [12].
C = {P : (1), (3), (4)} . (6)
The capability chart for the idealised design Ω is found by
The advantage of the capability chart in terms of active powers replacing the leg reconfiguration constraint (3) with the con-
P (instead of current phasors Iϕ ) is that, in the general case of tinuous constraint
arbitrary P (i.e., an Interconnected 4W VSC), the dimension of |Iϕ (P )[i]| ≤ 1 . (10)
the set (dim(C) = 3) is the same as the number of variables in
P. 2.3 Calculating Capability Chart Area or Volumes
The total power injected into the ac grid, PTtl. , is
Determining the CCA or CCV requires the evaluation of the
PTtl. = P [1] + P [2] + P [3] . (7) multi-dimensional integrals (8). In general, such integrals can
be computationally challenging [12]. However, considering
For the case of a Standalone 4W VSC (Fig. 2a), ac-side pow- uniform converters U(m) or converters with four legs, these
ers must balance, i.e., PTtl. must have value of zero; for an integrals can be calculated very conveniently as we demon-
Interconnected 4W VSC, PTtl. can have a non-zero value. strate.
Given the definition of a capability chart as a set, it is nat-
ural to consider the size of the set as a property of interest 2.3.1 Brute Force Grid Methods: In this work, CCA or CCV
when comparing converters with different α. The capability values are determined by first considering a regular grid of
chart volume (CCV) or capability chart area (CCA) are points over a region R that is known to completely enclose
Z Z the capability chart C. An indicator function F (P, α) is then
CCV = dV , CCA = dA , (8) defined that returns value 1 if the point lies within the capability
C C chart and zero otherwise. The CCA or CCV is then conve-
niently found via a normalised sum of the indicator function
with the former used for an Interconnected 4W VSC, and the values at all points across the grid.
latter a Standalone 4W VSC (whose capability chart C is two- If there are four legs of the converter, then the indicator
dimensional,
P as the three powers P are constrained to lie in the function is
plane P [i] = 0). (
It is worth noting that the CCV and CCA scale with the cube 1 if α[i] ≤ |Iϕ [i]| ∀ i ∈ {1, 2, 3, 4} ,
and square of the converter power rating, respectively (as can F4 (P, α) = (11)
0 otherwise ,
be observed by considering the units of the integrals). For a
given converter, the scaling factors ηV , ηA that yield the same where P is linked to phase currents (4), Iϕ has been assumed
volume or area, respectively, for given CCA or CCVs are [12] as in per-unit, and (with slight abuse of notation) α and |Iϕ |
r r have been ordered according to their size. For converters with
3 CCV2 CCA2
ηV = , ηA = . (9) uniform leg sizing U(m), the indicator function is
CCV1 CCA1
( P4
For example, if the value of the CCA of a second converter is 1 if i=1
ceil(m|Iϕ (P )[i]|) ≤ m ,
FU (P, α) = (12)
double that of a first,
√ then the first converter would need to be 0 otherwise ,
a factor of ηA = 2 larger to have the same CCA.
where ceil(·) is the ‘ceiling’ function (i.e., non-integer argu-
2.2.1 HB Sizing Case Studies: To explore potential benefits of ments are rounded up to the nearest integer).
the proposed approach for 4W current unbalance mitigation,
we consider four sizing and reconfiguration approaches. 2.3.2 Evaluating the Boundary of the Capability Chart: A
• An m-leg reconfigurable converter, each with uniform leg complementary approach for exploring the capability chart C
sizes of capacity 1/m pu, denoted U(m). is by finding its boundary. This has two potential uses for anal-
• Three-leg and four-leg converters with uniform HB leg
ysis and visualization. Firstly, it can be used as a method to
capacities and legs hard-wired to phase wires, denoted determine the CCA or CCV. For example, if the radius r of the

5
boundary of the capability chart C of a Standalone 4W VSC is Table 1 The number of legs m and leg capacities α for
known as a function of the angle θ in the plane, then each design.
Z 2π Symbol m (no. legs) α (leg capacities)
CCA = r(θ)dθ . (13) S Opt.
4 (0.12, 0.22, 0.26, 0.4)
4
θ=0
Opt.
I 4 4 (0.13, 0.21, 0.3, 0.36)
This could be an advantage as it reduces the dimension of the 

grid of points that must be evaluated over (e.g., from 2 to 1 for 1 m×
U(m) m m
CCAs). Secondly, when the total power injection PTtl. is non- 

Fix 1 3×
zero, the boundary is not necessarily connected to the interior U (3) 3 3
of the capability chart (as we show in Section 3.2). It therefore 

1 4×
provides different information than that which is provided by U Fix (4) 4 4
considering the size of the set (i.e., the CCA or CCV) alone.
To efficiently determine the radius at the boundary r, it is Ω n.a. limm→∞ U(m)
proposed to consider the capability charts in αβγ co-ordinates
via the Clarke transformation, The radius at the boundary of the capability chart can be
found either by defining an appropriate indicator function and
P̂ = TClarke P , (14) then using a root-finding method, or via an optimization to
maximise r or rSph. . For example, for a fixed ψ and θ, the radius
where the Clarke transformation matrix TClarke is rSph. at the boundary of the capability chart can be determined
r  1  as the solution of a mixed-integer linear program,
−1/2 −1/2
√ √
2
TClarke =
3
0√ √ − √
3/2 3/2 . (15)
max rSph. (19a)
1/ 2 1/ 2 1/ 2
s.t. (1), (3), (4), (14), (18) . (19b)
The transformed powers P̂ in αβγ co-ordinates can be inter-
preted as follows. The third element, P̂ [3], is proportional to 2.3.3 Capability Chart Discontinuities: If the current injected
the total power injected PTtl. , as into one of the wires has value of zero, then no leg capacity
needs to be connected to that wire. This means that at there are
PTtl. regions of the capability chart which, according to the CCA
P̂ [3] = √ , (16) and CCV definitions, do not contribute to the size of the set,
3
as these subsets will be of lower dimension. Nevertheless, as
as can be seen by considering (7), (15), (18). This means we show in Section 3, these points are non-trivial, and could be
that co-ordinates P̂ [1], P̂ [2] lie in the plane of the Standalone very useful for supporting some mission profiles.
VSC system’s capability chart and correspond to unbalanced The subspaces of these points in a given co-ordinate system
power injections. Secondly, the radius r in polar or cylindrical can be found by setting one of the powers P to zero (or the
co-ordinates, neutral current INtrl. ), then solving for the parameters that cap-
r2 = P̂ [1]2 + P̂ [2]2 , (17) ture that locus. For example, for a fixed azimuth angle ψ (in
spherical co-ordinates (18)), the powers are a function of polar
is proportional to magnitude of the neutral current |INtrl. |. This angle θ as
can be seen by considering the magnitude of the neutral cur-
rent (4) mapped through the Clarke transformation (14). From
 
cos(ψ) 0  
(17), constraints on neutral currents are therefore circles in αβγ ⊺ sin(θ)
TClarke r sin(ψ) 0
  =P. (20)
co-ordinates (rather than ellipses in nominal co-ordinates). cos(θ)
0 1
Therefore, visualizations in αβγ co-ordinates have additional
symmetry and explainability. The angle θ that will yield the ith element of P to be zero can
For an Interconnected VSC, it is useful to represent powers be found by setting the ith row of (20) to be zero, then solving
P̂ in spherical co-ordinates, for θ.

P̂ [1] = rSph. sin(θ) cos(ψ) (18a) 3 Results

P̂ [2] = rSph. sin(θ) sin(ψ) (18b) In this section, we evaluate the size of the capability charts
P̂ [3] = rSph. cos(θ) , (18c) for the converter sizing approaches given in Table 1, consid-
ering both Standalone and Interconnected 4W VSC operation.
where rSph. is the spherical radius, ψ the azimuth angle, and θ This demonstrates how the proposed approach can improve
the polar angle. In contrast to cylindrical co-ordinates, if a point converter flexibility. Both qualitative and quantitative proper-
at rSph. is feasible for a given ψ, θ, then all points between the ties of these capability charts are considered to explore the
origin and rSph. will also be feasible. performance of these designs.

6
 
&RQVWUDLQW  
© ª
|Iφ [i]| 1/4, i ∈ 1, 2, 3
P[2]SX

 |Iφ [4]| 1/4

P̂ 2 SX
P2 SX
X
|Iφ [i]| 1  

 

   U(5) U(8) S 4Opt.
P[1]SX      
P1 SX P̂ 1 SX
Fig. 5: Constraints that create capability chart C for a fixed
converter U Fix (4) and the idealised converter Ω. (a) Nominal co-ordinates (b) αβγ co-ordinates

Fig. 6: Capability charts for a 5- and 8- converter uni-

3.1 Capability Chart Areas for Standalone Four-Wire VSCs form designs, U(5), U(8) and the optimal 4-leg reconfigurable
design S4Opt. . For comparison, the feasible set of the idealised
In the first instance, we consider the boundary of the capability and fixed four-leg design are also plotted as dashed lines.
chart C for a Standalone VSC for a conventional, fixed design
U Fix (4) and compare this against the idealised design Ω, as


shown in Fig. 5. The constraints on power injection on the
&&$EDVHGFRQYHUWHUVL]H
















phase legs form a hexagon (in direct analogy to the balanced




UDWLRηA 

three-terminal SOP case [11]). In contrast, the constraint on 


the neutral current forms an ellipse. Interestingly, this ellipse
is completely contained within the hexagon. This highlights 
that a converter with all four legs sized symmetrically highly 
underutilizes the three phase legs–only the neutral leg is active


on the boundary of the capability chart. The idealised capabil- 
U(5)
U(6)
U(7)
U(8)
U(9)

U (3)
U Fix (4)
Ω
U(10)
U(11)
U(12)
U(13)
U(14)
U(15)

I4 .
.
Opt
S Opt
ity chart completely encapsulates the 0.25 pu hexagon, and so

Fix
even if the neutral current were not the limiting factor (or an /HJVL]LQJDSSURDFK
alternative means was provided for the neutral return, such as
a split-capacitor), the idealised design Ω would still improve Fig. 7: Bar chart of the CCA-based converter size ratio ηA
performance as compared to a fixed four-leg VSC U Fix (4). (from (9)) for a range of leg sizing approaches.
In Fig. 6, the boundary of the capability charts for three
designs are illustrated in nominal and αβγ co-ordinates, with
as compared to the idealised converter Ω. It is also interesting
the capability chart of the benchmark cases Ω, U Fix (4) shown
to see that designs with uniformly-sized converters are rela-
in dashed lines. It can be seen in this figure that the capability
tively inefficient at increasing the capacity of the CCA, with 15
chart varies significantly depending on the sizing α. For the 5-
uniformly-sized converters required to achieve a larger CCA
converter uniform design U(5), 0.4 pu is allocated to the neutral
than the optimal design S Opt. (4). This highlights how effec-
and so the capability chart is a hexagon. For a uniform, 8-
tive choice of leg sizes α can yield substantially increased
converter uniform design U(8), the capability chart is enlarged
flexibility, echoing previous findings [11, 12].
significantly, although, it can be seen that that design is mostly
contained within the optimal 4-converter design S4Opt. . Addi-
3.2 Capability Chart Volumes for Interconnected Four-Wire
tionally, Fig. 6 shows that many of the properties of capability
VSCs
chart areas of reconfigurable balanced VSCs can be observed:
non-convexity; existence of capability charts as strict supersets The boundary of the capability chart for the conventional, fixed
of other capability charts; and, large increases in area compared design U Fix (4) and the idealised design Ω are plotted in Fig. 8
to a conventional design [11]. and Fig. 9, respectively, for a range of values of total power
The CCA-based converter sizing ratio ηA for the leg sizing injection PTtl. . As expected, it can be observed a total power
approaches collected in Table 1 are plotted in Fig. 7, consider- injection PTtl. with value zero matches the Standalone 4W
ing uniform designs with up to 15 legs m. From this figure, the VSC (Fig. 5); but, as the total power injection PTtl. increases,
poor performance of conventional designs U Fix (3), U Fix (4) the shape of the capability chart changes significantly. For
can clearly be seen–for the former, there is no neutral return, example, as PTtl. increases above 0.5 pu, the set of feasible
and so power can only be transferred between two phases at injections quickly reduces, reaching a single point at 1 pu injec-
a time (and so the CCA is trivially zero). For the latter, it tion. A power injection PTtl. of 0.5 pu leads to the greatest
can be seen that the fixed converter U Fix (4) requires 57.8% per-phase power injection potential of 0.5 pu (by comparison, if
capacity increase to achieve the same capacity as the opti-
mal design S4Opt. , with an upper limit of an increase 75.3%

7
 (a) ψ = 0◦ (b) ψ = 3◦ (c) ψ = 45◦
(a) Nominal co-ordinates (b) αβγ co-ordinates
Fig. 10: Capability charts C for three values of azimuth angle
Fig. 8: Boundary of the capability chart C for fixed values of
ψ for the interconnected VSC optimal four-leg design I4Opt. .
total power injection PTtl. for the conventional design U Fix (4).

(a) PTtl. = 0.45 pu (b) PTtl. = 0.6 pu (c) PTtl. = 0.75 pu

(a) Nominal co-ordinates (b) αβγ co-ordinates
Fig. 11: Capability charts C for three values of PTtl. for the
Fig. 9: Boundary of the capability chart C for fixed values of interconnected VSC optimal four-leg design I4Opt. .
total power injection PTtl. for the idealised design Ω.
charts when plotted for a fixed PTtl. , as shown in Fig. 11b and
the power injection PTtl. is 1 pu, the per-phase power injection Fig. 11c. For example, close to the origin, it may be the case
is 1/3 pu, i.e., 50% lower). that even with the smallest leg connected to the neutral, there is
For reconfigurable designs, the capability charts of Inter- not sufficient capacity to inject PTtl. in an (almost) balanced
connected 4W VSC systems are more complex than those way; so, there must be a reasonable amount of unbalanced
of Standalone systems. For example, the capability chart for power injected to enable the full value of PTtl. .
three angles of azimuth ψ are plotted in Fig. 10 (only the Secondly, it is interesting to note that when there is a fixed
first quadrant is plotted due to symmetry). The filled areas are PTtl. , there can be isolated points (Fig. 11a), isolated lines
calculated using Section 2.3.1’s brute-force method, with the (Fig. 11b), and isolated areas (Fig. 11c). This could lead to
red outer boundary line calculated by determining the bound- more complex operation as compared to a conventional power
ary in spherical co-ordinates (Section 2.3.2); the black dashed converter–a more complicated procedure would need to be run
line indicating the boundary of the idealised design. It can to determine the connection of the converters as the mission
be observed that the Interconnected VSC can have isolated profile changes, for example.
points, as described in Section 2.3.3. Due to the complex non- Finally, the CCV-based converter size ratios ηV for various
convexities in the capability chart, only snapshots at given ψ designs are shown in Fig. 12. Compared to the Standalone 4W
are given (as compared to the full capability charts which can VSC case, the gap between the size ratio ηV of the conven-
be plotted in Fig. 8 and Fig. 9 for the conventional and idealised tional design U Fix (4) and the idealised design Ω, the potential
designs, respectively). for improvement is slightly reduced to 62.7%. Nevertheless, it
For a fixed value of PTtl. , the capability charts can have remains a significant increase, and this ratio is still larger than
more complexity again. The capability charts for three values that of the three-terminal SOP considered in [11] (which has a
of PTtl. are shown in Fig. 11, with solid coloured areas are plot- value of ηA of 150% for the idealised design). In contrast to the
ted using Section 2.3.1’s brute-force methods (coloured area); Standalone VSC, uniform sizing is also much more effective at
red lines calculated using Section 2.3.2’s boundary method in increasing the CCV compared to optimal design I4Opt. , with the
cylindrical co-ordinates; and, black dashed lines showing the nine-converter design U(9) surpassing the CCV of the optimal
boundary of the capability chart of the idealised design Ω and design I4Opt. .
benchmark U Fix (4). There are two main differences between
these spherical and cylindrical capability charts representa-
tions. Most strikingly, there can be ‘holes’ in the capability

8
 [3] K. Ma, R. Li, I. Hernando-Gil, and F. Li, “Quantifica-


&&9EDVHGFRQYHUWHUVL]H

 tion of additional reinforcement cost from severe three-













 phase imbalance,” IEEE Transactions on Power Systems,
UDWLRηV 





vol. 32, no. 5, pp. 4143–4144, 2016.

[4] O. Pereira, J. Quirós-Tortós, and G. Valverde, “Phase
 rebalancing of distribution circuits dominated by single-


 phase loads,” IEEE Transactions on Power Systems,
U(5)
U(6)
U(7)
U(8)
U(9)

U Fix (3)
U Fix (4)
Ω
U(10)
U(11)
U(12)
U(13)
U(14)
U(15)

.
.

I Opt
S Opt
vol. 36, no. 6, pp. 5333–5344, 2021.

4
4
/HJVL]LQJDSSURDFK [5] F. Shahnia, P. J. Wolfs, and A. Ghosh, “Voltage unbalance
reduction in low voltage feeders by dynamic switching of
Fig. 12: Bar chart of the CCV-based converter size ratio ηV residential customers among three phases,” IEEE Trans-
(from (9)) for a range of leg sizing approaches. actions on Smart Grid, vol. 5, no. 3, pp. 1318–1327,
2014.
[6] B. Liu, F. Geth, N. Mahdavi, and J. Zhong, “Load balanc-
4 Conclusion ing in low-voltage distribution networks via optimizing
The reconfiguration of power converters using low-cost selec- residential phase connections,” in 2021 IEEE PES Inno-
tor switches in applications with variable mission profiles has vative Smart Grid Technologies-Asia (ISGT Asia). IEEE,
previously shown significant promise in balanced distribution 2021, pp. 1–5.
systems. In this work, the approach has been shown to be [7] X. Cui, G. Ruan, F. Vallée, J.-F. Toubeau, and Y. Wang,
effective as a potential approach for current unbalance miti- “A two-level coordination strategy for distribution net-
gation, with a conventional design requiring between 62.7% work balancing,” IEEE Transactions on Smart Grid,
and 75.3% increased capacity to match an idealised converter. vol. 15, no. 1, pp. 529–544, 2024.
Efficient numerical methods have been given to determine the [8] K. Girigoudar, M. Yao, J. L. Mathieu, and L. A. Roald,
feasible sets and boundary of the capability charts, with more “Integration of centralized and distributed methods to
complex capability charts as compared to previously consid- mitigate voltage unbalance using solar inverters,” IEEE
ered reconfigurable converters, particularly when the VSCs are Transactions on Smart Grid, vol. 14, no. 3, pp. 2034–
interconnected to a dc-side power source or load. 2046, 2022.
It is concluded that power converter reconfiguration can be [9] A. S. de Morais, F. L. Tofoli, and I. Barbi, “Model-
an effective tool for reallocating capacity of individual legs ing, digital control, and implementation of a three-phase
of power converters in several different applications. Future four-wire power converter used as a power redistribution
work could explore hardware approaches to implement the device,” IEEE Transactions on Industrial Informatics,
proposed reconfiguration approach to consider more clearly vol. 12, no. 3, pp. 1035–1042, 2016.
the practical challenges of implementing the selector switches. [10] M. Stecca, T. B. Soeiro, A. K. Iyer, P. Bauer, and P. Palen-
In addition, other optimal design and operational approaches sky, “Battery storage system as power unbalance redis-
will be considered to further increase converter flexibility and tributor in distribution grids based on three legs four wire
utilization. voltage source converter,” IEEE Journal of Emerging and
Selected Topics in Power Electronics, vol. 10, no. 6, pp.
5 Acknowledgements 7601–7614, 2022.
[11] M. Deakin, “Optimal hybrid multiplexed ac/dc/ac power
M. Deakin was supported by the Royal Academy of Engineer- converters,” in 19th International Conference on AC and
ing under the Research Fellowship programme. X. Deng was DC Power Transmission (ACDC 2023). IET, 2023.
supported by Newcastle University Academic Track (NUAcT) [12] ——, “Multiplexing power converters for cost-effective
Fellowship scheme. and flexible soft open points,” IEEE Transactions on
Smart Grid, vol. 15, no. 1, pp. 260–271, 2024.
6 References [13] C. Lou, J. Yang, T. Li, and E. Vega-Fuentes, “New phase-
changing soft open point and impacts on optimising
[1] A. Speakman, O. Harris, C. Birkinshaw-Doyle, unbalanced power distribution networks,” IET Genera-
and D. Mills, “Low voltage network capacity tion, Transmission & Distribution, vol. 14, no. 23, pp.
study,” UK Department for Business, Energy 5685–5696, 2020.
and Industrial Strategy (BEIS), Tech. Rep., July [14] M. Vanin, H. Ergun, R. D’hulst, K. Vanthournout, and
2022, https://www.gov.uk/government/publications/ D. Van Hertem, “Analysis and insights from reactive
low-voltage-network-capacity-study. power measurements of low voltage users,” in 27th Inter-
[2] K. Ma, L. Fang, and W. Kong, “Review of distribution national Conference on Electricity Distribution (CIRED
network phase unbalance: Scale, causes, consequences, 2023). IET, 2023.
solutions, and future research directions,” CSEE Journal
of Power and Energy systems, vol. 6, no. 3, pp. 479–488,
2020.