Accepted Manuscript

A unified control strategy for power sharing and voltage balancing in

bipolar DC microgrids

Saman Dadjo Tavakoli, Mohammad Mahdavyfakhr, Mohsen Hamzeh,

Keyhan Sheshyekani, Ebrahim Afjei

PII: S2352-4677(16)30211-9
DOI: http://dx.doi.org/10.1016/j.segan.2017.07.004
Reference: SEGAN 112

To appear in: Sustainable Energy, Grids and Networks

Received date : 17 December 2016

Revised date : 6 June 2017
Accepted date : 20 July 2017

Please cite this article as: S.D. Tavakoli, M. Mahdavyfakhr, M. Hamzeh, K. Sheshyekani, E. Afjei,
A unified control strategy for power sharing and voltage balancing in bipolar DC microgrids,
Sustainable Energy, Grids and Networks (2017), http://dx.doi.org/10.1016/j.segan.2017.07.004

This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to
our customers we are providing this early version of the manuscript. The manuscript will undergo
copyediting, typesetting, and review of the resulting proof before it is published in its final form.
Please note that during the production process errors may be discovered which could affect the
content, and all legal disclaimers that apply to the journal pertain.
 A Unified Control Strategy for Power Sharing and Voltage Balancing in
Bipolar DC Microgrids

Saman Dadjo Tavakolia , Mohammad Mahdavyfakhra , Mohsen Hamzeha , Keyhan Sheshyekanib,∗,

Ebrahim Afjeia
a
Department of Electrical Engineering, Shahid Beheshti University, Tehran, Iran.
b
Department of Electrical Engineering, Polytechnique de Montreal, Quebec, Canada.

Abstract

This paper investigates the challenges of power sharing and voltage balancing in bipolar dc micro-
grids equipped with voltage-source-inverters (VSIs) and voltage balancers. This is done by unifying
the control systems of voltage balancer and VSI so that the VSI is able to participate in both power
sharing and voltage balancing of the bipolar dc microgrid. The VSI operates in dc-link voltage
control mode whose reference signal is generated by the control system of the voltage balancer.
Two different schemes namely the dynamic reference and the equivalent droop are proposed for
the power sharing purpose. These schemes, in contrast with the traditional droop control, do not
deteriorate the voltage balancing of the microgrid. In the dynamic reference scheme, the reference
voltages generated by the control system of the voltage balancer are transmitted to distributed
generations (DGs) using a high-speed master-slave controller area network (CAN). Then, the droop
characteristics of DGs and voltage balancer are shifted downwards or upwards according to the
voltage references. In the equivalent droop scheme, however, the voltage references are calculated
locally by DGs and voltage balancer which share the values of their output currents over a low-speed
CAN bus. The performance of the proposed power sharing schemes and the unified control system
is verified using MATLAB simulation.
Keywords: Bipolar dc microgrid, droop control, power sharing, voltage balancing.

1. Introduction

DC microgrids are steadily attracting attention mainly due to the proliferation of power-electronic
loads and the dc nature of the majority of distributed energy resources (DERs) such as PV, fuel cell,
and wind generation systems. In fact, since microgrids are replete with dc loads, it is technically

∗
Corresponding author
Email address: keyhan.sheshyekani@polymtl.ca (Keyhan Sheshyekani )

Preprint submitted to Elsevier July 21, 2017

and economically preferred to feed these loads directly by dc DERs avoiding multiple conversion of
voltage from dc to ac and vice versa. Moreover, compared to ac microgrids, dc microgrids generally
provide a better power quality to the loads as they are not prone to reactive power imbalances
and harmonic distortions [1] - [4]. Furthermore, dc microgrids are simpler to control as they need
no frequency and reactive power regulations. As dc converters are easier to work in parallel, dc
microgrids offer a more modular and scalable system than ac microgrids which makes them more
flexible to expand [5].
Recently, bipolar dc microgrids have been proposed offering additional benefits compared to the
conventional dc microgrids. The bipolar dc microgrid can be considered as an effective solution to
the applications requiring a high level of reliability and efficiency. Although the concept of bipolar dc
microgrid, for the first time, has emerged in 2010 [6], there is not still adequate research being done
to address their associated challenges. Some of the challenges have been extensively investigated as
they are common between conventional dc microgrids and bipolar dc microgrids. Hence, the same
approach from conventional dc microgrids can be customized to be applied to bipolar dc microgrids.
The majority of studies on bipolar dc microgrid have focused on suggesting new converter
topologies; in [7], a combination of SEPIC and Cuk converters has been proposed for bipolar
dc microgrid. Some other topologies have been also introduced in [8] - [11]. In [12], a power
converter (current redistributor) has been proposed to improve the stability of bipolar dc microgrids.
However, operation and control of bipolar dc microgrid have not been investigated in detail. In
particular, an in-depth study on the challenges associated with power sharing (droop controller)
and voltage balancing is not available in literature. As known, in conventional dc microgrids, the
voltage regulation is deteriorated with increasing the droop gain. Although reducing the droop
gain improves the voltage regulation, it worsens the accuracy of power sharing [1]. This is also a
challenge in bipolar dc microgrids with even more complexities.
The objective of the proposed unified control system is twofold: i) voltage-balancing in place of
voltage balancer and ii) involving the VSI in the power sharing. Since the traditional power sharing
method (i.e., droop controller) deteriorates the voltage balancing, two different schemes namely
the dynamic reference and the equivalent droop are proposed for the power sharing purpose. These
schemes, while provide a good power sharing, do not sacrifice the voltage balancing of the microgrid.
This work is a continuation of the concept developed in [13].

2
 PV Panels Energy
Storage

Main AC
Grid DC DC
vabc DG#3 DG#1
DC DC
Interlinking Voltage Ig3 Ig1,u
Ivb,u Ig1,w
Converter Balancer
Line#3 Line#1
+Vdc
AC DC Upper Tr.
+
V
- m,u Vdc 2Vdc
0
C.B Pin Lower Tr.
+
DC DC -Vdc V
- m,w
Vdc

Line#4 Line#2
VSI VB Ivb,w
Ig2,u Ig2,w

DC
DG#2
DC

Energy
DC Load Storage

Figure 1: Topology of bipolar dc microgrid equipped with a central voltage balancer (VB), a traditional boost converter
(DG#3), and bipolar boost converters (DG#1, DG#2).

Bipolar DC
Traditional VSI DC Link Voltage Balancer
Microgrid

Main AC
Grid + +
vabc C1
rf Lf S1 S3 S5 S7 vu iu
rLm Lm rc1 -
v DC
+
C2
S4 S2 vw iw
S6 S8
- rc2
-

Figure 2: Topology of voltage balancer connected between a voltage-source inverter (VSI) and a bipolar dc microgrid.

2. Bipolar dc microgrid structure and its components

The topology of a bipolar dc microgrid is presented in Figure1. As it can be seen, the connections
between + Vdc and 0 is called upper terminal, whereas the connections between 0 and − Vdc is
considered as the lower terminal.The salient feature of this type of microgrid is the availability of
two voltage levels, including Vdc and 2Vdc . Furthermore, in case of loss of one wire, the microgrid
continues to supply the loads connected between two remaining wires. Furthermore, it is possible to
connect each of loads to both terminals; while one terminal feeds the load, the other is kept as the
spare terminal. Hence, if one terminal is lost, the load is supplied via other terminal. In addition, a
voltage of 2Vdc can be provided for heavy loads connected between the upper and lower terminals.
This reduces the ohmic power loss and improves the overall efficiency of the microgrid.
However, there are two certain challenges associated with bipolar dc microgrids: voltage bal-
ancing and power sharing. Both of challenges are due to the three-wire topology of bipolar dc

3
microgrids. Whenever the loads on the upper and lower terminals have different nominal values,
the voltages of the two terminals tend to drift from the rated value making the bipolar dc microgrid
unbalanced. Therefore, bipolar dc microgrid requires a central voltage balancer aiming to keep the
voltages of the upper and lower terminals equal, in particular, under unbalanced loading [6], [14].
The need for a voltage balancer is mainly due to the fact that the common wire (denoted by 0 in
Figure1) is shared among several power converters. Hence, the proper function of the power convert-
ers is ensured only if the zero potential of this wire is enforced by a voltage balancer. The structure
of the voltage balancer connected to a voltage-source inverter (VSI) is presented in Figure2. The
main duty of voltage balancer is to keep the voltages across C1 and C2 equal, particularly in case
the load currents of iu and iw are unequal.
Besides performing the voltage-balancing function, the voltage balancer has to participate in
power sharing in bipolar dc microgrid. It is a common practice that only voltage source converters,
interfacing ac or dc sources to the microgrid, participate in power sharing. However, it is possible to
consider the voltage balancer as a voltage-source converter through which the VSI is connected to the
microgrid. In this case, the dc link of the VSI works as a voltage-controlled dc source participating
in power sharing only through the voltage balancer. This means that the upper terminal voltage,
vu , and the lower terminal voltage, vw , should be regulated by two voltage control loops (see
Figure2). The voltage references of two control loops have to be modified by droop controllers to
fulfil the power-sharing requirements. However, the function of droop controllers adversely affects
the voltage-balancing performance. The adversary effects of droop controllers on the performance
of voltage balancing are fully explained in Section 3.

2.1. A unified control system for VSI and voltage balancer

As it is shown in Figure2, the voltage balancer is located between the dc-link of the VSI and
the terminals of the bipolar dc microgrid. The dc-link voltage, vDC , is equal to the sum of vu and
vw in steady-state conditions. These two voltages are controlled by the switches S7 and S8 while
the inductor Lm accounts for power exchange. The stored energy in C1 can be transferred to C2
and vice versa, hence, the voltage balancer is able to exchange power between the upper and lower
terminals of bipolar dc microgrid. In order to perform voltage regulation and balancing in bipolar
dc microgrid, three variables should be controlled, including vDC , vu , and vw . The VSI controls
vDC , whereas vu and vw are regulated by the voltage balancer. In fact, two variables, vDC and
vu , are controlled, which results in an indirect control of the third variable, vw . The proposed
control system of the voltage balancer and VSI is shown in Figure3. The voltage control loop of

4
 Gff(s) Vsq
ref
Qs
\ + Kq(s) + ref -2/(3Vsd)
mq - iq
dq0 LW 0
VDC /2 iq
to 0
mabc LW 0 id
abc m0
- -
\ + Kd(s) + ref
2/(3Vsd)
md id
Gff(s) Vsd
ref
ref
V DC Ps
(.)2 + -kv(s) +
- +
vDC 2
(.) PDC
VSI
0 Voltage Balancer
Vw - iw
Voltage
+ kw LPF Islanding
+
Restoration 0 Detection
Unit Vu I
+ ref + Tcu(s) d
-
Vu II +
S7
vu -
iu
LPF ku 0 + Tci(s) fs S8
-
|V u - Vw|

Figure 3: The proposed control system of the voltage balancer and the VSI.

Bipolar DC
VSI Microgrid

+
C1
S7 vu iu
vDC rLm Lm rc1 -
+
- +
C2
S8 vw iw
rc2
-

Figure 4: Equivalent circuit of the voltage balancer.

the balancer and VSI regulate, respectively, vu and vDC . The lower terminal voltage, vw , is then
regulated indirectly,

ref
VDC = vu + vw = (Vu0 − ku iu ) + (Vw0 − kw iw ) (1)

where ku and kw are the droop gains of, respectively, upper and lower terminals. No-load voltage
references of the upper and lower terminals are Vu0 and Vw0 , respectively. The VSI operates under
the dc-link voltage control mode. In this mode of operation, vDC is regulated by a voltage control
loop, which in fact, controls the stored energy of the dc-link capacitor. As electrical energy is
proportional to the square of voltage in capacitor, it is easier to regulate voltage using its square
value. The voltage control loop generates the active power reference of VSI, Psref , which is fed to

5
 Bipolar DC Bipolar DC
VSI Microgrid VSI Microgrid

+ +
S7 ic1 C1 S7 ic1 C1
vu iu vu iu
vDC rLm Lm rc1 -
vDC rLm Lm rc1 -
+ +
- + - +
iLm C2 iLm C2
S8 vw iw S8 vw iw
ic2 ic2
rc2 rc2
- -

(a) (b)

Figure 5: Graphic circuit models of the voltage balancer used for state-space averaging (a) state I, and (b) state II.

ref
the current controller in dq frame. vDC is generated by the control system of the voltage balancer.
In the grid-connected mode, the VSI exchanges power with the ac grid and maintains vDC at
ref
the vDC value. The ac-side voltage and frequency of VSI are imposed by the ac grid so it injects the
active power of Psref to the ac grid. The references for the inner current control loops, iref
d and iref
q ,

are generated, respectively, from Psref and Qref

s . Since it is desirable to inject only active power to

the main ac grid, Qref

s is set to zero. As the currents in dq-frame are decoupled, the proportional-
integral (PI) controllers Kd (s) and Kq (s) have equal coefficients. Vsd and Vsq are the dq-frame
voltages of the ac grid, feed forwarded to the current control loops to decouple the dynamics of the
ac grid from the control loops. The feed forward transfer function is assumed to be Gf f (s) = 1;
however, it may include low pass filters as well. The control system of VSI is well documented in
many references [15], [16]. The equivalent circuit of the voltage balancer in grid-connected mode is
shown in Figure4, where the VSI is replaced by a controlled voltage source, vDC , and the current
sources iu and iw are connected, respectively, to the upper and lower terminals of the balancer to
model the bipolar dc microgrid. The circuit has two states: (I) S7 is open while S8 is close, and (II)
S7 is close while S8 is open. The equivalent circuit for state (I) and (II) are given in Figure5(a) and
Figure5(b), respectively. Obtaining the state-space equations for two states, and averaging over one
switching cycle, the plant transfer functions of the voltage balancer can be derived. In state (I),

dx
= A1 x + B1 u (2)
dt

where
 0  0
x = vc1 vc2 iLm , u = iu iw vDC

6
  
−1/2Crc −1/2Crc 1/2C 
 
A1 = 
−1/2Crc −1/2Crc −1/2C 

 
−1/2Lm 1/2Lm (rc + 2rLm )/2Lm
 
 −1/2C 1/2C 1/2Crc 
 
B1 = 
 −1/2C −1/2C 1/2Crc 

 
rc /2Lm −rc /2Lm 1/2Lm
In the above mentioned matrices A1 and B1 , the values of the capacitors and the associated
equivalent series resistance are considered to be equal,

C1 = C2 = C and rc1 = rc2 = rc

In state (II):
dx
= A2 x + B2 u (3)
dt
where A2 and B2 are equal to A1 and B1 , respectively; except that B1 (3, 3) = − B2 (3, 3). This is
due to the symmetrical topology of the circuit and the assumption that the capacitors have equal
value. Averaging the state-space equations over one switching period and ignoring the second-order
terms [17], the small-signal equation is obtained as,

dx
b
=Ax b + {(A1 − A2 )X + (B1 − B2 )U } db
b+B u (4)
dt

where
A = D A1 + (1 − D) A2

B = D B1 + (1 − D) B2

In (4), D is the duty cycle of switch S8 , X is the matrix of steady-state values, and U is the
matrix of electric sources. The plant transfer functions relating state variables to d can then be
obtained as,
b = (s I − A)−1 {(A1 − A2 )X + (B1 − B2 )U } db
x (5)

Since the upper terminal voltage, vu , is regulated by d, as it is shown in Figure3, the plant
transfer function relating vu to d is derived from (5). Based on this transfer function, an integral-
double-lead compensator is designed in Section 4.

7
 iu

iL
vu

iL

vw

iw

(a) (b)

Figure 6: Bipolar, bidirectional dc/dc boost converter, (a) converter topology, and (b) switching pattern [13].

iL iL iL
iC1 vu iC1 vu iC1 vu

iC2 vw iC2 vw iC2 vw

(a) (b) (c)

Figure 7: Graphic circuit models of the bipolar boost converter used for state-space averaging (a) state I, (b) state
II, (c) state III.

2.2. Bipolar, bidirectional boost converter

As mentioned earlier, the bipolar dc microgrid cannot operate unless the voltage balancing-
function is fulfilled. So far, this duty has been performed by the voltage balancer. However, in
case this central device fails, the microgrid stops working. To obviate the extra dependency on the
voltage balancer, a bipolar boost dc/dc converter is added to the microgrid (DG#1 and DG#2 in
Figure1). By doing this, as it is comprehensively described in [13], the voltage-balancing function
is performed in a decentralized manner. Therefore, the reliability of the microgrid is increased.
Since a detailed discussion about the operation and switching algorithm of bipolar boost converter
has been already given in [13], we briefly discuss its small-signal model and voltage control system.
The topology of the converter and switching algorithm are, respectively, given in Figure6(a) and
Figure6(b).
Adopting the state-space averaging method, the small-signal model of the converter is obtained,
from which the plant transfer functions are derived. The converter has three states associated with
the time periods ts , tb , and to as are shown in Figure6(b). The equivalent circuit of each state is
given in Figure7. It should be noted that the inductor Ls plays its role only in the balancing state

8
 Voltage Restoration
Unit
0 ref
Vu Vu d2
+ + Tcu(s) +
S22
- - -
Ku vu S12
fs
LPF iu
0 ref
Vw Vw d1
+ + Tcw(s) +
S11
- - -
Kw vw fs
S21

LPF iw

Figure 8: Voltage control loops of bipolar boost converter.

(tb ). The roles of diodes Ds1 and Ds2 are only to provide an alternative path for discharging the
energy of Ls . Since the bipolar boost has three states, three sets of state-space equations should be
derived and averaged over one switching cycle,

dxb c1 +
=Ax b+B u b + {(A12 − A2 )X + (B12 − B2 )U } d
dt
c2
{(A12 + A1 )X + (B12 + B1 )U } d (6)

where
 0
x = iL vc1 vc2 , u = Vi

A = D2 A1 + D1 A12 − (1 − D2 ) A12 + (1 − D1 ) A2

B = D2 B1 + D1 B12 − (1 − D2 ) B12 + (1 − D1 ) B2

where, referring to Figure6(b), D1 and D2 are the duty cycles of S11 and S22 , respectively. A1 and
B1 are, respectively, state and input matrices associated with ts ; A12 and B12 are associated with
tb ; and finally, A2 and B2 are associated with to . Consequently, vu and vw can be regulated by
two independent voltage control loops, as shown in Figure8. Based on the plant transfer functions
derived in 6, the compensators Tcu (s) and Tcw (s) are calculated in Section 4.

3. The proposed Power sharing schemes

Various power-sharing schemes for conventional dc microgrids have been proposed in the lit-
erature, which are categorized as central control (also known as hierarchical control), distributed
control with communication network, and local control without communication network [2], [18].
Although using central controller provides the most precise power sharing, it requires network in-
frastructure and additional investment, which might not be readily available in many dc microgrids

9
 Voltage
(V)
Vhi
V0
Vlo
Current
(A)
-I I

Figure 9: Droop characteristics employed by voltage control loops.

[18]. Moreover, a control system may achieve precise power sharing and acceptable voltage regula-
tion in conventional dc grid, but it may not have satisfactory results in bipolar dc microgrid. This
is mainly due to the three-wire topology of bipolar grid and certain converter topologies required to
provide voltage-balancing function. For example, in a conventional dc grid with rated voltage of 400
V, the allowable voltage may range from 380 to 420 V. However, this voltage variation may produce
undesirable results in bipolar dc microgrid; in a plausible scenario, the upper terminal voltage can
reach 420 V, while the lower terminal voltage falls to 380 V. The voltage difference of 40 V between
upper and lower terminals causes serious damage to some equipment.

3.1. Constant Reference without communication network

Since the voltages of the upper and lower terminals of the voltage balancer and the bipolar
boost converter can be regulated by two independent voltage control loops, two individual droop
controllers are used to modify the reference voltages of the upper and lower terminals. The conven-
tional droop characteristics used to modify the voltage reference of the upper (or lower) terminal is
shown in Figure9, from which the droop gains can be obtained,

Vhi − Vlo
ku = kw = (7)
2I

where Vhi is the upper limit for the microgrid voltage, while the lower limit is Vlo . I is the full-load
current of each terminal of each converter. The droop gains ku and kw will then be used to calculate
voltage references for the control systems of the voltage balancer and bipolar boost converters, as
they are shown in Figure3 and Figure8,

Vuref = Vu0 − ku × iu (8)

Vwref = Vw0 − kw × iw (9)

Assuming the voltage restoration units, shown in Figure3 and Figure8, generate constant refer-
ences, Vu0 and Vw0 , the problem raises when the upper terminal current, iu , is negative and the lower
terminal current, iw , is positive, and vice versa. In other words, the worst-case scenario happens

10
 (V) (V) (V) (V)
0 0 0 0
Vu 0 Vu 0 Vu 0 Vu 0
Vw Vw Vw Vw

+ΔV ku -ΔV ku -ΔV ku +ΔV ku

kw kw kw kw
(A) (A) (A) (A)
iw iu iu iw iu iw iw iu

(a) (b) (c) (d)

Figure 10: Droop characteristics in the dynamic reference scheme, (a) iu is positive and greater than iw resulting in
positive ∆V , (b) iu is positive and smaller than iw resulting in negative ∆V , (c) iu is negative while iw is positive
resulting in negative ∆V , and (d) iu is positive while iw is negative resulting in positive ∆V .

when one terminal is absorbing full-load current, while the other is generating full-load current.
This may happen when the generated power in one terminal of bipolar microgrid exceeds the load
demand. This challenge has been also obserevd in [13]. In worst-case scenario, iu and iw are at
maximum value (full load) but one of them is negative,

if iu = − If l and iw = + If l

k = ku = kw

|∆V | = |Vuref − Vwref | = 2 k If l (10)

For typical values of k = 2 ohm and If l = 10 A, the voltage difference between upper and lower
terminals, |∆V |, becomes 40 V, which is unacceptable. This implies that the traditional power
sharing based on the constant reference approach suffers from poor voltage regulation although it
provides almost accurate power sharing among converters.

3.2. Dynamic Reference using communication network

As discussed earlier, challenges of power sharing and voltage regulation are more challenging
in bipolar dc microgrid compared to conventional dc microgrid. Hence, the use of communication
network in bipolar dc microgrids seems to be inevitable. To achieve an accurate power sharing and
precise voltage regulation, various communication networks have been investigated in the literature.
Some communication networks require infrastructure and additional investment, such as the local
area network and Internet. A very suitable network for distributed communication within microgrid
is controller area network (CAN) whose viability has been studied in [18] and [19]. Hence, CAN
bus is used here to share the output currents and voltages of DGs within the microgrid. In order
to improve the voltage balancing performance (and consequently voltage regulation), the following

11
 Voltage Restoration Unit
0
Droop V
Gain
0
iu LPF + 1/2 k + Vu
-
0
iw LPF -1 + Vw

0
V

Figure 11: Voltage restoration unit.

algorithm can be adopted locally in the voltage restoration units where the values of Vu0 and Vw0 are
modified continuously,
∆V = 0

Vu0 − Vw0 = k (iu − iw )

k (iu − iw )
Vu0 = Vu0 + (11)
2
k (iu − iw )
Vw0 = Vw0 − (12)
2
Half of ∆V is added to Vu0 and subtracted from Vw0 . However, it must be noted that the sign
of ∆V is determined based on the signs and magnitudes of iu and iw . Some possible scenarios are
shown in Figure10. For example, referring to Figure10(a), while iu is greater than iw and both are
positive, the voltage difference ∆V becomes positive. Therefore, the voltage restoration unit shown
in Figure11 dynamically sets the values of Vu0 and Vw0 as they are given in (11) and (12), respectively.
By doing this, the droop characteristics with slope ku is shifted upward while the one with slope
kw is shifted downward. The arrows in Figure10 indicate the direction of displacement of the droop
characteristics. In case iu is smaller than iw , as indicated in Figure10(b), ∆V becomes negative,
so the direction of displacement is reversed as compared to Figure10(a). The same interpretation
applies to Figure10(c,d), where one terminal absorbs current and the other generates current. This
is the worst case happening in bipolar dc microgrids, since the voltage difference ∆V can reach to
an unacceptable large value. If the reference values, Vu0 and Vw0 , are modified locally by each DG,
the displacement of droop characteristics becomes unequal, which worsens the accuracy of power
sharing.
Therefore, the values of Vu0 and Vw0 generated by the voltage restoration unit of the voltage
balancer are transmitted to other DGs in microgrid. In CAN network, the voltage balancer acts as
the master converter while other DGs are slaves. Hence, the voltage restoration units of all DGs are
disabled and the references for their droop controllers are sent by the voltage balancer, which are

12
 Vt

ku1 Ig1 ku2 Ig2 ku3 Ig3 ... Ic1 Ic2 IL

...
0 0 0
Vu +- Vu +- Vu +- ...

Figure 12: Equivalent circuit of bipolar dc microgrid (upper terminal).

received by DGs over every control period. However, only DGs participating in voltage regulation
of the microgrid collect the transmitted voltage references (i.e. DG#1 and DG#2).

3.3. Equivalent Droop using communication network

Ignoring line impedances, the upper (or lower) terminal of a typical bipolar dc microgrid can
be represented by the equivalent circuit shown in Figure12. The DGs participating in voltage
regulation and voltage balancing can be modelled via a series connection between a voltage source
with nominal value of Vu0 and a resistance with the value of droop gain, Ku [18]. Other DGs operating
under current-control mode are modelled by current sources Ic . For the ease of calculation, total
load current is also represented by current source IL . Since the voltage-source DGs are assumed to
have equal voltage references, Vu0 , and the equivalent circuit is supposed to work at its steady-state
condition, the following equations can be simply derived,
m1
X m2
X
IT = Ign = IL − Icn
n=1 n=1
Pm1
n=1 Ign
∆Vu = Vu0 − Vt = (13)
KT
where,
1
KT = 1 1 1
Ku1 + Ku2 + Ku3 + ···
KT may be considered as the equivalent droop of the microgrid. As a result, all voltage-source
DGs shown in Figure12 can be represented by only one equivalent DG with reference voltage Vu0
and equivalent droop gain KT , as it is shown in Figure13(a).
∆Vu is calculated locally by each DG. Then, the droop characteristics of the upper terminals are
shifted downward or upward by the value of ∆Vu . The same calculation can be done to obtain ∆Vw
in the lower terminals of DGs. To calculate ∆Vu , each voltage-source DG should share its upper
terminal current on CAN bus, while receive the upper terminal currents of other voltage-source
DGs. Then, the total net currents of voltage-source DGs (negative sign for absorbing current and

13
 Vt (V)
0
Vu
KT IT Ic1 Ic2 Vt KT
... IL
0
Vu + (A)
-
IT
(a) (b)

Figure 13: Equivalent circuit of bipolar dc microgrid (upper terminal), (a) replacement of voltage-source DGs with
equivalent DG, and (b) equivalent droop characteristic.

positive for injecting) is computed locally by each DG. From (13) the value of ∆Vu is calculated and
added to Vu0 . The voltage references in DGs are updated periodically to ensure a proper voltage
regulation and power sharing. The droop equations (8) and (9) are periodically updated to,

Vuref = Vu0 + ∆Vu − ku × iu (14)

Vwref = Vw0 + ∆Vw − kw × iw (15)

Since the calculation results of ∆Vu and ∆Vw by each DG have the same values, the droop char-
acteristics of all DGs are shifted upward or downward equally. Therefore, this scheme would never
negatively affect the accuracy of power sharing while it provides a very good voltage regulation. It
is worth mentioning that in order to reduce overshoots and prevent voltage and current oscillations,
the calculated references (14) and (15) are passed through a simple smoothening function, known
as ”sigmodial function”, to smoothen the step changes.

4. Simulation results

To investigate the effectiveness of the proposed power-sharing schemes, the bipolar dc microgrid
shown in Figure1 is simulated in MATLAB software. The main ac grid is connected to the bipolar
dc microgrid through a traditional VSI and a voltage balancer. The main ac grid acts as an infinite
bus maintaining the ac voltage and frequency. The VSI operates in dc-link voltage control mode,
so it maintains vDC at the input terminal of the voltage balancer. DG#3 is a PV system connected
to the microgrid by means of a conventional boost converter operating in maximum power point
tracking mode. The PV system has the rated voltage of 200 V and generates a maximum power of
10 kW. However, it must be noted that DG#3 injects current (Ig3 ) to the upper terminal only. The
output current of DG#3 , Ig3 , along with the loading profiles of the upper and lower terminals are

14
 Table 1: Specifications of the bipolar dc microgrid.
Item Specifications

DG#1, #2 Vin =200 V, Vout =2 × 400 V, Pout =2 × 4 kW

Lm =474 µH, rL =0.2 Ω, C=50 µF, rC =0.05 Ω,
Ls =40 µH, fs =100 kHz
DG#3 Vin =200 V, Pmax =10 kW
VSI Pout =16 kW, vDC =800 V,
Lf =1 mH, rf =0.1 Ω, dclink=5 mF, fs =4 kHz,
VB Lm =400 µH, rLm =0.1 Ω, C=10 mF, rc =0.1 Ω
Pout =2 × 8 kW, vu =vw =400 V, fs =40 kHz
AC Grid vL−L = 380V rms, f = 60Hz
Line#1 R=0.07 Ω, L=70 µH
Line#2 R=0.05 Ω, L=50 µH
Line#3 R=0.15 Ω, L=170 µH
Line#4 R=0.1 Ω, L=100 µH

shown in Figure14. ILu and ILw are the load currents of the upper and lower terminals, respectively.
DG#1 and DG#2 are battery storage systems which are connected to the microgrid through the
bipolar boost converters presented in Figure6(a). Each battery system is capable of supplying a
maximum of 8 kW to the microgrid. The detailed specification of DGs, VSI, voltage balancer, ac
grid, and line impedances are given in Table (1).
Based on the information given in Table (1), the control system parameters are derived. In
Figure3, the current controllers of VSI, kq (s) and kd (s), have the same coefficients. The reference
ref
signal for the dc-link voltage control system,VDC , is generated by the voltage balancer. The voltage
compensator, kv (s), is a lead-lag controller which is designed based on the VSI transfer function
relating Psref to VDC
2 . The current controllers and the dc-link voltage compensator are described

as, [15]
5 × 10−3 s + 53.69
kd (s) = kq (s) = 0.5 + , kv (s) = 0.8 (16)
s s + 745.2
The step-by-step design procedure is available in [15]. Referring to the operating point given in
Table (1), the plant transfer function of the voltage balancer is derived from (5),

vu (s) 4 × 107
Tp (s) = = 2 (17)
d(s) s + 250 s + 5 × 104

Having the plant transfer function of (17), an integral-double-lead compensator, Tcu (s), is designed,

3.54 × 103 s2 + 9 × 105 s + 5.73 × 107

Tcu (s) = (18)
s3 + 1.55 × 103 s2 + 6 × 105 s

15
Using (6), the plant transfer functions of the bipolar boost converter are calculated,

vw (s) vu (s) 5.48 × 104 s2 + 8.5 × 109 s + 1.04 × 1013

Tpw (s) = = Tpu (s) = = 3 (19)
d1 (s) d2 (s) s + 2.01 × 103 s2 + 2.22 × 107 s + 1.33 × 1010

Then, an integral-double-lead controller, shown in Figure8, is derived,

kp (s + ωz )2
Tcu (s) = Tcw (s) = (20)
s (s + ωp )2

Making reference to the operating point values associated with DG#1 in Table (1), the coefficients
of the voltage compensators, at the loop-gain crossover frequency of 4.43 × 103 rad/sec, are as
follows: kp = 89.4 × 103 , ωz = 986.8 rad/sec, and ωp = 19.8 × 103 rad/sec. The first-order lowpass
filters (LPFs), used in the control systems to attenuate the ac component of converter current, have
the cutoff frequency of 300 rad/sec. At this cutoff frequency, the LPF attenuates the input power
by 3dB. To have the cutoff frequency of 300 rad/sec, the time constant of LPF, T should be about
3msec,
1
LP F (s) = (21)
T s+1
The upper and lower terminals of the microgrid have the nominal voltage of 400 V and the voltage
regulation is designed to confine voltage deviation to ±5 percent. Thus, the upper and lower voltage
limits, Vhi and Vlo , become 420 and 380 V, respectively. Referring to (7), the droop gains of the
voltage balancer and the bipolar boost can be calculated; since each terminal of voltage balancer
has the full-load current of 20 A, a droop gain of 1 Ω is obtained. The bipolar boost has the full-load
current of 10 A in each terminal, so the droop gain is 2 Ω.
Before evaluating the performance of power sharing and voltage regulation in the subsequent
sections, it is useful to study the impact of LPFs on the dynamics of the control systems. As shown
in Figure3 and Figure8, LPFs are widely used in the control systems of the voltage balancer and
bipolar boost converter. Since the plant transfer functions and compensators are already derived, the
dynamics of control systems which include LPFs can be studied using step response and frequency
response. Referring to Figure8 and Figure6(a), the output voltages of the bipolar boost converter,
vu and vw , are regulated by two independent voltage control loops. Hence, the effects of LPF on
each voltage control loop can be investigated independently. For example, the complete small-signal
model of the voltage control loop which regulates vu is presented in Figure15. The same small-signal
model can be derived for the control loop of vw as well. The step response and frequency response
of the close-loop control system are presented, respectively, in Figure16(a) and Figure16(b). As it
can be seen, for various values of filter time constant, T , the impact of LPF on the dynamics of

16
 30 Ig3
ILu
25

Current [A]
ILw
20
15
10
5
0 0.4 0.5 0.6 0.7 0.8 0.9 1 1.1 1.2
Time [second]

Figure 14: The output current of DG#3 and the loading profiles of the upper and lower terminals.
0 ref
Vu Vu d2 vu iu
+ + Tcu(s) Tpu(s) 1/R1
- -

Ku LPF

Figure 15: The small-signal model of the voltage control loop which regulates the upper terminal voltage, vu , of the
bipolar boost converter

the close-loop control system is trivial. From T = 0 msec (no LPF) to T = 10 msec, the step
and frequency responses of the close-loop control system do not vary remarkably. However, when
T tends to the higher values, the step response becomes slower showing an overshoot. The impact
of LPF on the dynamics of the voltage balancer is also negligible.

4.1. Scenario 1: constant reference

The voltage references, Vu0 and Vw0 , in droop controllers of the voltage balancer, DG#1, and
DG#2 are considered to be constant and equal to 400 V. The upper terminal currents of the voltage
balancer, DG#1, DG#2, and DG#3 are, respectively, Ivb,u , Ig1,u , Ig2,u , and Ig3 . Gradually, the
output current of the PV system is increased (Ig3 in Figure14) until it exceeds the upper terminal
load current, ILu , and its surplus is then transferred to the lower terminal by means of the voltage
balancer and the bipolar boost converters. The currents of the upper and lower terminals are shown,
respectively, in Figure17(a) and Figure17(b). As it can be seen, Ig1,u and Ig2,u are approximately
equal to each other since DG#1 and DG#2 have the same droop characteristic and equal nominal
power, and so are Ig1,w and Ig2,w . Moreover, as the droop gain of the voltage balancer is half of the
that of DG#1 and DG#2, it injects or absorbs current twice as much as the current of DG#1 or
DG#2. From t = 0.6 sec, Ivb,u , Ig1,u , and Ig2,u become negative since Ig3 exceeds ILu . The surplus
of Ig3 is then transferred to the lower terminal, which leads to a reduction in power generation of
the battery systems DG#1 and DG#2.

17
 1

T=10ms T=5ms
0.5
1
T=3ms
T=0ms
0
0 0.005 0.01 0.015 0.02 0.025 0.03
Time [second]
(a)

40
Magnitude (dB)

20 T=10, 5, 3ms
0
-20 T=0ms
-40 1 2 3 4 5 6
10 10 10 10 10 10
Frequency (rad/s)
(b)

Figure 16: The impact of LPF on the control system of bipolar boost converter, (a) step response, and (b) frequency
response.

An almost accurate power sharing is achieved using constant reference approach. For example, at
t = 0.9 sec, when the effective load current as seen by voltage-source DGs is (ILu − Ig3 = −12.5 A)
and ILw = 20 A, the ideal power sharing requires Ivb,u , Ig1,u , and Ig2,u to absorb, respectively,
− 6.26 A, − 3.12 A, and − 3.12 A. Referring to Figure17(a), the mentioned currents have approx-
imate values of − 5.34 A, − 3.34 A, and − 3.82 A, which is almost an acceptable accuracy. The
lower terminal currents, presented in Figure17(b), have the same power-sharing accuracy. However,
a perfectly accurate power sharing is not fulfilled due to the negative effects of line impedances.
Although voltage-source DGs perfectly balance their terminal voltages, the negative effects of droop
controllers, as discussed thoroughly in the preceding section, cause an unacceptable voltage differ-
ence between voltages of the upper and lower terminals. The upper and lower terminal voltages of
the balancer, vu and vw , are shown in Figure18. As it can be seen, the voltage difference becomes
greater as Ig3 increases. The reason is discussed and formulated in (10). Due to the voltage drop
over the line impedances, the deterioration of voltage regulation is more pronounced for main busbar
voltages, Vm,u and Vm,w , as they are shown in Figure19. At t = 1.1 sec, for instance, the voltage
difference becomes about 30 V. In bipolar dc microgrids with higher current level and greater line

18
 4 Ivb,u
2 Ig1,u
0 Ig2,u

Curent [A]
-2
-4
-6
-8
-10 0.4 0.5 0.6 0.7 0.8 0.9 1 1.1 1.2
Time [seconds]

(a)

12 Ivb,w
10 Ig1,w
Ig2,w
Current [A]

8
6
4
2
0 0.4 0.5 0.6 0.7 0.8 0.9 1 1.1 1.2
Time [second]
(b)

Figure 17: The output currents of voltage balancer, DG#1, and DG#2 in constant reference scheme, (a) upper
terminal currents, and (b) lower terminal currents.

vu
415 vw
410
Voltage [V]

405
400
395
390
385
380 0.4 0.5 0.6 0.7 0.8 0.9 1 1.1 1.2
Time [second]

Figure 18: The terminal voltages of the voltage balancer while references are constant.

impedances, the voltage difference could reach to even higher values. Therefore, although the con-
stant reference scheme provides almost accurate power sharing, it has a negative impact on the
voltage regulation (voltage balancing).

4.2. Scenario 2: dynamic reference using CAN network

Voltage regulation (voltage balancing) is improved using dynamic reference scheme in which the
values of Vu0 and Vw0 are modified continuously by the voltage restoration unit as it is formulated in
(11) and (12). As mentioned earlier, it is only the duty of the voltage balancer (master converter) to
generate Vu0 and Vw0 which are transmitted to DG#1 and DG#2 through high-speed CAN network.
In fact, Vu0 and Vw0 are updated over every control period (order of tens of microseconds) utilizing

19
 420
415 vm,u
vm,w
410

Voltage [V]
405
400 ΔV
395
390
385
380
0.4 0.5 0.6 0.7 0.8 0.9 1 1.1 1.2
Time [second]

Figure 19: Main busbar voltages using constant reference scheme.

high-speed CAN network.

The voltage references Vu0 and Vw0 associated with the voltage balancer are shown in Figure20(a),
resulting in terminal voltages shown in Figure20(b). A comparison between the voltage regulation
performances of the constant reference shown in Figure18 and the dynamic reference shown in
Figure20(b) reveals the effectiveness of the dynamic reference in providing a good voltage regulation.
In fact, Vu0 and Vw0 are modified in such a way that the voltage difference, ∆V , becomes zero.
However, voltage drop over line impedances causes a minor fluctuation in main busbar voltages,
Vm,u and Vm,w , which are presented in Figure21. Compared to the constant reference scheme where
∆V reaches to about 30 V at t = 1.1 sec, this value is limited to 10 V in the dynamic reference
scheme.
The currents of upper and lower terminals, as are shown, receptively, in Figure22(a) and
Figure22(b), reveal an accurate power sharing. In fact, power sharing accuracy is similar to the
constant reference scheme (see Figure17(a) and Figure22(a)). For the lower terminal currents, the
same comparison can be made. These observations confirm the simultaneous fulfillment of both
voltage regulation and power sharing by the proposed control system.

4.3. Scenario 3: equivalent droop using CAN network

As it is described earlier, the output currents of the balancer, DG#1, and DG#2 are transmitted
over CAN bus. Following to the calculation of IT and KT , the values of ∆Vu and ∆Vw are computed
locally. Then, they are added to Vu0 and Vw0 according to (14) and (15), respectively. As it is shown
in Figure23, Vu0 and Vw0 are updated every 0.4 sec, so a low-speed low-cost CAN network can
be implemented. Since the calculations of IT and KT , performed by each DG, have the same
results, the voltage references produced by the voltage restoration units of DG#1, DG#2, and
voltage balancer have the same waveforms as shown in Figure23. As the equivalent droop scheme is
activated every 0.4 sec, we double the simulation time (from 1.2 sec to 2.4 sec) in order to provide

20
 410 0
Vu
0
405 Vw

Voltage [V]
400

395

390 0.4 0.5 0.6 0.7 0.8 0.9 1 1.1 1.2

Time [second]

(a)

402 vu
vw
401
Voltage [V]

400
399
398
397
396
395
0.4 0.5 0.6 0.7 0.8 0.9 1 1.1 1.2
Time [second]

(b)

Figure 20: The terminal voltages of the voltage balancer utilizing dynamic reference scheme, (a) voltage references
generated by the voltage restoration unit, and (b) the voltages of the upper and lower terminals.

415 Vm,u
Vm,w
410
Voltage [V]

405
400 ΔV
395
390
385
380
0.4 0.5 0.6 0.7 0.8 0.9 1 1.1 1.2
Time [second]

Figure 21: Main busbar voltages using dynamic reference scheme.

a clearer illustration. The timing of the load profile shown in Figure14 is also doubled. The main
busbar voltages are shown in Figure24. At t = 0.6, 1, 1.4, 1.8, 2.2 sec voltage references are
updated. The biggest voltage difference happens at t = 2.2 sec; however, it is limited to only 4
V, which is very satisfactory as compared with 30 V in the constant reference scheme and 10 V in
the dynamic reference scheme. The equivalent droop provides an accurate power sharing too. The
currents of the upper and lower terminals are shown in Figure25(a) and Figure25(b), respectively,
which are as accurate as those of constant reference scheme.
It is important to notice that before and after updating voltage references, the output currents
of DGs do not change, which means voltage regulation does not adversely affect the power sharing
accuracy. Equivalent droop equipped with CAN communication network provides both accurate

21
 4
2 Ivb,u
Ig1,u
0 Ig2,u

Current [A]
-2
-4
-6
-8
-10 1
0.4 0.5 0.6 0.7 0.8 0.9 1.1 1.2
Time [second]

(a)

10
9 Ivb,w
8 Ig1,w
Ig2,w
Current [A]

7
6
5
4
3
2
1
0
0.4 0.5 0.6 0.7 0.8 0.9 1 1.1 1.2
Time [second]
(b)

Figure 22: The output currents of voltage balancer, DG#1, and DG#2 using dynamic reference scheme, (a) upper
terminal currents, and (b) lower terminal currents.

412
0
408 Vu
Voltage [V]

0
Vw
404
400
396
392
388
0.6 0.8 1 1.2 1.4 1.6 1.8 2 2.2 2.4
Time [second]

Figure 23: Reference voltages generated by the voltage restoration units in equivalent droop scheme.

410
Load changes
408 vm,u
vm,w
406
404
Voltage [V]

402
400 ΔV
398
396
394
392 Update voltage references
390
0.6 0.8 1 1.2 1.4 1.6 1.8 2 2.2 2.4
Time [second]

Figure 24: Main busbar voltages using equivalent droop scheme.

22
 6
Ivb,u
4 Ig1,u
2 Ig2,u
Current [A]

0
-2
-4
-6
-8
-10
0.6 0.8 1 1.2 1.4 1.6 1.8 2 2.2 2.4
Time [second]

(a)

10 Ivb,w
Ig1,w
Ig2,w
8
Current [A]

0
0.6 0.8 1 1.2 1.4 1.6 1.8 2 2.2 2.4
Time [second]

(b)

Figure 25: The output currents of voltage balancer, DG#1, and DG#2 using equivalent droop scheme, (a) upper
terminal currents, and (b) lower terminal currents.

Table 2: Comparison of power-sharing schemes

Scheme Power- Voltage Max. ∆V
sharing regulation
accuracy
Constant reference high very poor 30 V
Dynamic reference high good 10 V
Equivalent droop high very good 4V

23
power sharing and very good voltage regulation. Compared to the dynamic reference approach,
which employs a master-slave scheme with high-speed CAN network, the equivalent droop scheme
provides a more modular control system utilizing a low-speed low-cost CAN network. A quick
comparison of power-sharing schemes is provided in Table (2). With regard to the voltage and
current waveforms obtained from three power-sharing schemes, all oscillations damp out within 0.1
sec under the worst conditions. There are also some switching transient spikes which appear on the
voltage and current waveforms. These transient spikes are due to the load changes; however, the
overshoot and spikes on the voltage waveforms remain within ±10 percent of the nominal voltage.
It is worth mentioning that the equivalent droop scheme can be also utilized in islanded mode
of operation. In this mode, the VSI is disconnected from the main ac grid; however, it is possible
to keep the voltage balancer acivated in the microgrid to assist balancing the voltages of the upper
and lower terminals. When the VSI is disconnected, the dc-link voltage, vDC , becomes unavailable
in the input terminal of the balancer. Hence, the balancer can not act as a voltage-source converter,
which means it cannot participate in power sharing. However, the balancer can still perform voltage
balancing duty. To that aim, the control system of the balancer is switched to islanded mode of
operation, which is presented in Figure3. When the islanded mode is detected and the VSI is
isolated from the main ac grid, the control system of balancer is switched to islanded mode and
the difference between the upper and lower terminal voltages is forced to zero. Since the equivalent
circuit of the voltage balancer shown in Figure4 is no longer valid in islanded mode of operation, we
need to perform the state-space averaging analysis on the modified equivalent circuit to derive the
relevant transfer functions, and subsequently, design the compensator Tci (s). However, we do not
discuss this step as the approach is quite similar to the grid-connected mode except that the vDC is
replaced with the dc-link capacitor and the current sources iu and iw are replaced, respectively, by
voltage sources vu and vw in Figure4. It must be noted that the bipolar dc microgrid can effectively
operate in islanded mode even without voltage balancer. As it is discussed in [13], the bipolar
boost converters (DG#1 and DG#2) are designed to ensure the proper operation of the microgrid
in islanded mode.

5. conclusion

In this paper, the challenges of voltage balancing and power sharing in bipolar dc microgrid have
been investigated. A unified control system for VSI and voltage balancer was proposed where the
VSI regulates its dc-link voltage based on the reference signal generated by the voltage balancer. It

24
was shown that the traditional droop controller using constant reference scheme would significantly
worsen the voltage balancing performance. In particular, it causes a noticeable voltage difference
between upper and lower terminals. Two power sharing schemes were proposed which improve
the voltage balancing and voltage regulation while maintaining the accuracy of power sharing.
In the dynamic reference scheme, voltage references are generated by the voltage balancer and
transmitted to other DGs over every control period using a high-speed master-slave CAN bus. The
restoration unit of the voltage balancer dynamically modifies voltage references to achieve zero
voltage difference between upper and lower terminals. However, the voltage regulation was found
to be further improved by utilizing equivalent droop scheme, where the voltage references were
calculated locally by DGs and the voltage balancer using the data available in the CAN bus. The
simulation results revealed that the proposed power sharing schemes have the same accuracy as the
constant reference scheme, while they show a better performance in terms of voltage regulation and
voltage balancing.

6. Acknowledgements

This work was supported by Iran National Science Foundation under the ”Grant 93034332”.

References

[1] S. Anand, B. G. Fernandes, and J. M. Guerrero, Distributed control to ensure proportional

load sharing and improve voltage regulation in low-voltage dc microgrids, IEEE Trans. Power
Electron., vol. 28, pp. 1900-1912, April 2013.

[2] J. M. Guerrero, J. C. Vasquez, J. Matas, L. G. de Vicuna, and M. Castilla, Hierarchical control

of droop-controlled ac and dc microgrids: A general approach toward standardization, IEEE
Trans. Ind. Electron., vol. 58, pp. 158-172, January 2011.

[3] N. Yang, D. Paire, F. Gao, A. Miraoui, and W. Liu, Compensation of droop control using
common load condition in DC microgrids to improve voltage regulation and load sharing, Int
J Electr Power Energy Syst., vol. 64, pp. 752-760, January 2015.

[4] A. Bracale, P. Caramia, G. Carpinelli, E. Mancini, and F. Mottola,Optimal control strategy of

a DC micro grid, Int J Electr Power Energy Syst., vol. 67, pp. 25-38, May 2015.

25
 [5] M. Hamzeh, M. Ghafouri, H. Karimi, K. Sheshyekani, and J. Guerrero, Power Oscilla-
tions Damping in DC Microgrids, IEEE Trans. Energy Conversion, March 2016, DOI:
10.1109/TEC.2016.2542266

[6] H. Kakigano, Y. Miura, and T. Ise, Low-voltage bipolar-type dc microgrid for super high
quality distribution, IEEE Trans. Power Electron., vol. 25, pp. 3066-3075, December 2010.

[7] M. B. Ferrera, S. P. Litran, E. Duran Aranda, and J. M. Andujar Marquez, A converter for
bipolar DC link based on SEPIC-Cuk combination, IEEE Trans. Power Electron., vol. 30, pp.
6483-6487, December 2015.

[8] J. Lago, J. Moia, and M. L. Heldwein, Evaluation of power converters to implement bipolar
DC active distribution networksdc-dc converters, in Proc. IEEE Energy Convers. Congr. Expo.,
Phoenix, AZ, September 2011, pp. 985-990.

[9] P. Prajof and V. Agarwal, Novel boost-SEPIC type interleaved dc-dc converter for low-voltage
bipolar DC microgrid-tied solar pv applications, in Proc. IEEE 42nd Photovoltaic Specialist.
Conf., New Orleans, LA, June 2015, pp. 1-6.

[10] W. M. dos Santos, T. A. Pereira, C. Knaesel, H. R. Mamede, and D. C. Martins, Modeling

and control of the new DC-DC step-up converter to bipolar DC microgrid, in Proc. IEEE 36th
Telecommunications Energy. Conf., New Orleans, LA, Oct 2014, pp. 1-8.

[11] H. R. Mamede, W. M. dos Santos, and D. C. Martins, A new DC-DC power converter derived
from the TAB for bipolar DC microgrids, in Proc. IEEE Energy Conversion (ECCE) Conf.,
Montreal, QC, Sep 2015, pp. 6217-6222.

[12] J. Lago and M. L. Heldwein, Operation and control-oriented modeling of a power converter for
current balancing and stability improvement of DC active distribution networks, IEEE Trans.
Power Electron., vol. 26, pp. 877-885, January 2011.

[13] S. D. Tavakoli, J. Khajesalehi, M. Hamzeh, and K. Sheshyekani, Decentralised voltage balancing

in bipolar dc microgrids equipped with trans-z-source interlinking converter, IET, Renewable
Power Generation, vol. 10, pp. 703-712, April 2016, DOI: 10.1049/iet-rpg.2015.0222.

[14] X. Zhang, C. Gong, and Z. Yao, Three-Level DC Converter for Balancing DC 800-V Voltage,
IEEE Trans. Power Electron., vol. 30, pp. 3499-3507, August 2014.

26
[15] A. Yazdani and R. Iravani, Voltage-Sourced Converters in Power Systems : Modeling, Control,
and Applications. Wiley-IEEE Press, 2010, ch. 7.

[16] YA. Yazdani and R. Iravani, ”An accurate model for the DC-side voltage control of the neutral
point diode clamped converter”, IEEE Trans. Power Delivery, vol. 21, no. 1, pp. 185-193, Jan
2006

[17] R. W. Erickson and D. Maksimovic, Fundamentals of Power Electronics. Springer, 2012, ch.
8,9.

[18] S. Anand, B. G. Fernandes, and J. Guerrero, Distributed control to ensure proportional load
sharing and improve voltage regulation in low-voltage DC microgrids, IEEE Trans. Power
Electron., vol. 28, pp. 1900-1913, April 2013.

[19] Y. Zhang and H. Ma, ”Theoretical and experimental investigation of networked control for
parallel operation of inverters”, IEEE Trans. Ind. Electron., vol. 59, no. 4, pp. 1961-1970, 2012

27