ESPARZA et al.
: COMPREHENSIVE DESIGN APPROACH OF POWER ELECTRONIC-BASED DISTRIBUTED GENERATION UNITS
A Comprehensive Design Approach of Power
Electronic-Based Distributed Generation
Units Focused on Power-Quality
Improvement
Miguel Esparza, Student Member, IEEE, Juan Segundo, Member, IEEE, C. Nu´n˜ez, Member,
IEEE, Xiongfei Wang, Member, IEEE, and Frede Blaabjerg, Fellow, IEEE
Abstract—The undesirable harmonic distortion produced by
distributed generation units (DGUs) based on power-electronic
in- verters presents operating and power-quality challenges in
electric systems. The level of distortion depends on the internal
elements of the DGUs as well as on the characteristics of the grid,
loads, and controls, among others. This paper presents a
comprehen- sive method, focused on power-quality indexes and
efficiency for the design of microgrids with multiple DGUs
interconnected to the ac grid through three-phase multi-
Megawatt medium-voltage pulsewidth-modulated-voltage-source
inverters (PWM-VSI). The proposed design method is based on a
least square solution using the harmonic domain modeling
approach to effectively consider explicitly the harmonic
characteristics of the DGUs and their di- rect and cross-coupling
interaction with the grid, loads, and the other DGUs. Extensive
simulations and analyses against PSCAD are presented in order
to show the outstanding performance of the proposed design
approach.
Index Terms—Harmonic analysis, design optimization, power
quality.
I. INTRODUCTION
I N recent years, there have been an increased number of
grid-connected DGUs through power electronic interfaces,
which, in turn, has led to harmonic distortion problems.
Adverse power-quality effects due to harmonic distortions
are, for in- stance, interfering with communication, control
and protection systems, reduction of the equipment lifespan,
create additional losses in the power system, resonance and
stability problems [1], [2]. The harmonic distortion not only
depends on the har- monic sources, but also on the interaction
with the grid, loads, controls, and the others DGUs. In
addition, in power networks with closed-loop controlled
power electronic converters, the har- monic content of the
harmonic sources is tied to the harmonics in the signals
measured for the control schemes; introducing undesirable
behavior on the control performance, which can
Manuscript received December 15, 2015; revised April 26, 2016; accepted
June 7, 2016. Date of publication June 23, 2016; date of current version
March 22, 2017. Paper no. TPWRD-01785-2015.
M. Esparza, J. Segundo, and C. Nu´n˜ez are with the Universidad Auto
´noma de San Luis Potos´ı, San Luis Potos´ı 78290, Mexico (e-mail:
miesparza@ hotmail.com).
X. Wang and F. Blaabjerg are with the Aalborg University,
Aalborg 9200, Denmark.
Color versions of one or more of the figures in this paper are available
online at http://ieeexplore.ieee.org.
Digital Object Identifier 10.1109/TPWRD.2016.2584616
1
lead even to stability problems [2], [3]. Therefore, the
harmonic generation and propagation must be properly
reduced within the power-quality limits [4]–[6]. By
considering the harmonic interaction in the modeling of power
electronics based systems, better design and analysis can be
carried out. Although it is not a generalized practice, it is
common to assume certain ideal operating conditions which
neglect the harmonic interaction, in order to simplify the
model when performing some specific analysis, designs or
studies.
Ideal operating conditions are increasingly difficult to sus-
tain as valid, since harmonic generation, interaction of
controls and harmonic components, resonances, harmonic
stability is- sues, etc, are phenomena commonly reported in
systems with high harmonic penetration [7]. In practical
networks with power electronic-based DGUs, this could lead
to erroneous or unfavor- able operating scenarios. This gap has
been recognized and ad- dressed in the open literature,
although not in a holistic way. For instance, [8] deals with the
power transfer capability problem of DGUs including
harmonic distortion, and considering power- quality
regulations at a specific point of the network. Reference
[9] brings out some pitfalls of electric power-quality indexes
allowed by international regulations showing that being within
the limits is not enough in order to ensure proper performance
of the DGUs electrically close to each other. In more sophisti-
cated approaches, the power quality on the distribution system
is improved by the use of harmonic compensation methods in
the control schemes of the interconnected DGUs [10]–[14].
In this way, one of the best practice to mitigate the adverse
harmonic effects is the proper design of the passive filters
connected to the end terminals of power electronic converters.
Several design methods or procedures, specially for the LCL
filter, have been proposed. However, most are based on a trial
and error process in which their difficulty and convergence
problems considerably increase for systems with multiple
DGUs based on power electronic converters, specifically from
medium to high power levels [15]–[20].
This paper proposes a comprehensive approach, based on
op- timization and the extended harmonic domain (EHD) [21],
for the design of multiple grid-connected multi-Megawatt
medium- voltage PWM-VSI with LCL filters. This is carried
out by means of a Nonlinear Least SQuares formulation
(NLSQ), which cal- culates the filter parameters and the
steady state control variables which meet certain proposed
2 IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 32, NO. 2, APRIL
2017

reference operating conditions and

0885-8977 © 2016 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See http://www.ieee.org/publications standards/publications/rights/index.html for more information.
ESPARZA et al.: COMPREHENSIVE DESIGN APPROACH OF POWER ELECTRONIC-BASED DISTRIBUTED GENERATION UNITS
3

Fig. 1. Simplified layout for design.

includes power-quality restrictions and efficiency. As an

exam- ple, the design of two DGUs, based on three-phase
PWM-VSIs, which are connected to a microgrid is presented.
Two case stud- ies are presented to show the proposed design
approach, one considering that the interconnections grid is
unknown and the other when is known. The obtained results Fig. 2. Relationship between designable elements (DE), design objectives
show the remarkable good performance of the proposed (DO) and external conditions (EC).
design approach on both cases, along with advantages over
other design methodologies, which rely on the comprehensive
respectively. The pointing arrows between these elements are
consideration of multiple de- sign objectives.
links with the elements that have influence on or relation to
The paper is divided as follows: Section II presents the
the final value of them. Solid and dashed lines represent
design challenges of DGUs, Section III Introduces the EHD,
strong and weak interactions, respectively.
Section IV describes the proposed design approach, Section V
Fig. 2 shows an insight of the challenges in the designing
presents the harmonic model of the microgrid test system,
of DGUs. Some of the most common practices used to tackle
Section VI explains the selection of design objectives and the
them are: (1) settle many of the designable elements based on
designable elements of the case studies, Section VII presents
experience and a priory knowledge, especially those in respect
the design results of two different case studies, and Section
the topology, (2) decouple the relationships by considering
VIII draws the final conclusions.
only the most relevant designable elements for each design
objec- tive, (3) neglect some design objectives focusing only
II. DESIGN CHALLENGES OF DGUS on the most important. However, it is clear that as long as
Three main elements could be identified in the design of a more and better mathematical representations of these
DGU. (1) The Design Objectives (DO) (power quality, oper- relationships are provided, better and more reliable designs
ating conditions, size limitations, cost, etc.), (2) the External will be obtained; this is the basis of the presented design
Conditions (EC) (distributed resource, grid equivalent, proposal, to use the vast amount of information provided by
weather events, faults, generation outages, etc.) and (3) the using the harmonic domain model in order to describe the
Designable Elements (DE) (topology, component values, relationships and formulate the nonlinear problems enabling
control parame- ters, etc). In this context, a proper design can to accurately calculate the des- ignable elements.
be summarized as the selection of certain DE that ensure the
fulfillment of the DO in the presence of some EC. This III. EXTENDED HARMONIC DOMAIN MODELING OF DGUS
requires to understand in detail the relationships and
Electrical systems can be modeled by Linear Time Periodic
interactions among these main elements.
(LTP) systems, even in the presence of power electronics if the
Fig. 1 shows a very basic representation of a typical DGU
switching functions that drive them are periodic. However, the
and some of the above identified main elements are shown
discrete nature of these functions make it difficult dealing
(DO, EC and DE). From Fig. 1 the DO could be established,
with time domain periodic discontinuous models. The EHD
for exam- ple: DC bus voltage, DC voltage ripple, RMS
modeling overrides this limitation and obtains LTI models by
voltage at PCC, active power at PCC, reactive power at
considering the harmonic content of the signals in their
PCC, THD voltage at PCC, current ripple at PCC, among
formulation. In this way, any time domain LTP system, can be
others. Some of the DE are: distributed resource topology,
modeled as an LTI EHD model [21].
power electronic topology, AC and DC filter topologies,
The LTI EHD model of an electric system can be straight-
control unit topology, switching frequency, power switches
forward obtained from the LTP time domain model, if this is
ratings, DC filter component values and AC filter component
expressed in the form
values, control unit gains, among oth- ers. In order to have a
selection of the DE that ensures that the reference design d x(t) = A(t)x(t)+ B(t)u(t) (1)
objectives (DOref ) are met under bounded variation of certain dt
EC, is then required to understand the rela- tionships between
these main elements. by
It is clear that the relationships between the DO and the DE d
are far from linear and decoupled. Fig. 2 shows schematically X(t) = AX(t)+ BU(t) (2)
dt
an insight of the intricate relationships among the DO, DE and
EC; depicted by grey circles, blue ovals and green squares,
4 IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 32, NO. 2, APRIL
Fig. 3. Design algorithm of power converter optimization.
In the EHD model each periodic variable is replaced by
an harmonic vector of size (2h × 1) in which each element
correspond to an harmonic of the time-domain signal. In this
way, the k-th harmonic of periodic time-domain signal x(t) is
represented by the X×k ∗ element of the harmonic vector X
= [X×−h ∗, . . . , X×−1∗, X×0∗ , X×1∗ , . . . , X×h ∗]T . In this way, the
switching functions are included in the time periodic matrix (A)
t of time model (1); hence, the equivalent EHD model (2) will
include their harmonic content in matrix A [22].
IV. DESIGN PROPOSAL
The design proposal can be used in power electronic
systems able to be modeled in the EHD as in (2). A proper
implementa- tion considers that the Designable Elements (DE)
are involved in the matrices A and/or B of the EHD model.
Hence, the ref- erence design objectives (DOref ) are able to be
calculated from the EHD model steady state solution given by
Xss = −A−1 BU
In this way, for a given set of DE arranged in the unknown
vector x, a set of calculated design objectives (DOcal) can be
obtained from Xss. Hence, an optimization problem based on
the nonlinear least square (NLSQ) methodology is proposed
where the unknown vector x is calculated in order to obtain an
EHD model performance in which the DOcal meet (as close as
possible) their corresponding reference values DOref . This is
done by proposing the following objective function:
N respectively. The grid is represented by its Thevenin
Σ . Σ2
minimize F (x) = ρ DO − DO (x)
p ref calp
n p
=1
where ρp is the p-th weight constant and N is the number of
DO to meet.
Fig. 3 depicts the proposed design algorithm. Notice that in
order to evaluate the objective function F (x), it is required
to calculate the values of DOcaln (x) at the k-th evaluation of
the iterative numerical solution. This information is obtained
from the steady state solution of the EHD model, which can
be calculated from the k-th unknown vector x of DE.
2017
Fig. 4. Microgrid test system.
The flexibility offered by the NLSQ gives the possibility to
obtain many different solutions for the same system, depend-
ing on the approach used in the formulation. These approaches
are mainly dictated by the proposed design objectives along
with their weight constants. As long as the number of DO is
equal or greater than the number of DE and a mathematical re-
lationship between them exist, the algorithm is able to be
used. More complex design objectives can be included, such
as, price, the reliability, size, weight, temperature and
switching losses, etc., if those are calculated from the EHD
model steady state solution X ss .
The DOref of this paper are based on power-quality indexes,
in order to show the capability of the proposed design
approach and give an insight of its potential.
(3)
V. HARMONIC MODEL OF THE MICROGRID TEST SYSTEM
A. EHD Model of the Case Microgrid Test System
The proposed microgrid to design is shown in Fig. 4.
Three nodes are clearly identified in this system, labeled 1,
2 and PCC respectively. PC1 and PC2 are power cables of
different lengths, modeled by a three phase pi equivalent
line which con- nect these nodes. Distributed generation
units Dtt1 and Dtt2 are connected to nodes 1 and PCC
equivalent and is connected to node by means of a power
PCC transformer T3. A capacitor bank
(4) CB is also connected to this node in order to improve the
power factor of the microgrid interconnection. A linear load,
by LL and RL, along with a three-phase harmonic load, modeled
by ih, are connected in node 2.
A modular approach, which can easily be automated, is
used in order to obtain the EHD modeling of the case study.
For this purpose, the test system can be divided into
individual subsystems in order to derive their individual EHD
matrices, and plug them in into the EHD matrices of the test
system model. Under this conditions the EHD model of the
ESPARZA et al.: COMPREHENSIVE DESIGN APPROACH OF POWER ELECTRONIC-BASED DISTRIBUTED GENERATION UNITS
5
proposed
 6 IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 32, NO. 2, APRIL
2017

by lowercase bold letters that in general represent vectors

of size (3 × 1) constructed with the phases a, b and c, i.e.,
in1 = [in1 , ain1 , in1b ]T ; cTn is the Y − Δ transformer with
parameters rtn , ltn and an referred to the primary wind- ing;
control variables mna and θn are the modulation in- dex and
phase shift, respectively, of the modulating signal from which
the PWM block generates the switching function
s. For a closed-loop operation, the control variables will be
generated by a control strategy for given reference operating
Fig. 5. Distribution generation unit of test system.
conditions.
The (4H × 1) EHD state variables vector for this
subsystem is given by X Dtt n = [I n1, Vnf , I n2 T , V ] . The
test system is given by ndc

system ma-
trix and coupling matrices ADttn , BU1Dttn and (BU2Dttn ),
dX =A X of size (4H × 4H), (4H × 5H) and (4H × 4H) respectively,
M M M + B M UM (5)
dt the system matrix is given by
where XM = [X Dtt1 , X Dtt 2 , XttE , XRL , X TS ]T is formed
AD ttn =
with the harmonic state variables of each subsystem in the
microgrid; the subscripts state for the following
⎡ (Rn f + Rn 1 ) 1 R nf [Gna ] ⎤
I d − D − Id Id
subsystems: distributed generation unit 1 subsystem Dtt 1, − Ln Ln 1 Ln 1
Ln 1 ⎥
distributed gen- ⎢ 1 1 1
eration unit 2 subsystem Dtt2, grid equivalent subsystem ttE, I −D − I N ⎥
⎢ d d

M nO 1 M nO f M nO 2 − D N
linear load subsystem RL and interconnection subsystem (9)
TS. Cnf Cnf
UM = [Vs, I o1 , Io2 , Ih] is the input vector which in this case is ⎢
⎥
formed with the harmonic content of the equivalent grid ⎢
voltage
⎣ [Γn ] ⎦
− ⎥
source Vs; the dc current source of each DGU, i 1o and ; and Cndc N N −D
i 2o
the harmonic load ih. System matrix (AM ) and input matrix
and the coupling matrices defined by their non null elements are
(BM ) are given by
given by
⎡ ⎤
ADtt 1 0 0 0 BU1 Dtt 1 BU1 Dtt1( 3 , 3) = BU1 Dtt2( 3 , 4) = TnO (10)
⎢ 0 A Dtt 2 0 0 BU1 Dtt 2 ⎥
A = ⎢⎣
M 0 0 A ttE 0 BU 1 ttE 1
0 0 0 ARL BU1 RL ⎦ ⎥ BU2 Dtt2( 4 , 2) = BU2 Dtt2( 4 , 3) = Cndc Id (11)
BU1 T S BU2 T S BU3 T S BU4 T S A TS
where Id is the identity matrix of size (H × H); N is null
(6)
ma- trix of size (H × H); D = diag(D) is the EHD
and
differential matrix whit D = [−jhω0 , . . . , −jω0 , 0, jω0 , . . . ,
jhω0 ]. Ma-
T trices MO 1 = diag(MnO 1), MnO f = diag(MnO f ), MO 2 =
BM = [BU2 Dtt , 1BU2 Dtt , BU 2 2 ttE , BU2 RL , BU5 TS ]
(7) diag(MnO 2 ) and TnO = diag(TnO ) are diagonal matrix of
where A sub and BUn sub are the system matrix and the n- th size (H H), in which the i-th element of their diagonal
coupling matrices of formed with the input vector Usub and vectors, × MnO 1 , MnO f , MnO 2 and TnO respectively, are
input matrix B sub , respectively, when each subsystem is defined by the index number (i). These diagonal vectors
modeled by are given by
MnO 1(h + 1 + m)
d ⎧
Rn f
Xsub = Asub X sub + Bsub Usub (8) ⎪⎪ ∀ m = 0, ±3, ±6,...
dt ⎨ + nt
=
L 2
l

(12)
⎪
with sub = (Dtt1, Dtt2, ttE, RL,TS). Rn f 2 +L
By defining h as the maximum harmonic considered in the columns
⎩ of harmonic
n vectors as n2
3
EHD model, every EHD vector variable is of size (H 1) M f (h + 1 + m)
× L nO
with H = 2h + 1. Hence the size of EHD subsytem matrices n ⎧
1
in (5) can be stated by the r rows of harmonic vectors and c ⎨ Ln 2 + lnt
3
⎪
a
=
1
 ESPARZA et al.: COMPREHENSIVE DESIGN APPROACH OF POWER ELECTRONIC-BASED DISTRIBUTED GENERATION UNITS
7
+ lnt ∀ m = 0, ±3, ∀ m = 0, ±3, ±6,...
±6,... (13)

matrix element of an EHD subsystem matrix is denoted by the ⎪

subscript |· |(r,c). The EHD subsystem matrices of this case ∀ m = 0, ±3, ±6,...
⎩ Ln 3 an + Ln 2 + lnt
2

3
study are defined in the following subsections. MnO 2(h + 1 + m)
⎧
Rn 2 + Rn f +
B. Distributed Generation Unit Subsystem Model ⎪ − L +l ∀ m = 0, ±3, ±6,...
n2 nt
rn t (14)
Fig. 5 shows the topology for Dtt ⎨
= 2
+R
3
+

n ⎪⎩ (14)
subsystem with n Rn 3 an n nf + nt
∀ m = 0, ±3, ±6,...
−=
{1, 2}. The time-domain three-phase variables are denoted 2 3L
n3 a 2+ L
n2 + lntr
n
n
8 IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 32, NO. 2, APRIL
2017

Fig. 7. Interconnection subsystem circuit.

Fig. 6. Grid equivalent circuit.

TOg (h +1+ m)
TnO (h + 1 + m) ⎧
⎪0 ∀ m = 0, ±3, ±6, . . .

⎧ ⎨
⎪⎨ 0 ∀ m = 0, ±3, ±6, . . . = √
a( 3i − (20)
3)

s
t

= √ (15) ⎪⎩
an ( 3i − − ∀ m ƒ= 0, ±3, ±6, . . .
⎪
⎩− ∀ m = 0, ±3, ±6,... 2(3L a2 + l )
3) Ln 3 a2 + + )
2(3 n
Ln 2
lnt
n
Ln 2 lnt
The switching information explicitly depends on the
control variables mna and θn and modulating frequency D. RL Load Subsystem Model
mnf . From these parameters, elements of vectors Sn ×a , b , c ∗ can The EHD state variable vector of size (H × 1) is given
be calculated from analytical relationships [22]. [Gna ] is a by XRL = I rl . Hence, system matrix of size (H × H) is
Toeplitz matrix constructed with the elements of vector given by ARL L=L − R L Id − D. The coupling matrices of size
Ga given by (H × ×

Ga = Sa − S + S + S 5H) and (H 4H) are sparse matrix BU1 RL and null matrix
a b c
(16) BU2 RL , respectively. The non null EHD element of BU1 R L
3
For matrix [Γn ], let Γm be the m–th row of [Γna ] and Smna is given by BU1 RL(1 , 5) = 1LIL d.
na
be the m–th row of [Sna ]. Then [Γn] is defined by
. E. Interconnection Subsystem Model
∀ m = 0, ±6,... , ±h (17)
Γh+1+m = S h+1+mna
n
0 ∀ m = 0, ±6,... , ±h Consider the equivalent interconnection system depicted in
Fig. 7. In this figure, the power cables CP1 and CP2 are
C. Grid Equivalent Subsystem Model replaced by their pi model parameters, given by Rtlp , Ltlp , Ca tlp
Fig. 6 shows the grid equivalent where T3 is the Y − Δ and Cb tlp , with p = 1, 2. In this way, the equivalent
capacitor is
transformer with parameters rt , lt and a3 are referred to the given by Ce = CB + Cbtl1 + Catl2
primary winding; and ×
3 3
Ls and Rs are grid Thevenin The EHD state variable vector is a (5H 1) vector given
equivalent by X = [I ,I ,V ,V , V ]. The system ma-
parameters. The EHD state variable vector is a (H × 1) vector TS TL 1 a TL 2 a 1a pcca 2a

given by XttE = Ip. Hence, the system matrix of size (H × trix vector is given by
H) is given by A TS = Mg − D. Coupling matrices of size
(H × 5H) and (H × 4H) are given by sparse matrices AT S
BU1 ttE and
⎡
BU2 ttE , respectively, which non null elements are given by Rtl1 1 I −1 ⎤
− I −D N
d
I
d
N
L
BU1 ttE(1 , 4) = Mvg and BU2 ttE(1 , 1) = T Og , respectively. tl 1
d
Ltl1 Ltl1
= ⎢ ⎥
Matrices M = ), M Rtl
diag(M diag(M
), and L
T = diag( g ) ⎢ 2
Og vp ( Ovg ), in N − I
ESPARZA et al.: COMPREHENSIVE DESIGN APPROACH OF POWER ELECTRONIC-BASED DISTRIBUTED GENERATION UNITS
9
−D N
− 1 I 1
T Og ⎥
Og are diagonal matrix of size H×H ⎢ L tl2
d
L tl2
d d ⎥
tl2
which the i-nth element of their diagonal vectors, MOg , MOvg
=⎢ 1 ⎥

this context these diagonal vectors are given tl1

⎥
by 1
1 ⎥
and TOg respectively, are defined by the index number (i). In ⎢− Id N −D N N ⎥
this context these diagonal vectors are given ⎢
by Ca 1
1

M Og (h + 1 + m) ⎢ Id − Id N −D N
Ce C

Cbtl 2
⎧ ⎣ ⎦
⎪ rt ∀ m = 0, ±3, ±6,... N 1
Id N N −D
⎪⎨
=−l
(18)
(21)
t

(21)

⎪− 3Rs a2 + R L2 + t
∀ m ƒ= 0, ±3,
⎩ 3Lsa2 + lt r ±6,... The coupling matrices BU1 T S , BU2 T S , BU3 T S ( 1 , 5) ,
M Ovg (h + 1 + m) BU4 T S and BU5 T S of size (5H × 4H), (5H × 4H), (5H ×
⎧ H), (5H × H) and (5H × 4H), respectively are sparse
∀ m = 0, ±3, matrices defined by their non null EHD elements given
⎪l1 ±6,...
=⎨ t (19) by BU1 T S (3 , 3) = TiO 1 , BU2 T S ( 5 , 2) = TiO 2 , BU3 T S (4 , 1) =

⎪⎩ 1 − CI1e d and BU4 TS = (5BU

, 1) 5 TS
= − 1 I(5d., 4)
∀ m = 0, ±3, Ca 4
3L
s a2 t ±6,...
+l
 10 IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 32, NO. 2, APRIL
2017

Matrix TiOn = diag(Tion ) is given by the diagonal vector The calculation of these proposed design objectives from (3)
defined by
are given by
⎧
Vn dc×0∗ = Re (V ndc (h + 1)) (23)
⎨ ⎪0 ∀ m = 0, ±3,
±6, . . .
(h +1+ m) =
TiO n
an m =0
3 6
‚
⎪⎩ − ∀ , ± , ± ,... .
2Ctl p vnA RM S = VnA ×n ∗ Vn A ×−n ∗ (24)
(22) ,n
and Σ=−h
h

VI. TEST SYSTEM DESIGN OBJECTIVES (DO) AND P

DESIGNABLE ELEMENTS (DE) PFnA = Sno (25)
no
For the test system of Fig. 4, the main intention is to de- where
sign Dtt1 and Dtt2 LCL filter parameters, which offer the h
best power-quality signals, for certain desired DO and Σ
Pno = 3 VnA ×n ∗ In 2a ×−n (26)
considering some given fixed parameters (switching ∗
n
frequency, transformer =−h
parameters, grid impedance, interconnected loads, etc.). Sno = 3v nA R M S i n2 RM S (27)
Hence,
on each DGU, the DE have to include the LCL filter
2) Power-Quality Indexes Riin 1 , THDin 2 and Riv n d : The
parameters. c
Since
char- these DE are highly related with many other behavior
acteristics of the system (such operating condition and in1 current ripple (Riin 1 ), the in2 THD (THDin 2 ) and the v ndc
power transfer capability), DO (and DE) non directly voltage ripple (Rivn d c ) are selected as power-quality indexes.
related with the power quality of the signals have to be Many benefits are obtained if those power-quality indexes are
included in the proposed design approach. The following maintained as close as possible to zero. Among the most
subsections give more detail on the established DE and DO impor- tant are the fulfillment of power-quality international
for design Dtt1 and Dtt2 on the proposed test system of Fig. standards
4. [5] for interconnected DGU units. Moreover, from low values
of Riin 1 and Rivn d c , switching devices with lower voltage and
current ratings could be selected, which impacts on the size,
A. Designable Elements price and many other characteristics.
1) Steady State Control Variables mna and θn: By con- The calculation of the ripple of a signal is mainly a time
sidering the control variables as DE, the closed-loop per- domain calculation. Hence a time domain transformation from
formance of each DGU is considered in the design. In this the EHD steady state is required and this is represented by
way, the control references can be included as DO, al-
X → x(t) (28)
lowing to consider the nominal operating conditions in the
design. The ripple of the time-domain variable x(t), defined as Rix,
2) Input dc Current Source ino and dc Side Capacitor C ndc : is given by
By designing these parameters, the power quality of the dc ∗ ∗
Rix = max (X → x (t)) − min (X → x (t)) × 100 %
∗ ∗

voltage bus along with power rating related to the operating

conditions (such efficiency and transfered power) are able to 2|X×r ∗ |
be (29)
introduced as DO. where the asterisk (∗) represents the EHD and time-domain
3) LCL Filter Passive Components Ln1 , Ln2 , Rnf and Cnf : signals, X∗ and x∗ (t), respectively, with the reference r
The LCL filter parameter are mainly related with the power element (r = 1 for fundamental component and r = 0 for dc
quality of ac signals. However, their design impacts not component) cleared.
only power-quality DO but also on the overall closed-loop The THDx is calculated by
perfor- mance of the DGU. One example is to increase the .
Σh
damping 2
|X×n ∗ |
resistance Rf in order to reduce the resonance peak, n =2
however
a large value will affect the DGU overall efficiency. Hence, THDx = × 100 % (30)
the proper design of these parameters consider power quality, |X×1∗ |
operating conditions and transfer power DO.
ESPARZA et al.: COMPREHENSIVE DESIGN APPROACH OF POWER ELECTRONIC-BASED DISTRIBUTED GENERATION UNITS
B. Design Objectives
1) Operating Conditions Vn dc×0∗, vn A R M S and PFn A : The
dc component of vn dc voltage (Vn dc×0∗), the line to line ef-
fective voltage at output bus A (v nA RM S ) an the power fac-
tor at bus nA (PFnA) are selected as the desired operative
conditions.
11
3) Power Transfer Capability (Po) and Efficiency (η): Fi-
nally, two designs for different power transfer objectives are
conducted in this example: the active output power at bus A
(Po) and the overall efficiency of the DGU (η), which
includes the converter and the filter losses. These two design
objectives ensure that the required amount of power (if
available) is within the transfer capability of the LCL filter
including the feed- ers, transformer and network impedance
with highest possible efficiency.
 12 IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 32, NO. 2, APRIL
2017

TABLE I
TABLE III
DOref FOR ISOLATED AND COMPREHENSIVE DESIGN CASE STUDIES ISOLATED DESIGN: DE RESULTS

Design objective Dtt1 Dtt2 θn in o Ln 1 Cn d c Ln 2 Cn f Rn f

mna (rad) (A) (mH) (mF) (μH) (mF) (m Ω)
DO r e f 1 V n d c ×0 ∗ (V ) 1000 1200
DO re f 2 v n A RM S (V l −l ) 440 440
Dtt 1 0.77461 2.50551 1046.8 0.435 2.1 1.33 5.3 6.3
DO r e f 3 P n o (MW ) 1 1.5 Dtt 2 0.90722 3.03772 1315.9 0.5135 4.9 0.6012 8.7 0.2484
DO r e f 4 Ri i n 1 (%) 0 0
1
DO r e f 5 THD i n 2 (%) 0 0 Recalculated values, individual m 1 a = 0.7131 and θ1 = 2.648 rads
2
DOre f 6 Riv n d c (%) 10 7 Recalculated values, individual m 2 a = 0.9500 and θ2 = 2.8089 rads
DO re f 7 PF n A 1 1
DO r e f 8 η (%) 100 100 TABLE IV
ISOLATED DESIGN: DO SIMULATION RESULTS

TABLE II
DGUS FIXED PARAMETERS FOR ISOLATED AND COMPREHENSIVE DO 1 DO 2 DO 3 DO 4 DO 5 DO 6 DO 7 DO 8
DESIGN CASE STUDIES V n d c ×0 ∗v n A R M S P n o Ri i n 1 THD i n 2 Ri v n d c PF n A η
(V) (V) (MW ) (%) (%) (%) (Dimensionless) (%)

Parameter Dtt1 Dtt2 Dtt 1 1000 440 1 0 0 10 1 100

EHD 1000.2 441.7 1.004 7.95 1.52 9.73 0.99 95.9

SB (MVA) 3 3 (h = 500)
Tn (%) 0.5 + j 4 1 + j 4.5 PSCAD 994.5 440.6 0.998 8.01 1.61 9.91 0.99 95.8
Rn 1 (%) 1 1.5 (ts = 1 μs)
Rn 2 (%) 0.5 1.1 Dtt 2 1200 440 1.5 0 0 7 1 100
Rn 3 (Ω) 0.11 0.22 EHD 1200.87 440.2 1.47 7.69 1.94 7.87 0.985 93.62

Ln 3 (m H) 1 2 (h = 500)
vn s (kVRM S L −L ) 13.8 13.8 PSCAD 1200.11 438.42 1.47 7.53 1.92 7.91 0.985 92.22
Modulation SPWM SPWM (t s = 1 μs)
mnf 27 21

The calculation of the design objective Pno was already de- harmonic load is Ih ×1∗ = 6.6150/ −26.56◦ A, with a harmonic
fined in (26). The efficiency η is calculated by component of 50% and 20% of the fundamental for the fifth
and seventh harmonic, respectively.
Pno
η= (31)
In o×0∗ Vn dc×0∗ A. Isolated Design Case Study
In this case, each DGU is designed individually, considering
VII. MICROGRID DESIGN CASE STUDIES their interconnection buses as ideal voltage sources (infinite
In order to show the flexibility of the proposed design The RL load subsystem has the parameters LL = 0.1212 H and
ap- proach, two design cases studies are presented in this RL = 45.7 Ω. The fundamental frequency current phasor of the
Section for the microgrid test system proposed in Fig. 4.
The first case considers that no information of the microgrid
test system where Dtt1 and Dtt2 will be connected is
available, named as isolated design. The second case
considers that the microgrid test system where Dtt1 and Dtt2
will be connected is known, named as comprehensive
design. The main objective in both cases is to find the
designable elements proposed in Section VI-A for Dtt1 and
Dtt2 which meet as close as possible the reference design
objectives of Table I over an steady state operation of the
case study system.
The considered fixed parameters for each DGU are given
in Table II. Notice how the design proposed for Dtt2 is more
exigent since the modulation frequency mf is smaller and the
output power is 50% greater as compared with Dtt1.
The fixed parameters for the microgrid test system (Fig. 4)
are the following. The interconnection subsystem has power
cables P C1 and P C2 which parameters are Ca tl1 = Cb tl1 =
5.6 μF, Ltl1 = 1.1 mH, Ca tl 2 = Cb tl 2 = 9 μF, Ltl2 = 1.9 mH
with a capacitor bank CB = 3 μF. The grid equivalent sub-
system is conformed by Ls = 12.6 mH and Rs = 0.2142 Ω.
ESPARZA et al.: COMPREHENSIVE DESIGN APPROACH OF POWER ELECTRONIC-BASED DISTRIBUTED GENERATION UNITS
13
bus), please refer to Fig. 5. Two individual designs are
performed using the EHD subsystem models Dtt1 and Dtt2
given by (7). The results are used to compute the periodic
steady-state solution of the microgrid presented in Fig. 4
through PSCAD and the EHD model (5) in order to
evaluate the performance and validate the obtained design
once the designed DGUs are interconnected to this
system.
Table III shows the Case Study design results for Dtt1 and
Dtt2. Notice that DE mna and θn for both DGUs are valid
only for isolated operation. When interconnected the control
strate- gies will find the correct values that reach the DOref
of each DGU. Hence these have to be recalculated in a
simpler design by considering as DE only the control related
designable ele- ments (m1a, m2a, θ1, θ2) in order to validate
the obtained DE for the results of this case study when
DGUs are interconnected to the test system. Notice that,
depending on the interconnection system, the isolated design
results could result in undesirable op- erating conditions (ma
close to one or over-modulation). These recalculated values
are also indicated in Table III.
Table IV shows the reference design objective values and
their simulated values for both DGUs. The simulation
results given in Table IV show a very good agreement
between PSCAD and EHD solutions. The PSCAD
simulation is introduced to verify the DE results due to
the high detail level used in
 14 IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 32, NO. 2, APRIL
2017

TABLE VI
COMPREHENSIVE DESIGN: DO SIMULATION RESULTS

DO 1 DO 2 DO 3 DO 4 DO 5 DO 6 DO 7 DO 8
V n d c ×0 ∗ v n A R M S P n o Ri i n 1 TH D i n 2 Ri v n d c PF n A η
(V) (V) (MW ) (%) (%) (%) (dimensionless) (%)

Dtt 1 1000 440 1 0 0 10 1 1

EHD 999.79 443.8 1 7.82 1.6 10.08 0.99 96.5
(h = 500)
PSCAD 996.4 441.3 0.995 8.005 1.64 10.8 0.995 96.5
(ts = 1 μs)
Dtt 2 1200 440 1.5 0 0 7 1 1
EHD 1199.3 442.5 1.5 7.62 1.86 7.04 0.99 93
(h = 500)
PSCAD 1199.4 439.6 1.495 7.75 1.89 7.17 0.99 92.59
(ts = 1 μs)

Fig. 8. PSCAD and EHD simulated waveforms for isolated design Case Study.
(a) Dtt1 i11 converter current. (b) Dtt2 i21 converter current. (c) Dtt1 v1dc
voltage. (d) Dtt2 v2dc voltage. (e) Dtt1 node 1A voltage. (f) Dtt2 node 2A
voltage.

TABLE V
COMPREHENSIVE DESIGN: DE RESULTS

θn in o Ln 1 Cn d c Ln 2 Cf Rn f
mn a (rad) (A) (mH) (mF) (μH) (mF) (m Ω)

Dtt 1 0.7998 2.49010 1035.7 0.4489 1.91 2.48 4.83 0.0003

Dtt 2 0.9305 2.99607 1346.5 0.5134 5.47 0.5993 8.6 7.5

their electric and electronic component models. Fig. 8,

shows some PSCAD and EHD steady state (EHD-SS)
simulated waveforms of Dtt1 and Dtt2 when connected to
the test system, considering the DE results of Table III.
From the simulations shown in Fig. 8 it can be seen that
both waveforms are practically overlapped. This validates the
EHD model used to obtain the designable elements and the
design approach proposed. The achieved power-quality
indexes are ex- cellent considering the high power capability
and low switching frequency considered in the design. When
interconnected to the microgrid, each DGU behaves very
close to an ideal harmonic free voltage source and their
overall harmonic distortion im- pact over the microgrid is
practically negligible. However, since each DGU was
designed without considering all the elements interconnected
to them, the obtained design is decoupled and the isolated
operating conditions have to be verified when in-
terconnected. From this point of view, a better design could
be obtained if the complete system model is considered in the
proposed design approach.
B. Comprehensive Design Case Study
In this case study, the design of both DGUs is performed
simultaneously, considering all the elements interconnected in
the microgrid, using the overall model derived in (5) for the
system of Fig. 4. The design results for this case are shown in
Table V.
As in the previous case, these DE results of Table V are
used to perform PSCAD and EHD-SS simulations of the test
system
Fig. 9. PSCAD and EHD simulated waveforms for comprehensive design
Case Study. (a) Dtt1 i11 converter current. (b) Dtt2 i21 converter current.
(c) Dtt1 v1dc voltage. (d) Dtt2 v2dc voltage. (e) Dtt1 node 1A voltage.
(f) Dtt2 node 2A voltage.
in order to validate the design. The resulting design objectives
are shown in Table VI.
Fig. 9 shows some PSCAD and EHD-SS simulated
wave- forms of Dtt1 and Dtt2 when connected to the test
system, considering the DE results of Table V.
As in the previous case study, the PSCAD simulation
validates the obtained results and the proposed design
approach. A brief discussion of the obtained results for both
case studies is carried out in the following subsection.
C. Results Discussion
The results of both case studies show a remarkable perfor-
mance of the proposed design approach. Multiple and diverse
DOref are closely met, while the grid side power-quality stan-
dards are easily fulfilled with a very reduced converter current
ripple; even in the presence of low switching frequencies and
harmonic loads, with the best efficiency possible. In both de-
sign Case Studies, the performance of each DGU is seen by
the network almost as an ideal harmonic free voltage source
and prevents any harmonic related issue in the network caused
by the operation of th DGUs. For this reason, the overall per-
formance of the system and the obtained DE are very close
950 IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 32, NO. 2, APRIL
2017

in both case studies. However important differences should be Power Electron., vol. 29, no. 12, pp. 6421–6432, Dec. 2014.
[4] I. Standards, “IEC61000-X-X-Electromagnetic compatibility (EMC),”
pointed out. 1994.
Regarding the performance of Dtt1, an improvement of
around 0.7% in the efficiency for comprehensive case
could be noticed. Despite the relatively small
improvement, notice that there are significant differences
among both DE, specially in the damping resistance. An
interpretation is that in the comprehen- sive design, the
damping is obtained from the grid resistances. This allows
to improve the efficiency by decreasing Rnf while all the
other DE are adjusted to meet the other DOref .
For Dtt2, the isolated design presents a difference in the
out- put power Pno of around 30 kW respect to the
reference value, which results from non consideration of
the interconnection grid. This difference is significantly
improved in the compre- hensive design. Additionally,
most of the simulated design ob- jectives in the
comprehensive design are slightly better than in the
isolated design. However a small decrease in the efficiency
can be noticed, resulting from an increased Rnf value. In
short, comprehensive design provides an improvement in
the overall performance of Dtt2. This is because the overall
microgrid el- ements have much more impact on Dtt2 since
it is connected to the PCC node, which concentrates most
of the relationships among all the connected subsystems to
the microgrid.

VIII. CONCLUSION
This paper has introduced a novel design methodology
based on optimization and the extended harmonic domain
(EHD) for interconnected distributed generation units (DGUs)
in which the harmonic distortion and its effects over multiple
design ob- jectives are explicitly considered. The design
results of the pre- sented case studies have shown a
remarkable performance when both, the grid parameters are
available and not available, offer- ing an excellent power
quality with the best efficiency possible in the presence of low
switching frequencies. Compared with other design
methodologies, this proposal offers an advanced performance,
which rely on the comprehensive consideration of multiple
design objectives.

ACKNOWLEDGMENT
The authors want to thank to the Universidad Auto´noma
de San Luis Potos´ı (UASLP) through the Facultad de Ingenier
´ıa and the projects FORDECYT 190966 and FAI 2015 C15-
FAI- 04-106.106 for the facilities granted to carry-out this
research. Miguel Esparza acknowledges the financial support
granted by CONACYT to finance this research.

REFERENCES
[1] A. Medina, J. Segundo, P. Ribeiro, W. Xu, K. Lian, G. Chang, V.
Dinavahi, and N. Watson, “Harmonic analysis in frequency and time
domain,” IEEE Trans. Power Del., vol. 28, no. 3, pp. 1813–1821, Jul.
2013.
[2] J. Segundo-Ram´ırez, A. Medina, A. Ghosh, and G. Ledwich, “Stability
boundary analysis of the dynamic voltage restorer in weak systems with
dynamic loads,” Int. J. Circuit Theory Appl., vol. 40, no. 6, pp. 551–
569, Jun. 2012.
[3] X. Wang, F. Blaabjerg, and W. Wu, “Modeling and analysis of harmonic
stability in an AC power-electronics-based power system,” IEEE Trans.
16 IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 32, NO. 2, APRIL
2017
[5] “IEEE Standard for Interconnecting Distributed Resources with
Electric Power Systems,” IEEE Standard 1547-2003, pp. 1–28, Jul.
2003.
[6] EN50160, “Voltage characteristics of electricity supplied by public
distri- bution systems,” 1994.
[7] X. Tang, W. Deng, and Z. Qi, “Investigation of the dynamic stability
of microgrid,” IEEE Trans. Power Syst., vol. 29, no. 2, pp. 698–706,
Mar. 2014.
[8] I. N. Santos, V. C´ uk, P. M. Almeida, M. H. J. Bollen, and P. F.
Ribeiro, “Considerations on hosting capacity for harmonic distortions
on trans- mission and distribution systems,” Elect. Power Syst. Res.,
vol. 119, pp. 199–206, Feb. 2015.
[9] X. Zong, P. Gray, and P. Lehn, “New metric recommended for IEEE
Std. 1547 to limit harmonics injected into distorted grids,” IEEE
Trans. Power Del., 2015.
[10] J. He, Y. W. Li, F. Blaabjerg, and X. Wang, “Active harmonic filter-
ing using current-controlled, grid-connected DG units with closed-
loop power control,” IEEE Trans. Power Electron., vol. 29, no. 2, pp.
642–653, Feb. 2014.
[11] X. Wen, Y. W. Li, and J. He, “A distribution system harmonic
compen- sation approach using DG-grid interfacing converters at low
switching frequency,” in Proc. 29th Annu. IEEE Appl. Power
Electron. Conf. Expo., Mar. 2014, pp. 995–1001.
[12] J. He, Y. W. Li, R. Wang, and C. Zhang, “Analysis and Mitigation of
Resonance Propagation in Grid-Connected and Islanding
Microgrids,” IEEE Trans. Energy Convers., vol. 30, no. 1, pp. 70–81,
Mar. 2015.
[13] H. Yoshida, K. Wada, and P.-T. Cheng, “Harmonic current
suppression using single-phase DG units in three-phase grids with
unbalanced har- monic sources,” in Proc. IEEE Energy Convers.
Congr. Expo., Sep. 2015, pp. 6239–6245.
[14] Z. Zeng, H. Yang, J. Guerrero, and R. Zhao, “Multi-functional
distributed generation unit for power-quality enhancement,” IET
Power Electron., vol. 8, no. 3, pp. 467–476, 2015.
[15] A. Rockhill, M. Liserre, R. Teodorescu, and P. Rodriguez, “Grid-filter
de- sign for a multimegawatt medium-voltage voltage-source
inverter,” IEEE Trans. Ind. Electron., vol. 58, no. 4, pp. 1205–1217,
Apr. 2011.
[16] A. Reznik, M. Simoes, A. Al-Durra, and S. Muyeen, “Filter design
and performance analysis for grid-interconnected systems,” IEEE
Trans. Ind. Appl., vol. 50, no. 2, pp. 1225–1232, Mar. 2014.
[17] H. Karshenas and H. Saghafi, “Basic criteria in designing LCL filters
for grid connected converters,” in Proc. IEEE Int. Symp. Ind.
Electron., vol. 3, Jul. 2006, pp. 1996–2000.
[18] T. H. Kim, S. H. Kim, B. K. Kwon, and B. S. Kim, “A procedure to
design LCL filter for energy storage system,” in Proc. IEEE 8th Int.
Conf. Power Electron. ECCE Asia, May 2011, pp. 2974–2978.
[19] M. Hajian and D. Jovcic, “30kw, 200v/900v LCL IGBT DC/DC
converter prototype design and testing,” in Proc. IEEE Power Energy
Soc. Innovative Smart Grid Technol. Conf., Oct. 2014, pp. 1–5.
[20] W. Lin and D. Jovcic, “LCL and L-VSC converters with DC fault
current- limiting property and minimal power losses,” IEEE Trans.
Power Del., vol. 29, no. 5, pp. 2359–2368, Oct. 2014.
[21] J. Rico, M. Madrigal, and E. Acha, “Dynamic harmonic evolution
using the extended harmonic domain,” IEEE Trans. Power Del., vol.
18, no. 2, pp. 587–594, Apr. 2003.
[22] M. Esparza and J. Segundo-Ram´ırez, “An exact method for analysis
and component design of grid connected VSC-based power devices,”
Int. J. Elect. Power Energy Syst., vol. 78, pp. 489–498, Jun. 2016.

Miguel Esparza (S’12), photograph and biography not available at the time
of publication

Juan Segundo (M’10), photograph and biography not available at the time of
publication.

C. Nu´ n˜ ez (M’94), photograph and biography not available at the time of

publi- cation.

Xiongfei Wang (S’10–M’13), photograph and biography not available at the

time of publication.

Frede Blaabjerg (S’86–M’88–SM’97–F’03), photograph and biography not

available at the time of publication.