Revised Reprint of Paper No.

201702280000024 of
The International Conference on Electrical Engineering (ICEE 2017) Weihai, China, July 4-7, 2017

Parameterisation of Frequency Domain Models of Power Electronic

Interfaced Renewable Energy Sources and VSC-HVDC for Harmonic Studies
BERND WEISE, GENEVIEVE LIETZ

DIgSILENT GmbH, Germany

Abstract: modelling wind turbines as Norton or Thevenin

This paper summarises typical frequency domain equivalents [8]. The parameterisation of these
models of power electronic (PE) interfaced renewable equivalents based on field measurements requires
energy sources (RES), such as wind turbines or information about the phase angles of harmonic
photovoltaic (PV) inverters, and of voltage-sourced currents and voltages, which are not usually
converters (VSC) of high voltage direct current
disclosed in harmonics measurements reports
(HVDC) for the analysis of harmonics emissions. The
focus of the paper is on the parameterisation of these according to IEC 61400-21 [1]. Suggested
models in practical applications. Dependencies of the procedures to determine phase angles from
harmonic spectra on VSC operating points (power measurements of time-varying harmonics exist [9,
bins) are considered, and discussion of the practical 10]. The usage of the prevailing angle is seen as a
challenges for a correct model parameterisation is practical solution in technical guidelines [6, 10].
provided. Special attention is given to the phase angles This paper is organised as follows: Section 2
of harmonic currents and voltages. Phase angles of introduces the above-mentioned models in detail.
interharmonics are also discussed. The paper Section 3 describes the parameterisation of the
highlights methods to parameterise the inner sources
models as follows: Subsection 3.1 and 3.2 introduce
of Norton and Thevenin equivalents, based on
measurements at the terminals of the device in the the basic means to parameterise the models.
field and at test benches while considering the Subsection 3.3 focuses on phase angles, Subsection
influence of the network impedance. Possibilities to 3.4 on dependencies on operating points.
parameterise models based on incomplete Subsection 3.5 addresses parameterisation in cases
measurements are discussed, and limitations shown. of incomplete measurements. Conclusions are
Drawbacks of insufficient models and of assumptions drawn in Section 4.
in the parameterisation (in cases of incomplete
measurements) are explained.
2. Frequency Domain Models of PE
Keywords: Interfaced RES and VSC-HVDC
Harmonics analysis, frequency domain models,
power electronic interfaced renewable energy sources The ideal current source, depicted in Fig. 1, is
(RES), high voltage direct current (HVDC), voltage- the simplest model to represent harmonic injections.
sourced converter (VSC), Norton equivalent, This model is commonly used and in many cases
Thevenin equivalent, phase angle, prevailing angle, provides a good representation of harmonics
interharmonics injected by line-commutated converters (LCC) such
as thyristor converters of HVDC systems [11].

1. Introduction

For the analysis of harmonic emissions of

power electronic (PE) interfaced renewable energy
sources (RES), such as wind turbines or
photovoltaic (PV) inverters, and of high voltage Figure 1 Ideal current source
direct current (HVDC) converters, their behaviour
is usually assumed to correlate with ideal harmonic For PE interfaced RES and VSC-HVDC
current sources. Measurement approaches and systems, and for devices with VSCs in general, a
existing certification procedures often follow this Thevenin or Norton equivalent (see Fig. 2) better
assumption (e.g. [1, 2]). Practical experiences over represents the harmonic behaviour of such devices
recent years have highlighted however, that (the term ‘harmonics’ used here includes integer
voltage-sourced converters (VSC) interact with the order harmonics and interharmonics unless stated
grid and therefore cannot be treated as ideal otherwise). These equivalents are characterised by
harmonic current sources [3, 4, 5]. Enhanced their open-circuit voltage and short-circuit current,
measurement procedures [6] and modelling as shown in Eq. 1 and 2. The inner voltage source
approaches [4, 5, 7, 8] have been proposed. The of the Thevenin equivalent equals the open-circuit
draft IEC technical report 61400-21-3 suggests voltage, while the inner current source of the

Paper submitted on April 30, 2017. Paper accepted on May 15, 2017. Page 1/9
Paper presented in Session 7-1 of ICEE 2017 in Weihai, China on July 7, 2017.
Reprint version revised by authors after ICEE 2017; text revisions are marked in green.
Revised Reprint of Paper No. 201702280000024 of
The International Conference on Electrical Engineering (ICEE 2017) Weihai, China, July 4-7, 2017

Norton equivalent equals the short-circuit current. imaginary parts of the admittance appear in the real
In case of a two-level VSC, the impedance part of the impedance (Eq. 7) as well as in the
represents the series reactor, the filters and in some imaginary part of the impedance (Eq. 8).

# $%∙&
cases the machine or converter transformer. In case ' '
" ( )*∙+
of a modular multi-level converter (MMC), which (6)

#
(
is usually used for VSC-HVDC, the impedance
(, )+,
mainly represents the arm reactors (half the value of (7)

&
one arm reactor). If the controller transfer function
-+
(, )+,
is considered and represented by the impedance, the (8)
equivalent can be used not only for harmonics
analysis, but also for impedance-based stability Fig. 4 shows the possible interaction of PE
analysis [12]. If the device behaves linearly, i.e. the interfaced RES or HVDC VSC with existing
impedance or admittance does not depend on the background harmonics, i.e. with a harmonic voltage
voltage or current, the Thevenin and Norton distortion pre-existing in the grid regardless the
equivalents are equivalent to each other, as connected RES or HVDC VSC. The background
indicated by Eq. 3. The inner sources can be easily harmonic voltages feed harmonic current
calculated from each other, see Eq. 4 and 5. components flowing into the admittance of the
, , ,
Norton equivalent, as shown in Eq. 9. The injected
(1)
, , , ,
harmonic current components of the RES or HVDC
VSC depend on the impedances given in Eq. 10.
, , ,
(2) The resulting current that can be measured is
, , , ,

1⁄ ⟺ 1⁄
written in Eq. 11.
(3)

,
,
, ∙ (4)

,
,
, ∙
"
(5)

Figure 4 Norton equivalent illustrating potential

interactions with background harmonics from the
Figure 2 Thevenin equivalent (left) and grid [3]
Norton equivalent (right)
, ,6 7
, ,(./0→23.4. ) 8
(9)
In order to use the same impedance 6 7 9.

, ,23.4.→(./0
representation in the Norton and Thevenin (10)
∙ ;1 < 6 7
=
equivalents, it is useful to express the Norton
, ,23.4. ) 8
6 7 9.
equivalent with a parallel impedance Z(f) instead of
<
a parallel admittance Y(f), see Eq. 6 and Fig. 3. This
way, the Thevenin equivalent can be used for the , ,>?@A , ,23.4.→(./0 , ,(./0→23.4. (11)
Norton equivalent as well. This is the approach The effective use of a Thevenin equivalent to
taken in the commercially available network model a type 4 wind turbine was demonstrated in
analysis software DIgSILENT PowerFactory. [13] as follows:
• The harmonic emissions from the converter
are represented as several Thevenin
equivalent circuits at relevant harmonic or
interharmonic frequencies.
• The equivalent impedances in the model can
be used to represent the physical
components (smoothing reactor, harmonic
filters) and the converter closed loop control,
thereby representing the interaction between
Figure 3 Norton equivalent with admittance the converter and background harmonics.
(left), and with impedance (right) • The inner voltage sources in the model
It should be noted that both the real and represent harmonic disturbances caused by
the switching of the power electronic valves.

Paper submitted on April 30, 2017. Paper accepted on May 15, 2017. Page 2/9
Paper presented in Session 7-1 of ICEE 2017 in Weihai, China on July 7, 2017.
Reprint version revised by authors after ICEE 2017; text revisions are marked in green.
Revised Reprint of Paper No. 201702280000024 of
The International Conference on Electrical Engineering (ICEE 2017) Weihai, China, July 4-7, 2017

In the case of VSC-HVDC with MMC, the 3.2 Norton and Thevenin Equivalents
equivalent impedance should be modelled as
accurately as possible, because it represents As the Norton and Thevenin equivalents
exclusively the converter impedance (as no ac consist of an inner source and an impedance (or
filters are present) and will determine any admittance), both must be parameterised.
background harmonic amplification caused by
resonance between the network impedance and the 3.2.1 Parameterisation of the Impedance
converter equivalent impedance.
As previously mentioned, the impedance
3. Parameterisation of Frequency mainly represents series reactors or arm reactors,
Domain Models filters and in some cases the machine or converter
transformer. The controller transfer function can be
This section describes the means to considered and represented by the impedance as
parameterise the models introduced in Section 2. well [8], which is particularly important for
impedance-based stability analysis [12].
3.1 Ideal Current Source The equivalent impedance can be determined
in a theoretical analytical way by deriving the
The ideal current source model is very easy to frequency characteristics from the device’s internal
parameterise, under the assumption that the current impedances (e.g. passive filters) and from the
injected by the source is independent from any outer controller transfer functions (small signal
condition, i.e. independent from the grid impedance representation) as indicated in [4, 5, 7]. It is also
or background harmonic voltage distortion at the possible to measure the impedance as seen from the
terminal. The value of the harmonic current source device terminal, as described in [14], if the inner
is equal to the harmonic current measured at the source is negligible in the frequency range of
source terminal. interest. In a three-phase system the sequences (pos.,
Measurement reports nowadays usually list the neg., zero sequence) have to be considered and
harmonic current magnitudes in percent, based on measurements taken accordingly. The equivalent
the nominal apparent current of the device [1, 10]. impedance can be different in the individual
Care has to be taken with the definition of the sequences, as described in [4, 5, 7] and indicated in
nominal apparent current, which is used for the [8].
conversion of the measured data. The nominal If dynamic models for EMT simulation exist,
apparent current is either the current at nominal which represent the behaviour with sufficient
apparent power and nominal voltage (Eq. 12), or it precision in the frequency range of interest, the
is the current at nominal active power (i.e. apparent frequency characteristic of the impedance can be
power with unity power factor) and nominal voltage determined via EMT simulations which emulate the
(Eq. 13). The latter approach is usually used for above-mentioned tests. The latter approach is
measurements and in measurement reports. advantageous especially for operating points which
Using the current at nominal apparent power are difficult to measure [14]. Validation of the
for the model while the measurement report uses the simulation models against available measurements
current at unity power factor errs on the side of is recommended.
caution but leads to a slight overestimation of the An example template for a possible
injected harmonic currents as indicated by Eq. 14. representation of the frequency-dependent
The recalculation of the harmonic current for the equivalent impedance in tabular format (i.e. look-
precise representation is given in Eq. 15. up table) is provided in [8].

DEF
B DGHIF
4@A? BC √K∙ DGHF
(12) 3.2.2 Parameterisation of the Inner

DEF
B ∙MNO P L ∙MNO Q
Source
4@A? LC √K∙ √K∙
(13)

DEF D%LC F ∙ DEF

To parameterise the inner source of the Norton
, , 4@A? BC
≥ D%LC F ∙ DEF
(14) or Thevenin equivalent, both harmonic currents and
, 4@A? LC voltages have to be measured. It is not sufficient to
DEF D%LC F ∙ ∙ DEF
L only use the magnitudes. Magnitudes and phase
, , B 4@A? BC (15)
D%BC F ∙ DEF
angles have to be taken into account. This means
, 4@A? BC
that harmonic currents and voltages have to be
measured simultaneously and phase-correct.
If the complex impedance of the equivalent is
known, the inner source can be parameterised as
shown in Eq. 16 (Thevenin equivalent) or Eq. 17

Paper submitted on April 30, 2017. Paper accepted on May 15, 2017. Page 3/9
Paper presented in Session 7-1 of ICEE 2017 in Weihai, China on July 7, 2017.
Reprint version revised by authors after ICEE 2017; text revisions are marked in green.
Revised Reprint of Paper No. 201702280000024 of
The International Conference on Electrical Engineering (ICEE 2017) Weihai, China, July 4-7, 2017

(Norton equivalent). 3.3.1 Reference Point for Phase Angles

, , ,>?@A $ ∙ , ,>?@A (16)
Measurements and complex network
, , ,>?@A $ , ,T U
(17) calculations usually use different reference points
for phase angles. As soon as harmonic sources of
If the measurements are performed in a more than one device type are considered in the
laboratory or at a test bench, for which the harmonic study, this fact has to be taken into
frequency-dependent impedance of the whole set- account and the phase angles of the sources
up is known and no other harmonic source exists parameterised correctly.
besides the device under test (DUT), the inner In measurements the zero crossing of the
source of the Thevenin equivalent (Eq. 18) or fundamental frequency voltage is usually taken as
Norton equivalent (Eq. 19) can be parameterised the reference point for all angles, because it is easy
without measurement of the harmonic voltage. It to detect. Fig. 5 shows the time domain voltages
should be noted that this is not possible for and currents and the corresponding angles as
measurements in the field, for which the frequency- defined in [6, 10]. The zero crossing as reference
dependency of the grid impedance and the point refers to a sine function. Currents and voltages
background harmonics sources which are present can be expressed as written in Eq. 20 to 23.
during the measurement are usually not known.

, , ,>?@A ∙; V 2 $ 2?AW+?CXY = (18)

, , ,>?@A ∙ ;1 $ 8 Z
= (19)
[\8

Example templates for a possible

representation of the inner source of the Norton and
Thevenin equivalents in a tabular format (i.e. look-
up tables) are provided in [8].

3.3 Phase Angles

Phase angles of harmonic currents and voltages Figure 5 Angles resulting from measurements
have to be considered if the Norton or Thevenin depicted as corresponding time curves
equivalent is parameterised, as described in Section
3.2. In the following subsections considerations ]' ^ √2 ∙ ' ∙ sincd' ∙ ^ $ e3,'
A/C
f (20)
√2 ∙ ' ∙ sin d' ∙ ^ with e3,'
A/C
0
referring to phase angles which are important for
harmonic studies are listed.
It should be noted that even if the phase angles h' ^ √2 ∙ ' ∙ sincd' ∙ ^ $ e/,'
A/C
f (21)

] ^ √2 ∙ ∙ sincd ∙ ^ $ e3, f
have to be considered for calculation of the values
A/C
for the inner sources, it would still be possible to (22)

h ^ √2 ∙ ∙ sincd ∙ ^ $ e/,A/C f
use only the magnitude of the resulting complex
number for the inner source in the model. This (23)
applies a magnitude only (instead of a complex The real part of the complex number used in
number) for the inner sources in Fig. 1, 2 and 3. By network calculations represents a cosine function,
following the latter approach, a summation law (for as indicated by the Euler function in Eq. 24 (with
example as defined in [1, 10, 15]) would be needed time function) and Eq. 25 (for steady-state
to investigate the harmonic propagation of several calculations without time function). Therefore, in
harmonic sources in a harmonic study. This is network calculations, the peak of the sinusoidal
preferable if it is not certain that the contributions waveform is usually taken to be the reference point,
of the harmonic sources are synchronised to their referring to the cosine function.
i* cos d^ $ e $ % ∙ sin d^ $ e
local angle reference points at the fundamental
jW)P
frequency (i.e. if the PAR is small; see Section (24)
i *P cos e $ % ∙ sin e
3.3.3), and for interharmonics in general. If the
(25)
harmonic sources show a synchronised behaviour
related to the local angle reference points, To represent the time curves of Fig. 5 by cosine
consideration of the phase angles in the harmonic functions instead of sine functions, Eqs. 20 - 23
models and using a complex harmonic load flow should be rewritten as Eqs. 26 - 29 with a shift of
calculation in harmonic studies will provide more the angle by -90 degrees (= -π/2) to keep the same
precise results and should be favoured. reference point in time.

Paper submitted on April 30, 2017. Paper accepted on May 15, 2017. Page 4/9
Paper presented in Session 7-1 of ICEE 2017 in Weihai, China on July 7, 2017.
Reprint version revised by authors after ICEE 2017; text revisions are marked in green.
Revised Reprint of Paper No. 201702280000024 of
The International Conference on Electrical Engineering (ICEE 2017) Weihai, China, July 4-7, 2017

]' ^ √2 ∙ ∙ cos ;d' ∙ ^ < =

m
'
synchronised with the current of the fundamental
n
(26)
frequency (for LCC this is always the case; for VSC
h' ^ √2 ∙ ∙ cos ;d' ∙ ^ $ e/,' < =
A/C m
'
it depends on the effect which causes the harmonic
n
(27)
and on the pulse pattern of IGBT switching), it
] ^ √2 ∙ ∙ cos ;d ∙ ^ $ e3, < =
A/C m makes sense to use an angle for the harmonic
n
(28)
currents which is not related to the fundamental
h ^ √2 ∙ ∙ cos ;d ∙ ^ $ e/,A/C < =
m frequency voltage, but to the fundamental
n
(29)
frequency current. Eq. 36 can be rewritten as Eq. 37;
Changing to the reference point (t = 0) from the Eq. 38 shows the angle representation. The
zero crossing to the peak of the sinusoidal harmonic current phase angle deviation is described
waveform changes the equations without any loss by Eq. 39, which would be the input to harmonic
of general information [16]. Note that the time point current sources in PowerFactory, if phase angles are
t = 0 is an arbitrarily defined point and can be taken from measurement reports. It should be noted
changed without loss of general information. Fig. 6 that the phase angle of the fundamental frequency
shows the corresponding angles. Eqs. 30 - 33 current is the same, independently of whether the
provide the mathematical representation [16]. zero crossing (measurement) or the peak of the
sinusoidal waveform (complex number
representation in calculation) is used as the
reference point.
h ^
√2 ∙ ∙ coscd ∙ ^ $ ℎ ∙ e/,' $ Δe/,XoA f
(37)

√2 ∙ ∙ coscℎ ∙ cd' ∙ ^ $ e/,' f $ Δe/,XoA f

Δe/,XoA e/,XoA < ℎ ∙ e/,' (38)

Δe/,XoA e/,A/C $ p < 1 ∙ < ℎ ∙ e/,'

m
n
(39)

3.3.2 Interharmonics

Figure 6 Angles in complex network calculation The accurate consideration of interharmonics,

depicted as corresponding time curves i.e. non-integer harmonic orders, presents particular

]' ^ √2 ∙ ∙ coscd' ∙ ^ $ e3,' f

challenges [17, 18]. For interharmonics it is
XoA
' (30) inappropriate to use a phase angle which is related
√2 ∙ ' ∙ cos d' ∙ ^ with e3,'
XoA
0 to the fundamental frequency, because this causes

h' ^ √2 ∙ ∙ cos ;d' ∙ ^ $ e/,' $ 1 < 1 ∙ n=

the angle to change with every period of the
A/C m
'
fundamental frequency, as the period of an
√2 ∙ ∙ coscd' ∙ ^ $ e/,'
XoA
f
interharmonic is not an integer multiple of the
' (31) fundamental frequency period.
] ^ √2 ∙ ∙ cos ;d ^ $ e3, $ p<1 ∙ =
A/C m In such cases, it is appropriate to use angles for
n interharmonics as well, however the information of
(32) interest is the phase angle between current and
h ^ √2 ∙ ∙ cos ;d ^ $ e/,A/C $ p < 1 ∙ =
m voltage of the same frequency (Eq. 40), i.e. of the
n same interharmonic. In other words, the voltage of
(33) the same interharmonic is the reference for the
The phase angles of the harmonics can be angle of the interharmonic current. It should be
rewritten as shown in Eq. 34 [16], leading to a noted that the angle used for the displacement factor
simplified representation in the time functions, cos(ϕ) is defined in the opposite direction, because
Eq. 35 and 36. Eq. 34 shows the re-calculation of the conjugated complex current is used in the power
the angles from measurements (corresponding to equations.
sine function) to network calculation input e/3, e/, < e3, (40)
(corresponding to cosine function).
e XoA e A/C $ p < 1 ∙
m In addition to the phase angle, the sequence in
n
(34) which the interharmonic voltages and currents exist
] ^ √2 ∙ ∙ coscd ∙ ^ $ e3,
XoA
f
(i.e. positive, negative or zero sequence) is of
(35) interest [18]. Therefore, analysis of the angles of the
h ^ √2 ∙ ∙ coscd ∙ ^ $ e/,XoA f
phases B and C referring to the interharmonic
(36)
voltage of phase A of the same frequency as shown
As the harmonic currents are often in Eqs. 41 - 43 is of importance.

Paper submitted on April 30, 2017. Paper accepted on May 15, 2017. Page 5/9
Paper presented in Session 7-1 of ICEE 2017 in Weihai, China on July 7, 2017.
Reprint version revised by authors after ICEE 2017; text revisions are marked in green.
Revised Reprint of Paper No. 201702280000024 of
The International Conference on Electrical Engineering (ICEE 2017) Weihai, China, July 4-7, 2017

e3I, 0
.? I needed for information regarding the quality of the
(41)

e3+, e3+, < e3I,

result. The PAR is the length of the vector sum of
.? I
(42) the N phasors, in relation to the arithmetic sum of

e3s, e3s, < e3I,

.? I
the magnitudes, as defined in Eq. 47.
eu
(43)
, v,{.?|@/}/C~ (45)
∑w
w
‚ Cz' †‡tu , v,C ˆ
ƒ^ƒ„ … w ‰ h Š #i‡tu , v,C ˆ ≥0
It should be noted that Eqs. 40 - 43 are valid if
€ ∑Cz' #i‡tu , v,C ˆ
the same reference point (regardless if it is the zero
crossing or the peak of the sinusoidal waveform) is Cz'
• ∑w
w
Cz' †‡tu , v,C ˆ
€‹ $ ƒ^ƒ„ … w ‰ h Š #i‡tu , v,C ˆ Œ0
used within each period of the fundamental

• ∑ Cz' #i‡tu , v,C ˆ

frequency (see Section 3.3.1).
Cz'
3.3.3 Prevailing Angle eu , v,{.?|@/}/C~ (46)
w

•hŽ„ •Š †‡tu , v,C ˆ•

A harmonic measurement contains a number of
Cz'
discrete Fourier transform (DFT) windows. The
∑w
Cz' #i‡tu
, v,C ˆ
∙ ƒ‘’• … ‰
result for each individual DFT window can vary

y∑Cz' tu , v,C y
w
from the others. Hence an aggregation of the
measurement results is needed.
‚ ƒ‘’• …∑Cz' #i‡tu , v,C ˆ‰ h
w w

Š †‡tu , v,C ˆ ≥0
For the magnitude of the harmonic voltage or
€ y∑wCz' tu , v,C y
current the mean value of all DFT windows is taken
Cz'
•
as aggregation as shown in Eq. 44. The variable C
∑w
w
#i‡t ˆ
€<ƒ‘’• … ‰ h Š †‡tu , v,C ˆ Œ0
Cz' u , v,C
represents either voltage or current. N is the number

• y∑wCz' tu , v,C y
of aggregated DFT windows.

∙ ∑w
Cz'
tu Cz'ytu , v,C y
'
, v w y∑”•– su , v, y y∑”•–c@u , v, )*∙4u , v, fy
(44)
“E# ∑”•–ysu , v, y ∑”•–yc@u , v, )*∙4u , v, fy
(47)
For the phase angles, a simple mean value
would not reflect results correctly (for example the
If the PAR is close to unity, the phase angle
mean value of 10 degrees and 350 degrees would
does not vary significantly [10] and can be
yield 180 degrees, which is the opposite angle
considered for harmonic studies. A PAR of 0.9 or
direction). IEC 61400-21-1 [10] and FGW
higher indicates a very stable phase angle [6]. The
guideline TR3 [6] suggest the prevailing angle as an
phase angle is considered as random if the PAR is
average of the angle, which is defined by Eq. 45 [8,
smaller than 0.3 [10].
10]. An alternative formulation is provided in
The prevailing angle and PAR can be applied
Eq. 46, which ensures that the range of the
to interharmonics as well. These should be applied
prevailing angle is always within -π to +π (-180° to to the angles related to the voltage of the same
+180°). interharmonic (as described in Section 3.3.2),
because angles of interharmonics related to the
fundamental frequency quantities are random by
nature and would always result in a very small PAR.

3.4 Dependencies of Harmonic Spectra

on Operating Points

The harmonic spectrum of a VSC or a PE

interfaced RES can vary depending on the operating
point of the device. Therefore IEC 61400-21
requires harmonics measurements at 11 power bins
Figure 7 Prevailing Angle (0%, 10%, 20%, …, 90%, 100%) [1, 10].
The prevailing angle represents the direction of 3.4.1 Ideal Current Source
the complex vector sum of all phasors of the N DFT
windows, as indicated in Fig. 7. If the individual
To achieve the worst-case conditions for the
phasors have a similar direction, the prevailing
harmonic analysis, the maximum values of all
angle gives their common direction. If the
power bins must be selected for each harmonic (and
individual phasors have large deviations in their
interharmonic) for the parameterisation of the ideal
directions, the prevailing angle will tend to be
current source model (i.e. resulting in maximum
random. The Prevailing Angle Ratio (PAR) is
current injection), if only magnitudes are

Paper submitted on April 30, 2017. Paper accepted on May 15, 2017. Page 6/9
Paper presented in Session 7-1 of ICEE 2017 in Weihai, China on July 7, 2017.
Reprint version revised by authors after ICEE 2017; text revisions are marked in green.
Revised Reprint of Paper No. 201702280000024 of
The International Conference on Electrical Engineering (ICEE 2017) Weihai, China, July 4-7, 2017

considered in the harmonic study. It should be noted magnitudes differ in phase angles and therefore
that the resulting worst-case spectrum does not result in a lower superposition), or by a resonance
reflect a particular power bin, but can comprise a of the impedance of the Norton or Thevenin
mix of different power bins at different orders (or equivalent (inner impedance) with the grid
frequencies). impedance (outer impedance). This means that if
If the harmonic study is carried out using a the impedance varies with power bins, the related
complex harmonic load flow calculation and the spectrum of the inner source as well as the
sources are considered with phase angles, the worst- frequency-dependent inner impedance must be
case condition at the node of interest not only considered for each power bin in a harmonic study.
depends on the magnitudes of the individual Again, the harmonic load flow calculation must be
sources, but also on their phase angles. See repeated for all power bins so as not to miss any
Section 3.4.2 for details. worst-case scenario if the Norton or Thevenin
equivalent is used.
3.4.2 Norton and Thevenin Equivalents In PowerFactory, network model variations
can be used for the representation of the individual
If the frequency-dependent impedance of the input data referring to power bins. Scripting or
Norton or Thevenin equivalent is the same for each automatic task execution is recommended to carry
power bin, the same approach as for the ideal out all necessary harmonic load flow calculations,
current source model also applies to the Norton or either in a loop or even in parallel, distributed on
Thevenin equivalent. However, from the several cores of the computer [19].
measurements of currents and voltages at the
terminal, the values for the inner source have to be 3.5 Parameterisation of Models based on
calculated first for each power bin, as explained in Incomplete Measurements
Section 3.2. Subsequently, the maximum value
(max. magnitude) of the inner source can be As described in Section 3.2, measurement of
selected per power bin to achieve the worst-case the harmonic current and voltage (both with
spectrum. It should be noted that this approach is magnitude and angle) is needed to parameterise the
only possible if only magnitudes are considered for Norton or Thevenin equivalent, unless the
parameterisation of the sources, but phase angles frequency-dependent impedance of the entire set-up
are neglected, and summation laws are used for the is known. If these measurements are not performed
summation of harmonic currents and voltages at the and only the harmonic current magnitudes are
point of connection or point of common coupling known, the models cannot be parameterised
(i.e. the node of interest) in the harmonic study. If sufficiently.
the harmonic study is carried out using a complex Taking the harmonic current measured at the
harmonic load flow calculation and the sources are terminal of the device as the inner current of the
considered with phase angles, the worst-case Norton equivalent or as short-circuit current for
condition at the node of interest does not only parameterising the inner source of the Thevenin
depend on the magnitudes of the individual sources, equivalent following Eq. 5 (or using Eq. 16 but
but also on their phase angles. This also applies to entering the harmonic voltage at the terminal as
the ideal current source model. Therefore, if phase zero), will in most cases lead to an underestimation
angles of sources are considered and the harmonic of the harmonic injection in the harmonic study.
sources in the network are not all of the same device This is because the magnitude of the inner source
type, each spectrum of power bin has to be must be larger in order to achieve the measured
considered in the harmonic study. This means that harmonic currents at the terminal.
the harmonic load flow calculation must be If no harmonic voltage measurement is
repeated for all power bins in order to not miss any available, a rough estimation is possible by
worst-case scenario. reproducing the impedances of the measurement
Assuming that the frequency-dependent set-up as accurately as possible for the
impedance of the Norton or Thevenin equivalent parameterisation. This should be done using the
may be different for each power bin (this may apply same assumptions for the frequency characteristics
if components of the device are non-linear or if the of the impedances as will be used later in the
controller transfer function changes for different harmonic study. With these assumptions, Eq. 18 or
operating points), for parameterisation of the 19 can be used to estimate the values of the inner
Norton or Thevenin equivalent, the worst-case sources.
conditions are very difficult to select. This is
because a worst case can be caused by maximum
current injection, by injection of different sources
with the same phase angle (not necessarily with
maximum magnitudes, if injections of max.

Paper submitted on April 30, 2017. Paper accepted on May 15, 2017. Page 7/9
Paper presented in Session 7-1 of ICEE 2017 in Weihai, China on July 7, 2017.
Reprint version revised by authors after ICEE 2017; text revisions are marked in green.
Revised Reprint of Paper No. 201702280000024 of
The International Conference on Electrical Engineering (ICEE 2017) Weihai, China, July 4-7, 2017

4. Conclusions Frequency Domain, 9th International

Conference on Compatibility and Power
This paper provides a detailed discussion of the Electronics, Portugal, 2015
Norton and Thevenin equivalent for representation [10] IEC 61400-21-1, Wind energy generation
of PE interfaced RES and VSC of HVDC system in systems - Part 21-1: Measurement and
harmonic studies. Correct parameterisation of the assessment of electrical characteristics - Wind
models is crucial for their use in harmonic studies. turbines, Committee Draft for Vote (CDV),
As the complex harmonic currents and voltages (or 2017 (IEC CDV 61400-21-1:2017)
impedances of the whole test set-up) are needed for [11] Cigré WG 14.39 Brochure 139: Guide to the
the parameterisation, consideration of the phase Specification and Design Evaluation of AC
angles is very important. Procedures to evaluate Filters for HVDC Systems, April 1999
phase angles from measurements and to transform [12] J. Sun: Impedance-based stability criterion for
them in a correct way for use in network grid-connected inverters, IEEE Trans. on
calculations are described in detail. The formulas Power Electronics, Vol. 26, No. 11, Nov. 2011,
needed for parameterisation of the models are pp. 3075-3078
provided. [13] L. Shuai, Ł. H. Kocewiak, K. H. Jensen:
Application of Type 4 Wind Turbine Harmonic
References Model for Wind Power Plant Harmonic Study,
15th Wind Integration Workshop (Intl.
[1] IEC 61400-21, Wind turbines – Part 21: Workshop on Large-Scale Integration of Wind
Measurement and assessment of power quality Power into Power Systems…), Vienna, 2016
characteristics of grid connected wind turbines. [14] M. Aberhard, M. Meyer, C. Courtois:
Edition 2.0, 2008 (IEC 61400-21:2008) EN 50388-2 – der neue Teil 2: Stabilität und
[2] FGW TR8, Technical Guidelines for Power Harmonische, Elektrische Bahnen, Vol. 112,
Generating Units – Part 8: Certification of the No. 10, 2014, pp. 574-581
Electrical Characteristics of Power Generating [15] IEC Technical Report 61000-3-6: Electro-
Units and Farms in the Medium-, High- and magnetic compatibility (EMC) – Part 3-6:
Highest-voltage Grids, Rev. 8, Dec. 2016, Limits – Assessment of emission limits for the
FGW e.V., Germany connection of distorting installations to MV,
[3] F. Santjer, B. Weise, T. Pausch, J. Brombach: HV and EHV power systems, Edition 2.0,
Aspects for Improvement of Measurement and 2008-02 (IEC/TR 61000-3-6:2008)
Assessment Procedures of Harmonic Emission [16] M. H.J. Bollen, I. Y.H. Gu: Signal Processing
of Wind Power Plants, DEWEK 2015, Bremen, of Power Quality Disturbances, IEEE Press,
Germany, 19-20 May 2015 Wiley-Interscience, 2006, ISBN 978-0-471-
[4] M. Cespedes, J. Sun: Modeling and Mitigation 73168-9
of Harmonic Resonance Between Wind [17] IEEE Task Force on Harmonics Modeling and
Turbines and the Grid, IEEE Energy Simulation (A. Testa, et al.): Interharmonics:
Conversion Congress and Exposition, 2011 Theory and Modeling, IEEE Trans. on Power
[5] J. Sun: Modeling and Mitigation of Harmonics Delivery, Vol. 22, No. 4, 2007, pp. 2335-2348
and Harmonic Resonance Involving Wind [18] D. Zhang, W. Xu, Y. Liu: On the Phase
Inverters, 12th Wind Integration Workshop Sequence Characteristics of Interharmonics,
(International Workshop on Large-Scale IEEE Trans. on Power Delivery, Vol. 20, No. 4,
Integration of Wind Power into Power Systems Oct. 2005, pp. 2563-2569
as well as on Transmission Networks for [19] DIgSILENT PowerFactory 2017 User Manual,
Offshore Wind Power Plants), London, 2013 DIgSILENT GmbH, Germany, 2017
[6] FGW TR3, Technical Guidelines for Power
Generating Units – Part 3: Determination of Authors’ brief introduction and contact
electrical characteristics of power generating information:
units connected to MV, HV and EHV grids. Dipl.-Ing. Bernd Weise (e-mail address:
Rev. 24, March 2016, FGW e.V., Germany b.weise@digsilent.de) obtained his Dipl.-Ing.
[7] M. Cespedes, J. Sun: Impedance Modeling and Degree (diploma degree) in electrical engineering
Analysis of Grid-Connected Voltage-Source from the Technical University of Berlin, Germany,
Converters, IEEE Trans. on Power Electronics, in 2002. In 2007 he joined DIgSILENT GmbH,
Vol. 29, No. 3, March 2014, pp. 1254-1261 Gomaringen, Germany, as an application and
[8] IEC Technical Report 61400-21-3, Wind consulting engineer. In 2013 he became manager of
turbine harmonic model and its application, the “Application Engineering” department. Since
Draft Version, 2016 (IEC DC 61400-21-3 TR) 2016 he leads the “Power Generation and
[9] K. Malekian, A. Gürlek, W. Schufft: Analysis Conversion” department, which is responsible for
and Modeling of Time-Varying Harmonics in application engineering and consulting in the field

Paper submitted on April 30, 2017. Paper accepted on May 15, 2017. Page 8/9
Paper presented in Session 7-1 of ICEE 2017 in Weihai, China on July 7, 2017.
Reprint version revised by authors after ICEE 2017; text revisions are marked in green.
Revised Reprint of Paper No. 201702280000024 of
The International Conference on Electrical Engineering (ICEE 2017) Weihai, China, July 4-7, 2017

of electric power generation and power conversion

systems within DIgSILENT. Mr. Weise represents
DIgSILENT in working groups of FGW e.V.
(encourage-society for wind energy and other
renewable energies), Germany, and is chairman of
a working group on Harmonic Models within FGW
e.V.
Dr. Genevieve Lietz (e-mail address:
g.lietz@digsilent.de) obtained the Bachelor of
Engineering (Hons) from the School of Electrical
and Computer Systems Engineering at RMIT
University, Melbourne, Australia, in 2001. She
obtained the PhD from the Faculty of Engineering
at the University of Melbourne in 2007. From 2006-
2007 she worked as a power systems consulting
engineer at DIgSILENT Pacific, Melbourne,
Australia, and since 2007 works as a power systems
engineer and software developer at DIgSILENT
GmbH, Gomaringen, Germany. Her primary areas
of interest are harmonic analysis and modelling, and
EMT cable/line modelling. She is a member of the
CIGRÉ Joint Working Group C4/B4.38 on
Network Modelling for Harmonic Studies.

Paper submitted on April 30, 2017. Paper accepted on May 15, 2017. Page 9/9
Paper presented in Session 7-1 of ICEE 2017 in Weihai, China on July 7, 2017.
Reprint version revised by authors after ICEE 2017; text revisions are marked in green.