Experimental Verification of a Virtual Synchronous

Generator Control Concept

Igor Cvetkovic, Dushan Boroyevich, Rolando Burgos,

Yi-Hsun Hsieh, Fred C. Lee, Chi Li Paolo Mattavelli
Center for Power Electronics Systems (CPES) Department of Management and Engineering
The Bradley Dep.of Electrical and Computer Engineering University of Padova - Vicenza, Italy
Virginia Tech, Blacksburg, VA 24061, USA Adjunct professor at CPES
Email: [email protected]

Abstract—This paper shows a comprehensive approach to a side-by-side.

synchronous machine emulation in the power electronics
converters. It shows how to establish the electromechanical- II. CONVERTER-LIKE MACHINE MODEL
electrical duality, re-structure the well-known machine model to The machine voltage equation (1) can be rewritten into the
resemble widely used power converter dynamic model, perform given form, defining flux derivatives per rotor position rather
experiments on the 30 kW synchronous machine, and at the end, than time (the main reason for this reformatting is to factor out
verify the model. Additionally, initial result of the concept the term dependent on the machine angular speed):
implementation into the power electronics converter will also be
shown for completeness. This paper deals with the four pole
synchronous machine, but presents a generalized approach on d d d
vd  rs id  e q   rs id   p( d  q )
how to control power converters as either isotropic or anisotropic dt d e
synchronous machines by entirely including dynamic effects of (1)
damper and field windings. d q d q
vq  rs iq  e d   rs iq   p(  d )
dt d e
Keywords — virtual synchronous machine; modeling;
synchronisation; experimental verification; Ω vF

rF
I. INTRODUCTION Tm kf J Te
iF
This paper essentially presents an addendum to the LF
reference [10], further generalizing the virtual machine model, Mechanical 3
subsystem rD iD MDF 2 MF id rs
and including some results of the concept experimental MF
verification. It also complements some of the previously
LD MD Ld vd
published research in this area [1]-[9]. 3M
2 D
One of the important steps in the attempt of finding the best

+
-
Electrical ωeψq
way to control power converters as synchronous machines is to subsystem rQ iQ iq rs
develop an electrical model of the machine completely dual to
commonly used – electromechanical, and then restructure it MQ
vq
LQ Lq
mathematically to fully resemble generic average model of the 3M
2 Q
power electronics converter. This would offer better
+
-

ωeψd
understanding on which power converter parameters relate to
which parameters of cylindrical or salient pole machine. One of
the goals in the effort of implementing machine operation in 
the power electronics converter, is to use us much as possible
id r L
resources/components of the chosen converter, leaving only
+
-

complementary parts to be implemented into the converter + ddvdc ωLiq

vdc Idc vd
control algorithm. This could be achieved by, first, defining -
electrical parameters that feature physical/mathematical duality L
Is G C iq r
with corresponding mechanical parameters, and second, by d di d dqiq
+
-

mathematically restructuring synchronous machine model to + dqvdc ωLid

vq
resemble power converter model (without losing generality) – -
as Figure 1 illustrates by showing generic grid-interface
converter [11] and synchronous generator model [12]-[13] Fig. 1. Establishing equivalence between two models;
Synchronous machine (top), and converter (bottom)
This work was primarily supported by CPES Wide-bandgap High-Power
Converters and Systems (WBG-HPCS) mini-consortium

978-1-5090-1815-4/16/$31.00 ©2016 IEEE

 Furthermore, active power at the machine terminals
comprises following components: joules losses, rate of change  d  ( Ld  L)id  M F iF  M DiD   d  Lid
of the energy accumulated in the magnetic field, and (5)
mechanical power converted to electrical:  q  ( Lq  L)iq  M QiQ   q  Liq
d d d q
Pdq  (id2  iq2 )  rs  id  iq  e ( d iq  qid )  This can be applied to the stator equations:
dt dt
(2)
 (i  i )  rs  P gap
2
d
2
q did d
vd  rs id  L  e Liq   p( d   q )
dt d e
Assuming lossless electromechanical conversion from rotor (6)
to stator, all of the power Pdq (except joules losses) is delivered d iq d q
from the rotor (Pgap). Lumping this power into the form of an vq  rs iq  L  e Lid   p(   d )
dt d e
“equivalent” torque, it could be written:
and to equivalent torque:
1  d d
Pgap d q  e
Te'    id  iq   ( d iq  qid )  d d d
   dt dt   Te'  id p(  q )  iq p( q  d ) 
(3) de de
d d d (7)
 id p(  q )  iq p( q  d ) d d d
de de  id p(  q )  iq p( q  d )
de de
Combining equations (1), and (3), the form that matches the
converter dc-ac conversion stage model is obtained. The new The above equation (7) is obtained using the following:
machine model is shown in Figure 2, where:
 d iq  q id   Lid iq  d iq  Liq id  q id 
 d    d iq  q id (8)
 d'  p d  q 
d
 e 
(4)
and assuming that
 d 
  p q   d 
'
did diq

q
d
 e  id  iq 0 (9)
d e d e
It can be noticed that machine model from Figure 2
structurally resembles the converter model from Figure 1 quote has quite insignificant impact to the system dynamic.
well, however, the difference is absence of the converter output
Fluxes ψd and ψq can be expressed as functions of id, iq, and
inductor. Reference [10] addresses this issue by extracting the
vF [12]:
leakage inductance Lγ out of the flux equations due to the
simple fact that it does not contribute the process of
electromagnetic power conversion in the machine (this is  d   Ld ( s )  id  G ( s )  vF 
addressed in equations (8), (9), (11), and (13) in the paper
[10]). Another option of extracting arbitrary inductance L will
be shown here.


Ld rD (rF  sLF )  LD rF s  LD LF s 2  M FD
2
s2
 id 

The fluxes ψd and ψq can be rewritten as shown in (5) (rD  sLD )(rF  sLF )  M FD2
s2
(without loss of generality).


3 / 2 2M D M F M FD s 2  M F2 (rD  sLD ) s
 id 
 (10)

id rs (rD  sLD )(rF  sLF )  M FD2

s2

+ ψ'd Ω M F  (rD  sLD )  sM FD M D

Ω Te' - vd   vF
(rD  sLD )(rF  sLF )  M FD
2
s2
Tm kf J iq rs
ψd' id ψq' iq
+ ψ'q Ω  q   Lq ( s )  iq 
p -
vq
ωe (11)
Lq (rQ  sLQ )  3 / 2 M Q2 s
θe   iq
1/s rQ  sLQ

Fig. 2. Restructured synchronous machine model

 It is obvious that operational inductances in (10), and (11) and Newton’s motion equation
contain all machine parameters illustrated in Figure 1. These
can be rewritten using (5): d
J  Tm  Te  k f  (15)
dt
 d  ( Ld ( s )  L)  id  G ( s)  vF can be directly equivalented as seen in Figure 4.
(12)
 q  ( Lq ( s )  L)  iq This practically means that dc-link voltage vdc [V] has to be
numerically equal to Ω [rad/s], input current Is [A]
The expression (4) now gets the new form that will be used numerically equal to the input torque Tm [Nm], etc., in order to
here: obtain purely equivalent electrical equivalent of the
mechanical system. However, due to the fact that in voltage
 d  source converters dc-link voltage has to be higher than peak
 ' d  p d  q 
 d e  line-to-line voltage on the ac side (modulation index lower
(13) than 1), it could be expected that vdc [V] has values multiple
 d  times higher (numerically) than Ω [rad/s].
  p q   d 
'
q Reference [10] already dealt with this, when, by applying
 d e  law of conservation of power/energy, it was shown how
As it can be seen in (12), the only control variable is increase in dc-link voltage affected other circuit parameters.
actually vF, while id, and iq, are measured and fed back for flux The constant K can be defined as
calculation. Figure 3 shows the final restructured machine vdc
model that will be used as a basis for this work. K (16)
s
III. MECHANICAL-ELECTRICAL SUBSYSTEM DUALITY where ωs is the angular frequency of the system (grid) the
converter connects to. In most of the cases that would be 377
Per [14], the following table describes duality of [rad/s] for 60 Hz, or 314 [rad/s] for 50 Hz system.
mechanical/electrical units.
It has quite known that in machines, there is a similar
Mechanical systems Electrical systems
relationship between mechanical and electrical angular
frequency:
Torque T ≡ Current Idc
e s
p  (17)
Angular speed Ω ≡ Voltage vdc  
Moment of Inertia J ≡ Capacitance C Figure 5 depicts above described relationships.
Friction kf ≡ Conductance G With all of this, (18) shows how choice of the dc-link
voltage level influences other parameters; in other words, in
the attempt to completely mimic machine dynamics with the
With this, dc-link equation converter (one-to-one mapping), one has to choose the
dvdc following values for input current, output admittance, and
C  I s  I dc  Gvdc (14) capacitance.
dt
Tm T k J
Is  , I dc  e , G  2 f 2 , C  2 2 (18)
id rs L pK pK p K p K
+
-

+ ψ'Δd Ω ωeLiq v
Ω Te' - d
Ω vdc
Tm kf J iq rs L
' id ψΔq
ψΔd ' iq Tm kf J Te ↔ Is G C Idc
+
-

+ ψ'Δq Ω ωeLid v
p - q
ωe Mechanical Electrical
subsystem subsystem
θe
1/s Fig. 4. Subsystem duality
id iq vF

' dψΔd
ψΔd p ψΔq
dθe – (Ld (s) – L)·id + G (s)·vF Ω [rad/s]= ωs /p ωs [rad/s] vdc [V] = K∙ ωs

' dψΔq – (Lq (s) – L)·iq

ψΔq p ψ
dθe + Δd Fig. 5. Graphical relationship between dc-link voltage, system angular
frequency, and actual mechanical angular frequency of the synchronous
Fig. 3. Final, restructured synchronous machine model machine
 Once connected to the grid, operating at the particular
operating point, synchronous machines self-synchronize to the
system, and precisely follow changes of the system grid
frequency. This is their well-known inherent feature due to the
physics behind the power balance. Newton’s motion equation
(3.11) describes this feature. However, this feature can be
replicated on the converters too, allowing them to
“synchronize” their dc-link voltages with the system
frequency.
After any perturbation, synchronous machines set their
internal angle θe (consequently - power angle δ) to the level
required to maintain the power balance between mechanical
and electrical systems. Hence, it is possible to obtain internal
converter angle in the same way it is obtained in the
synchronous machines – only now by integrating converter’s
dc-link voltage:

d e v
v dc  K  e    e   e   dc dt   o (19)
dt K
Figure 3 showed restructured synchronous machine model
that features a resemblance with the converter model from
Figure 1. The same figure now clearly indicates what has to be
done to operate converter as a machine; Consequently,
Figure 6 now shows model of a converter that mimics Fig. 7. 30 kW synchronous generator testbed and
machine behavior. 200 kVA 3-level ANPC converter testbed system

It is evident that converter duty-cycles dd and dq should be

controlled in the following way in order to mimic machine IV. SYNCHRONOUS MACHINE MODEL VERIFICATION
operation: In order to perform verification of the virtual synchronous
1  d d  machine model, and proceed with hardware implementation of
d d    q 
K  d e the concept, it is necessary first to match the machine model

(20) from Figure 3 with the experimental results obtained using
real synchronous machine. The 30 kW testbed used in this
1  d q 
d q    d  work comprises a 30 HP variable speed drive, 22 kW
K  d e  induction motor, and 27 kW synchronous generator with field
and damper windings. The testbed is a part of the CPES’
Wide-bandgap High-Power Converters and Systems (WBG-
HPCS) architecture, and is showed in Figure 7 together with
the 200 kVA 3-level converter here used for the concept
kf
G= C= 2J 2 id r L verification purpose [16]. Limited information about the
p 2K 2 pK
+
-

machine parameters had been provided by the manufacturer,

+ ddvdc ωeLiq v
vdc Idc - d thus the process of machine characterization (parameter
extraction) had to be done according to IEEE guide for
Tm iq r L standstill frequency response (SSFR) testing - Std. 115A-
G C
pK d d id d q iq
+

1987 [15].
-

+ dqvdc ωeLid v
q
1/K - The IEEE SSFR standard proposes measurement of several
ωe
frequency response curves while machine is at standstill.
θe
1/s Some type of a frequency response analyzer with an amplifier
id iq vF is required to perturb the machine at the frequency range from
a few mHz to a few hundred Hz. Most readily available
dψΔd
dd 1/K ψΔq amplifiers operate only in the audible range (from about 20 Hz
dθe – (Ld (s) – L)·id + G (s)·vF
to 20 kHz), and may not be the best choice for machine
dψΔq – (Lq (s) – L)·iq characterization. Simple signal (function) generator was used
dq 1/K ψ
dθe + Δd here for this characterization task. The machine terminals were
connected in series with the 50 Ω impedance, thus overloading
Fig. 6. Virtual synchronous machine the unit was prevented. Currents in the level 50-70mA were
 fact the impedance between terminal a, and terminal b was
a
measured (Figure 8), the obtained measurement result
iinj iF represents two Zd impedances in series:

1 vinj ( j ) 1
Z d ( j )   Z de ( j ) (24)
2 iinj ( j ) 2
vinj
Simultaneously with Z de ( j ), jGe ( j ) can be measured.
That transfer function has a form
iF ( j )
b jGe ( j )  (25)
iinj ( j )
vF 0

2 i ( j ) vd ( j )
j G ( j )  jGe ( j )  F  (26)
Fig. 8. IEEE Std. 115A-1987 proposed measurement setup for machine 3 id ( j ) v 0
v F ( j ) i 0
F d
characterization in d-q-axes
As seen in Figure 8, current iF can be measured at the same
injected into machine, while terminal voltages were kept in the time as iinj. It is easy to prove that two transfer functions from
range of a few volts (10 V peak-to-peak max). (26) are equal. If rotor of the machine is aligned in q-axis, and
measurement set up as shown in Figure 8, output impedance in
Figure 8 illustrates the setup and measured variables. The q-axis is obtained point-by-point, by injecting perturbation
standard proposes a simple technique to position (by hand) the points from 10 mHz to 1000 Hz.
rotor of the machine in order to align it with d-, or q- axis,
before performing the frequency sweep point-by-point (only Similarly as in the case of d-axis impedance, frequency
necessary steps and explanations will be given here for response between terminals a, and b was measured, and the
obtained measurement result represents two Zq impedances in
inclusiveness).
series:
As seen in the chapter (12), three frequency response curves
are needed in order to obtain a full dynamic response of the
machine: 1 vinj ( j ) 1
Z q ( j )   Z qe ( j ) (27)
2 iinj ( j ) 2
Ld (s)– the direct-axis operational inductance. It represents the
ratio of the direct-axis armature flux linkages to the direct-axis Ld(jω), G(jω), and Lq(jω), can now be obtained in the
current with the field winding shorted: following way:
~ Z d ( j )  Ra
Ld ( s)   ~d (21) Ld ( j )  (28)
id v 0 j
F

G (s) – the armature to field transfer function. It represents the jG ( j )

G ( j )  (29)
ratio of the direct-axis armature flux linkages to the field j
voltage, with the open armature:
~ Z q ( j )  Ra
G ( s)  ~ d (22) Lq ( j )  (30)
v F id  0
j
where Ra is the stator resistance. IEEE SSFR standard suggests
Lq (s) – the quadrature-axis operational inductance. It
that Ω-meter with very good accuracy is used (measurement
represents the ratio of the quadrature-axis armature flux
resolution of 1 part in 1000 is required), as this parameter
linkages to the quadrature-axis current: significantly influences calculated operational inductances
~ Ld(s), and Lq(s), in the low frequency range (consequently, this
Lq (s)   ~q (23) influences synchronous inductances Ld, and Lq).
iq
Transfer functions (21), (22), and (23), cannot be measured Using (28) – (30), the required transfer functions can finally
directly. Instead, they are calculated out from the measured be extracted from the measured data. Results are shown in
frequency responses, as it will be seen later in this section. Figure 9.
If rotor of the machine is aligned in d-axis, and (1  s 7.289)(1  s 0.062)
measurement set up as shown in Figure 8, output impedance in Ld ( s )  0.0438 (31)
(1  s15.618)(1  s 0.450)
d-axis is here obtained point-by-point, by injecting perturbation
with resolution of about 10 frequency points per decade,
ranging from 10 mHz to 1000 Hz (in 51 points). Due to the
 (1  s 0.016)
Lq ( s )  0.0135 (33)
(1  s 0.038)
Magnitude (dB)

Expressions (31) – (33) actually contain information about

machine time constants in the following form
(1  sTd' )(1  sTd'' )
Ld ( s )  Ld (34)
(1  sTdo' )(1  sTdo'' )

(1  sTkd )
G ( s)  Go (35)
(1  sTdo' )(1  sTdo'' )
Phase (deg)

(1  sTq'' )
Lq ( s )  Lq (36)
(1  sTqo'' )

Above transfer functions unquestionably match chosen

machine model from Figure 1, as rotor of the synchronous
machines can be represented with multiple damper windings
in both d-, and q- axis, the order of transfer functions (34) -
(36) (and (31) - (33)) would reflect that; d-axis here comprises
Magnitude (dB)

one field, and one damper winding, thus transfer functions

Ld(s) and G(s) are second order (with the same denominator),
while q-axis features only one damper winding, hence Lq(s) is
the first order transfer function.
V. SYNCHRONOUS MACHINE CONVERTER-LIKE
MODEL VERIFICATION
Phase (deg)

Having performed frequency response measurements

reported above, machine model from Figure 3 can now be
experimentally verified, by performing transient response tests
of the machine. Variety of different tests have been completed
with the generator loaded with the resistive load. The variable
speed drive was set to the speed mode with the reference speed
of 60 Hz (377 rad/s). Figure 10 shows experimental results of
one of the test where at the time 1.5 s load step was made,
changing the machine resistive load from 7.1 Ω to 5.6 Ω;
following this transient, field voltage step was made at 5 s,
stepping vF from 6.85 V to 3.85 V, and recovering it back to
Magnitude (dB)

6.85 V at 9.7 s.
In order to perform machine model verifications, d-q
currents and voltages are needed; id, iq, vd, and vq cannot be
measured directly, but can be calculated out from the measured
iabc, and vabc. Angle θe for d-q transformation (19) was easily
obtained from the measured speed (lower-left subfigure in
Figure 10). Initial angle θo was obtained via d-axis position
sensor installed at the shaft of the synchronous generator.
Phase (deg)

Calculated d-q variables are showed in Figure 11.

If used exactly as identified, transfer functions Ld(s), G(s),
and Lq(s) do not automatically give very good matching with
the experimental results. There are many reasons for that –
measurement accuracy, error in determining the exact machine
resistance, insufficient core excitation, etc. Numerous authors
suggest that obtained transfer functions (31)-(33) unfortunately
Fig. 9. From top: d-axis operational inductance; Armature to field freq.
response; q-axis operational inductance;
require some minor tuning before satisfactory results could be
dots – frequency response data, blue – s-domain identification achieved. Shown in (37) – (39) are adjusted transfer functions.

(1  s15.830)
G( s)  0.162 (32)
(1  s15.618)(1  s0.450)
 Lab. C
DC-AC
L RL
Power vdc Power
Electronics
Supply Converter

Modulation θe
Tabc/dq vF
θe T id iq
dq/abc

– (Ld (s) – L)·id + G (s)·vF

1/K
ωe – (Lq (s) – L)·iq
θe
1/s dd
dψΔd
1/K ψΔq
dθe
dq dψΔq
1/K ψ
dθe + Δd

Fig. 12. Virtual synchronous machine – as implemented into the power

Fig. 10. Experimental results of the synchronous machine testing electronics converter

was obtained, transfer functions (37)-(39) may require some

more refinement.

VI. INITIAL RESULTS OF THE VIRTUAL SYNCHRONOUS

MACHINE HARDWARE IMPLEMENTATION
The virtual synchronous machine concept has been initially
implemented into a 200 kVA three-level Active Neutral Point-
Clamped (ANPC) converter [16] shown in Figure 7. Although
it features several times higher rated power than machine
modeled and characterized here, this power converter was
chosen for experimental verification of the virtual synchronous

Fig. 11. Calculated d-q currents and voltages: blue – from the generator
testing; red – from converter-like machine model

(1  s13.245)(1  s0.043)
Ld ( s)  0.0438 (37)
(1  s13.369)(1  s0.669)

(1  s13.974)
G ( s )  0.205 (38)
(1  s13.369)(1  s0.669)

(1  s 0.012)
Lq ( s )  0.0178 (39)
(1  s 0.036)

Finally, Figure (11) shows currents id, iq, and voltages vd, vq
obtained from the machine model plotted over the experimental Fig. 13. d-q currents and voltages obtained from converter, plotted
results from Figure (10). Although relatively good matching together with the virtual machine model (green –experimental results;
black – model results)
machine concept, and will be later used for system-level [2] Z. Qing-Chang and G. Weiss, "Synchronverters: Inverters That Mimic
studies within CPES’ Wide-bandgap High-Power Converters Synchronous Generators," Industrial Electronics, IEEE Transactions on,
vol. 58, pp. 1259-1267, 2011.
and Systems (WBG-HPCS) architecture. The virtual
[3] J. Driesen and K. Visscher, "Virtual synchronous generators," in Power
synchronous machine control structure was implemented in and Energy Society General Meeting - Conversion and Delivery of
this converter, using transfer functions (37)– (39), as illustrated Electrical Energy in the 21st Century, 2008 IEEE, 2008, pp. 1-3.
above in Figure 12. Repeated are the same transients described [4] Z. Lidong, L. Harnefors, and H. P. Nee, "Power-Synchronization
earlier for the generator (Figure 10), with the only difference of Control of Grid-Connected Voltage-Source Converters," Power Systems,
being conducted as a separate, consecutive tests. Only load step IEEE Transactions on, vol. 25, pp. 809-820, 2010.
transient results are shown in this paper overlaid in Figure 13 [5] M. Torres and L. A. C. Lopes, "Virtual synchronous generator control in
with previously verified virtual machine model response autonomous wind-diesel power systems," in Electrical Power & Energy
(Figure 11). The sensor gain offsets were first applied to the Conference (EPEC), 2009 IEEE, 2009, pp. 1-6.
measured waveforms before being plotted in Figure 13. [6] N. Phi-Long, Z. Qing-Chang, F. Blaabjerg, and J. M. Guerrero,
"Synchronverter-based operation of STATCOM to Mimic Synchronous
The excitation voltage vF step tests (second and third Condensers," in Industrial Electronics and Applications (ICIEA), 2012
transient steps in Figure 11) are not shown in this paper as they 7th IEEE Conference on, 2012, pp. 942-947.
did not provide as good matching with the model, due to the [7] P. Rodriguez, I. Candela, and A. Luna, "Control of PV generation
systems using the synchronous power controller," in Energy Conversion
fact that the power supply (“ideal” voltage source) was used as Congress and Exposition (ECCE), 2013 IEEE, 2013, pp. 993-998.
converter’s front-end (as depicted in Figure 12). In the case of [8] U. Tamrakar, D. Galipeau, R. Tonkoski, and I. Tamrakar, "Improving
the synchronous machine, front-end (induction motor with the transient stability of photovoltaic-hydro microgrids using virtual
motor-drive) introduces a non-negligible dynamic impact on synchronous machines," in PowerTech, 2015 IEEE Eindhoven, 2015,
the behavior of the machine, hence ideal voltage source could pp. 1-6.
not have been used for experimental verification, leaving this [9] S. D. Arco, J. A. Suul, and O. B. Fosso, "Small-signal modelling and
part of the concept verification for future publications where parametric sensitivity of a Virtual Synchronous Machine," in Power
estimated output impedance of the synchronous machine’s Systems Computation Conference (PSCC), 2014, 2014, pp. 1-9.
front-end is to be implemented into converter’s front-end (dc- [10] I. Cvetkovic, D. Boroyevich, R. Burgos, C. Li, and P. Mattavelli,
"Modeling and control of grid-connected voltage-source converters
dc boost converter). emulating isotropic and anisotropic synchronous machines," in Control
and Modeling for Power Electronics (COMPEL), 2015 IEEE 16th
VII. CONCLUSIONS Workshop on, 2015, pp. 1-5.
[11] Silva Hiti, “Modeling and control of three-phase PWM converters,”
August 1995, PhD dissertation.
This paper describes a generic d-q model of the grid-
[12] P. Kundur, Power System Stability and Control, 1994, McGraw-Hill.
interface converter that behaves as a synchronous machine of
[13] S. N. Vukosavic, Electrical Machines, 2013, Springer.
any type, and showed a comprehensive approach of four-pole
[14] H.F. Olson, Dynamical Analogies, 2nd ed., Van Nostrand, pp. 27–29.,
synchronous machine emulation. Model verification results 1958.
were showed, as well as some initial experimental results from [15] "IEEE Standard Procedures for Obtaining Synchronous Machine
the concept implementation in the power converter. Parameters by Standstill Frequency Response Testing”IEEE Std 115A-
1987, p. 0_1, 1987.
VIII. REFERENCES [16] Y. Jiao and F. C. Lee, "New Modulation Scheme for Three-Level Active
[1] H. P. Beck and R. Hesse, "Virtual synchronous machine," in Electrical Neutral-Point-Clamped Converter With Loss and Stress Reduction,"
Power Quality and Utilisation, 2007. EPQU 2007. 9th International IEEE Transactions on Industrial Electronics, vol. 62, pp. 5468-5479,
Conference on, 2007, pp. 1-6. 2015.