This paper describes objective technical results and analysis.

Any subjective views or opinions that might be expressed

in the paper do not necessarily represent the views of the U.S. Department of Energy or the United States Government.

SAND2019-7187C

Simulation of Grid-Forming Inverters Dynamic Models using a Power

Hardware-in-the-Loop Testbed.
Javier Hernandez-Alvidrezl, Adam Summers% Matthew J. Reno% Jack Flicker% Nataraj Pragallapati2
1Sandia National Laboratories, Albuquerque, NM,87123, USA.
2New Mexico State University, Las Cruces, NM,88001, USA.

Abstract — Modern power grids include a variety of renewable peer-to-peer coordination between devices. This is typically
Distributed Energy Resources(DERs)as a strategy to comply with accomplished by the use ofdroop control schemes in frequency
new environmental and renewable portfolio standards (RPSs)
imposed by state and federal agencies. Typically, DERs include the
and voltage [7].
use of power electronic(PE)interfaces to interact with the power Recently, a workshop related to GFMIs for low inertia power
grid. Recently this interaction has not only been focused on systems [8] gathered members of academia, researchers of
supplying maximum available energy, but also on supporting the national laboratories, utility engineers, and representatives of
power grid under abnormal conditions such as low inverter and protective relaying manufacturers. In this
voltage/frequency conditions or non-unity power factor. Over the
last few years, grid-following inverters(GFLIs) have proven their
workshop, presenters addressed the state-of-the-art in GFMI
value while providing these ancillary grid-support services either power electronics, reliability, and control. Furthermore, design
at residential or utility scale. However, the use of grid-forming engineers of protective relaying equipment expressed their
inverters(GFMIs) is gaining momentum as the penetration-level concerns about the negative influence of the Phase Locked
of DERs increases and system inertia decreases. Under abnormal Loop (PLL) of GFLIs in terms of protection, loss of
operating conditions, GFMIs tend to better preserve grid stability
due to their intrinsic ability to balance loads without the aid of
synchronism, and inability to supply negative sequence current
coordination controls. In order to gain and propose fundamental during fault scenarios. Even though protective relay
insights into the interfacing of GFMIs to real time simulation, this manufacturers are able to design protective relays for both types
paper analyzes the dynamics of two different GFMI simulation ofinverters either at transmission or distribution level [9],[10];
models in terms of stability and load changes using a Power design engineers clearly expressed their preference towards
Hardware-in-the-Loop(PHIL)simulation testbed.
protection of GFMIs scenarios, since these inverters inherently
Index Terms — grid-forming inverter, grid-following inverter,power
behave in a similar way as synchronous machines without the
hardware-in-the-loop,photovoltaic systems.
need ofPLL for synchronization purposes.
To date, almost all GFMI behavior and incident operational
I. INTRODUCTION benefits have been shown in simulation. More research into
Existing interconnection standards for DERs [1], [2], [3] hardware demonstration is needed. While demonstration of
mandate regulations for the dynamics of grid-tied power GFMI in application environments is ideal, it is difficult to
converters in terms oftransient and steady-state behavior. They tractable test hardware in a wide variety of operation
require DERs to be capable ofassisting the grid under abnormal conditions. Power Hardware-In-Loop (PHIL) is a hybrid
condition such as: low voltage conditions, voltage imbalance, simulation/hardware testbed that can apply simulated grid
and frequency perturbations. DERs perform such assistance conditions to actual hardware. It is a flexible, high fidelity
through the use of ancillary functions such as: volt-var, extension of simulation results that are more tractable to
frequency-watt, and high/low voltage ride-through [4]. implement for a wide variety of operating conditions than a
In scenarios where intentional islanding is desired in order to pure hardware testbed.
energize critical loads, the anti-islanding functionality of the Even with a very broad and in-depth analysis ofthe benefits
DER equipment is disabled with permission of the electric of GFMIs, topics about testing such devices using PHIL are
power system operator. In these cases, without additional often not discussed. While there is abundant literature that
generator sources, GFLIs cannot be used, as they require the explain the different methods of interfacing GFLIs using PHIL
presence of an external voltage signal to synchronize and interfaces [11],[12],[13],[14],[15], the process of interfacing
operate. GFMIs are necessary in such islanded operation since GFMIs using PHIL interfaces presents its own challenges in
they can regulate the voltage magnitude and the frequency terms ofstability and control. These challenges, so far, have not
similarly to synchronous generators. As the increasing been properly addressed in literature.
penetration ofrenewable energy resources is starting to displace This paper presents simulation results of two GFMI
synchronous generators, the amount of inertia in such systems simulation models interfaced with physical GFLIs using a PHIL
is decreased. For these low inertia scenarios, GFMIs models setup. The main idea behind these simulations is to recognize
have been proposed and tested to enhance grid stability [5],[6]. and understand the intrinsic dynamics ofthe simulation models
One of the most important features of GFMIs is their in terms oftransient stability, voltage and frequency regulation,
capability to instantaneously balance load changes without and load sharing capabilities. Results of the PHIL tests will

Sandia National Laboratories is a multimission laboratory managed and operated by National Technology & Engineering Solutions of Sandia, LLC, a wholly owned
subsidiary of Honeywell International Inc., for the U.S. Department of Energy’s National Nuclear Security Administration under contract DE-NA0003525.
provide technical feedback to inverter manufacturers that can This setup provides a flexible platform for the proposed
further discussions and analyses regarding errors between validation tests of GFMI models since all the aforementioned
simulation models and actual GFMIs available in the market. simulation metrics can be implemented in software inside the
Metrics of the validation process focus mainly on the RTS box. Furthermore, any type of abnormal grid condition
following dynamics• i) transient stability during and after that involves significant amounts of energy, such as faults or
interconnection to the grid and under abrupt load changes, ii) large loads, takes place merely at simulation level, thus
grid-forming capabilities under islanded conditions in protecting the DUT against potential catastrophic damages.
distribution systems with low inertia, and iii) proper load
sharing during islanding cases with more than one GFMI is III. DESCRIPTION OF SIIVIULATION MODELS
present in the system.
The simulation models used in this paper are presented and
After a brief PHIL overview, this paper describes, in a
described in this section. Both models have the same generic
detailed manner, the two proposed GFMIs to be tested under
GFMI control scheme shown in Fig. 2, but they differ in the
the PHIL setup. Later, the experimental setup consisting of a
voltage regulation scheme. Notice that droop control is a
low inertia PHIL system is described, followed by the
fundamental part of the control scheme if the inverters are to
experimental results analyses and conclusions.
share loads with other generating sources connected to the
system. The equations ofthe droop characteristics are based on
II. PHIL OVERVIEW the power flow equations from the circuit in Fig. 3:
Advances in parallel computing have allowed the EV EV
proliferation ofstate-of-the-art real-time simulators(RTS)[13], P = —sin g —8 (1)
XL XL
which, in turn, provide the means to implement a systematic
process for testing a variety of devices, including power = E2 — EV cos g E(E —V)
converters. Such testing not only facilitates the rapid Q (2)
prototyping of hardware and its respective control schemes via XL XL
Hardware-in-the-Loop(HIL), but also permits the use ofpower
Where the accuracy of the approximations shown in (1) and
amplifiers while testing commercial devices such as inverters.
(2)are valid only if the X/R ratio ofthe coupling impedance of
The inclusion ofpower amplifiers allows the devices under test
the GFMI is sufficiently large (z10) to a point where only the
(DUTs)to transfer real and reactive power in a process called
inductive reactance can be considered in the analysis [7].
Power Hardware-in-the-Loop (PHIL).
Experiments were performed in the PHIL testbed located at Q
the Distributed Energy Technologies Laboratory (DETL) at
Sandia National Laboratories. Fig. 1 shows a simplified NvVo Vo vdci
diagram ofthis testbed, where a distribution feeder is simulated re 10
—> Droop Voltage
— Cric 444-y-yY\4_Vo
inside the RTS box. The RTS box, in turn, exchanges current Power Control and/or Emod Lf
Calculation current —
and voltage measurements with the DUT to incorporate the
measured signals into the simulated distribution feeder. Notice
Eref
—> control SPWM cT
A
4% 4% A generator H-Bridge
also that the RTS controls the power amplifier according to the o
¡0
-Pg-
feeder dynamics,this allows the proper power flow between the Vset--'
4
1\11\10
power amplifier and the DUT.

Fig. 2. Generic control scheme of a GMFI.

For cases where the X/R ratio does not meet the
DUT
aforementioned assumption, different droop characteristics
Simulator
(inverter need to be implemented([6],[16], and [17]) to counteract for
analog the low X/R. Therefore, if the linearity of the approximations
output
Power in(1) and (2)hold, the real power provided by the inverter can
Amplifier
be controlled by the frequency, whereas the reactive power can
Simulator be controlled by the voltage amplitude, as depicted in the plots
analog of Fig. 3.
inputs
Current & Voltage
The first model (called Model A), uses the CERTS control
measurements and scheme described in [5], whereas the second model (called
scaling
Model B)performs voltage and current control in the dq-frame
Fig. 1. PHIL test setup.
[17].
 introducing a second control loop that reduces the inverter's
frequency in cases where it may become overloaded [5]. For
simplicity, this control loop will be omitted since proper
precautions will be taken (moderate loads) during simulations
to avoid overloading the inverter under steady state conditions.
TABLE I shows the main simulation parameters ofModel A,
where all the droop-control and PI regulators are tuned based
Pset = 0.5 p.u. on per unitized quantities.
fo set = 1 p.u.
0 q
MP TABLE I
SIMULATION PARAMETERS OF MODEL A.
-->
Prated = 1 AU. Qmin Qmax Sbase 50 kVA KP,oi 6.1
Fig. 3. Power flow circuit and droop characteristics diagrams for Vbase 480 Vrms Kiwi 0.001
generic GFMI controls with droop. inp 0.01 Lf 0.8 mH
mq 0.05 Cf 1500 i.tF
A. Description ofControl Scheme ofModel A Cd, 1000011F Vdr 900 V
The control scheme of Model A is shown in Fig. 4. This grid-
forming controller uses frequency-droop and voltage droop
B. Description ofControl Scheme ofModel B.
characteristics to regulate the real and reactive power,
respectively. Before the droop control takes place, the output The control scheme of Model B is shown in Fig. 5. As with
variables io and vo are mapped into the a[3-frame. Such Model A, the frequency-droop scheme dictates the frequency
transformed values of voltage and current are used to calculate of the system according to the load demands. However, in
P and Q in real time, as per the following expressions [18]. Model B, this scheme also provides the system's angular
displacement (A) needed to perform the dq-frame
P =-3[vo io -E vo io ] (3) transformation in the output voltages and currents, which in
2 a a ß ß return are used to calculate the real and reactive power in real
time as per the following expressions [18]:
r
Q = 2 I.— Vocti013 +1;013i0a] (4)
r
P =— ly0diOd + VOqi0q] (6)
Also,a PI controller is used to regulate the voltage reference 2
provided by the voltage droop stage, this improves voltage
Q =-
r (7)
regulation by minimizing circulating currents throughout the
2 I.— vOdiOq + VOqi0d 1

system. The output peak voltage provided to this control stage

is also calculated in the aP-frame as follows: The voltage droop control scheme provides the d-component
reference(Vdref)to the corresponding voltage regulator, whereas
2
VOpeak = A V2
Oa + V0,3 (5) the q-component reference(Vq,ef) is normally set to zero if a
virtual impedance loop is not used [17]. The outputs of the
voltage PI controllers are used as a reference for the current
Power and Peak regulators. As with the grid-following control schemes, the
Voltage Calculation
in ap-Irame DroopControl current control loops are the innermost and fastest loops [19].
, Therefore,the dynamics ofthe current regulators are set at least
Pset tOref

Vaabc a -wOal3
one order of magnitude faster than the droop and voltage
e cos —
J _41 regulators.
abc 1_3 p,Q,
iOabc co _r> Vopeak E„,o
,
Since a dq-reference frame is used, compensation of the
—.> q PI cross-coupling toLf terms must be implemented in order to
abc 64 ---
VOpeak decouple the dynamics of each current control loop from each
vs„ ) other. This compensation reduces the order of each current
s___,
Voltage control
control loop to one single variable, which in return facilitates
Fig. 4. Control scheme of Model A(CERTS droop control).
and simplifies the current compensators design and stability
analysis [18], [19]. TABLE II shows the main simulation
A more advanced feature of the CERTS control scheme parameters of Model B.
includes the use of an overload mitigation regulator, which
helps to prevent the GFMI from DC bus voltage collapse by
 lOdy
P,Q, Hardware
devices
calculation
abc HI L Simulator Box
in dq-frame iOd VOd I $t4/1
TL TL2 Power
'dee
Pse VOabc Amplifier 1 GFLI
- 3-phase

NZ
Vdre

1
Iq PV
Vsei Vod
Simulator
Droop Control Vq
VOa

Fig. 6. PHIL testing setup for three-phase inverters.

Voltage control Current control
Fig. 5. Control scheme of Model B. B. PHIL setupfor single phase experimental tests.

TABLE II This PHIL setup is shown in Fig. 7. Inside the simulation,

SIMULATION PARAMETERS OF MODEL B. two GFMI models are cormected in split phase using a master-
slave configuration [20], where the master device regulates the
Sbase — 50 kVA - V KPvoi - 1.0
voltage and frequency of the system, whereas the slave device
Vbase 480 Vrms Kiva, 10.0 is controlled by the modulation signal provided by the master,
mi, 0.01 Lf 3.0 mH but 180 degrees out of phase, this way the voltage at the split
mq 0.05 Cf 100 µF phase is twice the rated value of each GFMI. Also, simulation
Cdc 10000 µF Vdc 900 V is interfaced with a commercial GFLI rated 3 kVA,240 Vrms.
KPcur, 5.663 KT., 1700 As described later, this setup mimics the configuration of the
single-phase experimental testbed used for testing only
commercial inverters without the use ofPHIL interfaces.
IV. EXPERIMENTAL SETUPS.
Three experimental setups were used for testing the HIL Simulator Box Hardware
simulations models. The first PHIL setup was suited for three- devices
phase systems whereas the second PHIL setup is for single-
phase systems. A third setup, involving only commercially
available devices, was used with the purpose ofcomparison and
validation against the simulations models.
Power
Amplifier
J_
(t
GFLI
= 1-phase
A. PHIL setupfor three phase experimental tests.
This PHIL setup is depicted in Fig. 6. The GFMI model is '\/
connected to two transmission lines (TL1 and TL2)and two PV
loads(Li and L2)implemented in the HIL simulator box, and a Simulator
commercial grid-following inverter is connected to the system
through the PHIL interface. Since no rotating machines are used
in the simulation, this PHIL setup is a low inertia system
(micro-grid) that can test and validate the grid-forming Fig. 7. PHIL setup for single-phase inverters.
capabilities of the simulation models in terms of load sharing
and voltage regulation. The commercial GFLI is a three-phase C. Experimental setupfor commercial single-phase inverters.
device rated at 24 kVA,and 480 V.s. This experimental testbed is located at the Southwest
Loads L1 and L2 are both 24 kW and are connected to the Technology Development Institute (SWTDI) at New Mexico
system through switches SW1 and SW2. Also, in order to State University. This setup is depicted in Fig. 8, where two
investigate the dynamics of the GFMI models under fault commercial GFMIs are cormected on the AC side in a split
scenarios, a low impedance (Zf = 0.1 CI), three-phase line to phase fashion using a master-slave configuration. On the DC
ground fault is applied between the two transmission lines. side, both GFMIs share an AGM battery bank of 48 V. Each
GFMI is rated 5.6 kVA at 120 Vrms. Therefore, the split phase
in the AC side provides the 240 Vrms needed by the GFLI(2.5
kVA), which is cormected on the DC side to a PV solar array of
3.4 kW.
 that the GFLI supplies most ofthe power demanded by Li. Once
L2 connects (after t2), the frequency of the system droops,
GFMI allowing the GFMIs to share both loads with the GFLI. Since
master the DQ GFMI uses two control loops, its response to a change
PV in load is intrinsically slower that the CERTS GFMI, which
GFLI Array
1 phase uses only one control loop. These slower dynamics of the DQ
GFMI are reflected in the momentary but exacerbated negative
glitch in the voltage ofthe system (compared to CERTS GFMI
2
response) when L2 connects, as depicted in the first plot of Fig.
GFMI
slave 10.
1_
, --> GFLI connects t2 ---> L2 connects
Fig. 8. SWTDI experimental testbed at NMSU.
280

E 260
— CERTS GFMI
V. EXPERIMENTAL RESULTS. 240 — DQ GFMI

A. Three-phase case results. 220

35 40 45 50 55 65
The first set of results are related to the dynamics of the time (s)

GFMIs models after a load change, where from Fig. 6, L2 is 60.3

connected at simulation level, forcing the GFMIs models and 11 60.2

,
— CERTS GFMI
the GFLI to supply their respective rated powers while sharing — DQ GFMI
12 60.1
the two loads (48 kW). The chronology of the events for this
test is better explained by analyzing the dynamics of the real 60
40
35 45 50 55 60 65
power provided by the GFMIs models and the interfaced GFLI, time (s)

as depicted in Fig. 9. Before t1, since the GFLI is disconnected, Fig. 10. Voltage and frequency ofthe GFMIs under load changes.
each one ofthe GFMIs supply the required 24 kW ofLi, while
also dictating the voltage and frequency ofthe system. At tl, the Another way to visualize the dynamics ofGFMIs is using the
GFLI connects and follows a period of transition. Since the power-frequency (P-f) plane as shown in Fig. 11, where the
GFLI is closer to L1,it supplies most ofthe required real power. different time stages of the load change test are plotted with
Notice also that the power provided by the GFMI droops different colors following the sequence: green(before GFLI
significantly, but still regulates the voltage and the frequency of connects)-blue(after GFLI connects)-red(after L2 connects).
the system so the GFLI stays online. At t2, the second load L2 The colored arrows indicate the direction of the trajectories as
connects, forcing the GFMI to provide the required real power time elapses. The starting point(t=0) of the trajectories is also
while sharing the loads with the GFLI. marked in both plots

—> GFLI connects t2 ---> L2 connects tn

4
x 4
60.3 - t< t1
—II< t t2
— CERTS GFMI model
,
'
171 60.2 - --12 < t
— GFLI
g-60.1 -
CERTS GERI
60 -
30 35 40 45 50 55 GO 65 70
0 0.5 1 1.5 2 2.5
time (s)
Real power (watts) x104
,...., 3
to
4
`A) '—
60.3* —0< t t1
2 —tl < t < t2
—12< t
— DQ GFMI model ,
74 60.2
o1 — GFLI
To
g60.1 -
0 DQ GFMI
60 - F
30 35 40 45 50 55 60 65 70
time (s) 0 0.5 1 1.5 2.5
Real power(watts) x104
Fig. 9. Real power delivered by the GFMIs.
Fig. 11. Power-frequency plane trajectories of GFMIs.
In terms of voltage and frequency regulation, Fig. 10 shows
the dynamics of both GFMIs. After t1, the frequency dictated From Fig. 11, it can be noticed that during the GFLI
by the GFMIs rises (second plot ofFig. 10)according to the P- connection (blue line), the path described by both inverters is
fdroop characteristic shown in Fig. 3. This is due to the fact the actual P-fdroop characteristic. This is a consequence ofthe
slow transition in power that the system undergoes during this
stage, where both GFMIs control schemes are faster than the
slew rate of real power dictated by the GFLI during its t ---> Fault inception
f
400
connection to the PHIL. During the load change (red line), both
350
trajectories deviate from the P-fdroop characteristic due to the
transient response ofthe control schemes, but in both cases the 300
-CERTS IA
system settles at the required P-foperating point in steady state. -CERTS IB
-CERTS IC
The second set of results are related to the response of the -DQ IA
y 200 -DQ IB
GFMIs under a balanced, remote three phase fault-to-ground -DQ IC
(Zf= 0.1 S2)as depicted in Fig. 6. The chronology ofthe events < 150
is explained by analyzing the dynamics of the real power 100
provided by both the GFMIs and GFLIs as shown in Fig. 12.
50
Before tf(pre-fault conditions), both inverters are supplying L1
and L2. Fault occurs at tf and it lasts 5 cycles. During fault 119.98 120 120.02 120.04 120.06 120.08
occurrence, the GFLI experiences a momentary cessation due time (s)
to the voltage collapse imposed by the fault. Notice that the DQ Fig. 13. Fault current contribution of GFMIs.
GFMI injects more power during the fault than the CERTS
GFMI, this is a result of a more stringent current regulation The frequency plots shown in Fig. 14 illustrate how the
imposed by the DQ control scheme [17]. control scheme mandates the behavior ofthe GFMIs during the
presence of a fault. In the case ofthe CERTS GFMI model, its
f Fault inception control scheme does not regulate the current, therefore it
delivers significantly more real power than the DQ GFMI. As a
-CERTS GFMI model
-GFLI consequence, the frequency of the inverter droops in response
to the increase of injected real power. The opposite dynamics
occur for the DQ GFMI, where the current regulation imposes
GFLI momentary cessation a limit to the fault current contribution. Since the voltage is
collapsed and the current limited, the power injected by the DQ
119.98 1 0 120.02 120.04 120.06 120.08
time (s)
GFMI reduces, and thus the frequency increases as depicted in
x104 Fig. 14.
G• 4
-DQ GFMI model
-GFLI tf ---> Fault inception
2 60.3
`<15
60.25
o
u
m GFLI momentary cessation 60.2
-CERTS pre-fault frequency
-2 _60.15 -CERTS fault frequency
119.98 120 120.02 120.04 120.06 120.08 -DQ pre-fault frequency
60.1 -DQ fault frequency
time (s)
cr
a.)
Fig. 12. Real power delivered by the GFMIs during a 3-phase fault. 60.05

60
The fault current contribution of each GFMI is depicted in
59.95 -
Fig. 13. As previously mentioned, the lower fault current
59 9 - 1
contribution from the DQ GFMI is due to its intrinsic current 119.98 120 120.02 120.04 120.06 120.08
regulation scheme, which also reflects a lower real power time (s)
injection during the fault when compared against the CERTS Fig. 14. Frequency dynamics of GFMIs under a 3-phase fault.
GFMI. The parameter AI, shown in Fig. 13, reveals a current
difference of about 200 A. between the two fault current A. Single-phase case results.
contributions. This current difference must be taken into For this case, the SWTDI experimental setup previously
account when designing the protection scheme for this system. described was used as a validation platform. Due to a limited
Moreover, if overcurrent relays are used, adaptive protection number of sensors, only voltage and frequency were recorded
might be a possible solution to overcome relay misoperation during these experimental tests at SWTDI. Also for safety
[9]. reasons, both commercial GFMI had their frequency-droop
From Fig. 13, it can be observed that the control scheme of characteristics deactivated. To comply with this constraint, the
the GFMI plays an important role in the design ofthe protection frequency-droop control ofthe models was deactivated as well.
scheme for low inertia systems. Where CERTS GFMIs seemed Fig. 15 shows the frequency response of the GFMI when L1
easier to protect using overcurrent relays than their DQ GFMIs and L2 are simultaneously connected to the system as shown in
counterpart. in Fig. 7 and Fig. 8. Both loads demand real and reactive power
(P=4.5 kW and Q=1 kVar, each). Since the droop A low inertia PHIL testbed was introduced to perform the
characteristics are deactivated, the frequency experiences a different testing scenarios. Such a testbed allows for the
transient and then goes back to the commanded value of60 Hz. application ofload changes along with low impedance faults to
The frequency spike is similar in magnitude for both models, the tested GFMI models. In order to preserve the low inertia of
but the CERTS dynamics respond much faster than the the system, a commercially available GFLI inverter was
commercial inverter. interfaced with the PHIL.
Simulation results showed that a GFMI model with DQ
current control shows an intrinsic slower response offrequency
60 and voltage regulation when compared to the response of a
59.9 - -COMMERCIAL GFMI
CERTS GFMI model. It was also shown that under fault
-CERTS GFMI scenarios, the DQ controlled GFMI performed better in terms
59.8 - -DQ GFMI
N of limitation of fault current contribution due to its inherent
59.7 - current control scheme.
a-
g 59.6 - The single phase models were validated against
59.5 -
commercially available inverters. The PHIL results showed
accurate results in steady state, but further research and
59.4 -
experiments are required to validate the corresponding
59.3 differences between experimental and PHIL results during
100 101 102 103 104 105
time (s) transients.
Fig. 15. GFMIs frequency response to a load change. The use of validated, generic models which can accurately
replicate the behavior of real GFMIs under several dynamic
Fig. 16 shows the voltage response of the GFMIs to the load operating conditions allow inverter manufacturers, integrators,
change previously described. CERTS GMFI shows a short and systems operators to obtain technical feedback about the
negative voltage spike, whereas the commercial GFMI shows a dynamic performance of GFMIs under a variety of abnormal
well voltage-regulated transient response. In steady state both operating conditions.
dynamics converge to the same voltage level, which is less than
the voltage before the loads connects. This is due to the V-Q ACKNOWLEDGMENTS
droop control and the presence of reactive power demanded by
Sandia National Laboratories is a multi-mission laboratory
the loads.
managed and operated by National Technology and
Engineering Solutions of Sandia, LLC. a wholly owned
subsidiary of Honeywell International, Inc., for the U.S
Department of Energy's National Nuclear Security
-COMMERCIAL GFMI
-CERTS GFMI Administration under contract DE-NA-0003525.
-DQ GFMI

90
REFERENCES
O
80 -
[1] IEEE 1547 Std. 1547,"IEEE Standard for
70 - Interconnecting Distributed Resources with Electric
60
Power Systems," Institute of Electrical and Electronic
99.5 100 100.5 101 101.5 102 Engineers, New York, 2008.
time (s)
Fig. 16. GFMIs voltage response to a load change.
[2] International Electrotechnical Comission,"IEC 61850-
90-7," International Electrotechnical Comission, 2013.
[3] UL-1741,"Inverters, Converters, Controllers and
IV. CONCLUSIONS Interconnection System Equipment for use with
The dynamic behavior of two simulation models of grid- Distributed Energy Resources," Underwriters
forming inverters are evaluated in terms oftransient and steady- Laboratories„ Northbrook, 2016.
state stability under abrupt load changes and low impedance [4] S. Gonzalez, J. Johnson, M. Reno and T. Zgonena,
faults. The two simulations models have added features to their "Small Commercial Inverter Laboratory Evaluations of
conventional control schemes (droop-control) that allows them UL 1741 SA Grid-Support Function Response Times,"
to improve their performances under abrupt load changes or in 43rd IEEE PVSC,New Orleans, 2016.
overloading conditions.
[5] W.Du, R. H. Lasseter and A. S. Khalsa, "Survivability [14] B. Palmintier, B. Lundstrom, S. Chakraborty, T.
of Autonomous Microgrid during Overload Events," Wlliams, K. Schneider and D. Chassin,"A Power
IEEE Transactions on Smart Grids, 2018. Hardware-in-the-Loop Platform With Remote
[6] H. Bevrani, B. Francois and T. Ise, Microgrid Dynamics Distribution Circuit Cosimulation," IEEE Transactions
and Control, Hoboken, New Jersey: Wiley & Sons, Inc, on Industrial Electronics, vol. 62, no. 4, 2015.
2017. [15] M.Panwar, B. Lundstrom, J. Langston, S.
Suryanarayanan and S. Chakraborty,"An Overview of
[7] P. Kundur,Power System Stability and Control, 2nd ed.,
New Delhi: Tata McGraw-Hill, 1993. Real Time Hardware-in-the-Loop Capabilities in Digital
Simulation for Electric Microgrids," in North American
[8] B. Johnson, Y. Lin, J. Eto, R. Lasseter, A. Ellis and D. Power Symposium (NAPS), Manhattan, KS,USA,2013.
Kirshen,"Low Inertia Grids," 29-30 April 2019.
[Online]. Available: https://lowinertiagrids.ece.uw.edu. [16] A. Bidram, V. Nasirian, A. Davoudi and F. L. Lewis,
Cooperative Synchronization in Distributed Microgrid
[9] H. J. Altuve Ferrer and E. O. Schweitzer III, Modern Control, Springer International Publishing, 2017.
Solution for Protection, Control and Monitoring of
Electric Power Systems, Pullman, WA: Schweitzer [17] A. Vinayagam, K. Swarna, S. Khoo, A. Oo and A.
Engineering Laboratories, 2010. Stojcevski,"PV Based Microgrid with Grid-Support
Grid-Forming Inverter Control-(Simulation and
[10] E. 0. Schweitzer III, B. Kasztenny, A. Guzman, V.
Analysis)," Smart Grid and Renewable Energy, vol. 08,
Mynam and H. J. Altve Ferrer, Locating Faults and
no. 01, pp. 1-30, 2017.
Protecting Lines at the Speed of Light: Time-Domain
Principles Applied, Pullman, WA: Schweitzer [18] A. Yazdani and R. Iravani, Voltage-Sourced Converters
Engineering Laboratories, 2018. in Power Systems. Modeling, Control, and
Applications., Hoboken, New Yersey: John Wiley &
[11] A. Summers, J. Hernandez-Alvidrez, R. Darbali-
Sons, 2010.
Zamora, M. J. Reno, J. Johnson and N. S. Gurule,
"Comparison ofIdeal Transformer Method and [19] R. Teodorescu, M.Liserre and P. Rodriguez, Grid
Damping Impedance Method for PV Power-Hardware- Converters for Photovoltaic and Wind Power Systems,
In-The-Loop Experiments.," in 46th IEEE Photovoltaic 1st ed., John Wiley & Sons, 2011.
Specialists Conference, Chicago, IL, 2019. [20] SMA Solar Technology AG,"Sunny Island
[12] J. Hernandez-Alvidrez, A. Summers,N. Pragallapati, M. Downloads," 2016.[Online]. Available:
J. Reno, S. Ranade ,J. Johnson, S. Brahma and J. https://files.sma.de. [Accessed May 2019].
Quiroz, "PV-Inverter Dynamic Model Validation and [21] M. B. Delghavi and A. Yazdani,"A Control Strategy for
Comparison Under Fault Scenarios Using a Power Islanded Operation ofa Distributed Resource Unit," in
Hardware-in-the-Loop Testbed," in IEEE 7th World IEEE Power and Energy Society General Meeting,
Conference on Photovoltaic Energy Conversion 2009.
(WCPEC), Waikolola Village, HI, USA,2018. [22] S. S. Venkata, M. J. Reno, W. Bower, S. Manson, J.
[13] S. S. Noureen, N. Shamim, V. Roy and S. B. Bayne, Reilly and G. W.Sey Jr, "Microgrid Protection:
"Real-Time Digital Simulators: A Comprehensive Study Advancing the State ofthe Art," Sandia National
on System Overview, Application, and Importance," Laboratories, SAND2019-3167, Albuquerque, 2019.
International Journal ofResearch and Engineering, vol.
IV, no. 11, pp. 266-277, November 2017.