This article has been accepted for publication in IEEE Transactions on Sustainable Energy.

This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/TSTE.2024.3520989

A Novel Design for Switchable Grid-Following and

Grid-Forming Control
Huazhao Ding, Student Member, IEEE, Rabi Kar, Student Member, IEEE, Zhixin Miao, Senior Member, IEEE,
Lingling Fan, Fellow, IEEE

Abstract—This paper presents the design of a novel grid- MIGRATE project for GFM evaluation in bulk power system
forming (GFM) control structure adapted from a typical grid- operation [12]. The multi-loop control has a similar cascaded
following (GFL) control structure with minimal edits, thereby structure as that of the GFM in [1]. The aforementioned GFM
enabling a switchable control structure for voltage sourced
converters (VSCs) to operate in either GFL or GFM mode control structures are different from a GFL in synchronization
by simply switching a flag manually. The VSC is shown to be methods, where the synchronization angle is generated by
able to operate in the GFL control mode synchronizing to the a power-based control, instead of the voltage-based method
main grid through a phase-locked-loop (PLL) and operate as (e.g., PLL). This difference has been recognized by the re-
a GFM controller with power-based synchronization for both search community [13], [14], the National Renewable Energy
grid-connected and islanded conditions. To guarantee smooth
operation, the control schemes and the mode switching logic have Laboratory’s GFM roadmap report [15], as well as the grid
been carefully designed and examined via a series of experiments. industry [8]. In terms of power-based synchronization, there
The experiment results show that the switchable control structure are multiple variations, including power-frequency droop [9]
can fulfill the desired control and operation functions and enable and virtual synchronous generator with inertia emulated [16].
smooth transition between control modes. It has to be noted that while both the GFL and the multi-loop
Index Terms—Voltage sourced converters, grid-following, grid- GFM employ inner current control, their outer controls are
forming, synchronization, islanding, grid integration. very different. The outer control of a GFL generates the current
orders for the inverter’s output currents, while the outer control
I. I NTRODUCTION of a multi-loop GFM [1], [11], [12] generates the orders for the
shunt capacitor filter’s currents, which have to be compensated
T HREE-phase voltage-sourced converters (VSCs) are in-
creasingly useful in modern power systems. An important
application for VSCs is inverter-based resources (IBRs), e.g.,
by feed-forwarding the external current to further produce the
inverter current orders. It can be seen that such GFM requires
solar photovoltaic (PV) and wind turbines [1], [2]. an additional current sensor. Additionally, the GFM’s outer
Control design for VSC is essential for IBRs to achieve de- control regulates the PCC bus voltage by the d-axis control
sired functionalities for microgrid operation or grid-connected while the q-axis control enforces the PCC bus voltage space
operation. In the past decade, significant progress has been vector to be aligned with the synchronizing frame. This is to
made in VSC control design to suite the needs of various be compared with the voltage control in the GFL, which is
operational requirements. The state-of-the-art VSC control realized by the q-axis outer control.
includes both the grid-following control (GFL) based on PLL
and the grid-forming control (GFM) based on a controlled A. Motivation
frequency [1]. In real-world bulk power grid operation, it The objective of this research is to design a converter
has been found that PLL-synchronized GFL introduces many controller and implement it in a chip. To make sure that the
operational issues, including large angle deviation upon grid customers receive maximum benefits, the controller chip has
disturbances [3], [4], and weak grid stability issues [5], [6]. been designed so that it can be used as either GFL (voltage-
The latter issues have also been identified for VSC-HVDC based synchronization) or GFM (power-based synchroniza-
operating in weak AC grids [7] in the 2000s. tion). To do so, we make sure that the hardware parts of the
The grid industry has put efforts in pushing for more GFM GFL and GFM are the same and the software controllers share
into the bulk power grids to have IBRs provide necessary common parts as much as possible. In this case, both GFL
frequency and voltage support [8]. In turn, quite a few new and GFM use the same sets of sensors, and share the current
designs of GFM have been proposed, including single-loop controllers.
control to directly control the converter’s output voltage and
its angle to emulate a synchronous generator [9], [10] and
B. Research goals
multi-loop control consisting of inner current control and outer
control to regulate the point of common coupling (PCC) bus This research has two goals: 1) a switchable VSC with both
voltage [11], [12]. The latter structure has been adopted by the GFL and GFM functions; and 2) the GFM design with the
capability of seamless operational condition transition between
The research is supported in part by Department of Energy DE-EE- grid-connected and standalone.
0011474. H. Ding, R. Kar, Z. Miao, and L. Fan are with the Department
of Electrical Engineering, University of South Florida, Tampa, FL, 33620. To date, the majority of the IBRs are either in GFL control
E-mail: [email protected]. mode or GFM control mode. In this research, we aim to

Authorized licensed use limited to: INDIAN INSTITUTE OF TECHNOLOGY GUWAHATI. Downloaded on January 06,2025 at 07:20:08 UTC from IEEE Xplore. Restrictions apply.
© 2024 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.See https://www.ieee.org/publications/rights/index.html for more information.
 This article has been accepted for publication in IEEE Transactions on Sustainable Energy. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/TSTE.2024.3520989

𝑣
𝑎𝑏𝑐

𝑑𝑞
𝑉
𝑃𝐼 +
+

𝜔
1
𝑠
𝜃
GFL&GFM
Synchronization
𝑃

𝑃 ,
−
+ 𝑚 +

1
+
𝜔
1
𝑠
𝜃

Fina
Phase − locked loop (PLL) Power Synchronization (PS)

if flag = 0
0
𝑖 , , if flag = 1 if true, flag = 1
𝑃𝐼 GFL control?
𝑃 , if flag = 1 if false, flag = 0
+
−

𝑃 GFL power control

𝑉
if flag = 1 𝑖 , + +
0 − 𝑃𝐼 +

𝑖 if flag = 0
−
𝑑𝑞
, ,
𝑃𝐼
if flag = 0 𝑖 𝜔 𝐿 𝑚
0 +
−

𝑖 𝜔 𝐿
𝑉 , 𝑉 GFM angle control
−
𝑃𝐼 +
+ 𝑎𝑏𝑐
+ +

if flag = 0
0 𝑖 , 𝐿𝑃𝐹
𝑖 , if flag = 1 𝑉 𝜃
𝑃𝐼
𝑉 , , if flag = 1 𝜃
−
+
if flag = 1

𝑉 GFL voltage control 𝜃

if flag = 0
if flag = 1
0
𝑖 if flag = 0 𝑣 𝜃 𝑃 𝜃
,
𝑃𝐿𝐿 𝑃𝑆
𝑃𝐼 Δ𝜃 Δ𝜃
𝑉 , , if flag = 0 +
+

−
+ 𝑣 𝜃 − 𝑣 𝜃
−

𝑃𝐿𝐿 𝑃𝐿𝐿
𝑉
GFM voltage control Angle for measurement

Fig. 1: The GFL and GFM switchable control strategy.

have both GFL control and GFM control in a single VSC frequency VSC presented in [1] (chapter 9) with additional
controller chip. A selection of GFL or GFM can be realized active feedback compensation and droop method. Similarly,
manually by simply pressing a button. The GFL and GFM the hybrid converter control proposed in [17] can work for
controls are based on the same two sensors: the converter’s both operating conditions.
exporting current and the PCC bus voltage. A selection of In this research, the designed VSC control will be analyzed
GFL or GFM can be realized by simply pressing a button. and tested using hardware experiments. Compared to the time-
This is a cost effective approach to have a VSC control in two domain simulation based testing methods employed in [17],
modes, providing more flexibility. Additionally, the control [21], conducting hardware experiments is a noteworthy effort
design philosophy is to minimize control structure changes to achieve high feasibility.
for the two types of controls. To this end, the GFL and GFM
C. Contributions
will not only use the same set of measurements but also
share the same the inner current control. While design, prototyping, and hardware experiments of
the GFLs and the multi-loop GFM have been carried out in
In a recent research article [17], Lima and Watanabe de- numerous research projects, including our own research on
signed a hybrid converter control emulating a parallel GFL IBR weak grid stability demonstration and control design,
and GFM. This design is also based on the assumption that e.g., [22]–[25], design, prototyping, and hardware experiments
the same sensors and measurements are used for GFL and of the proposed GFM with minimal edits from GFL have
GFM. On the other hand, none of the outer controls share the not been conducted. And while its performance for grid-
same structure. connected condition has been checked in computer simulation
This second goal of this research is to have the GFM in the senior authors’ new book [26] (Chapter 6), islanded
proposed in this paper operate robustly in both operation operation has not been checked. The current research will give
conditions. In the previous decade, distinct controls are a detailed examination of this control structure, configure each
switched upon operation condition change, with GFL been parameter, and provide prototyping results and experiment
used for grid-connected operation while GFM been used for results for such a GFM under various operation conditions.
the standalone operation [18]–[20]. This type of switching The contribution of this research is threefold.
strategies requires fast detection of operation conditions and 1) First, a novel GFM control structure is designed, ana-
communication. Later on, research has been developed to lyzed, and tested. This GFM control structure evolves
have one controller for both operating conditions. In [21], from a standard GFL control structure by using the
Delgahavi and Yazdani developed a unified control strategy same set of measurement sensors and utilizing the same
for this purpose. The control is based on the controllable inner current control structure and q-axis outer control

Authorized licensed use limited to: INDIAN INSTITUTE OF TECHNOLOGY GUWAHATI. Downloaded on January 06,2025 at 07:20:08 UTC from IEEE Xplore. Restrictions apply.
© 2024 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.See https://www.ieee.org/publications/rights/index.html for more information.
 This article has been accepted for publication in IEEE Transactions on Sustainable Energy. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/TSTE.2024.3520989

structure. Only the synchronizing unit and the d-axis It is worthwhile to note that the GFL and the GFM use
outer control are re-designed. Compared to the con- the same sensors. Compared to the conventional GFM control
ventional design [1], [11], [12] which is based on three proposed in the literature, e.g., [1], [11], [12], the current one
sensors, the new design is based on two sensors and has a few distinct features leading to easy implementation for
is apparently more economical. While the outer control switchable GFL-GFM control.
in the conventional design regulates the synchronizing • The GFM in [1], [11], [12] has its outer control’s d-axis
angle through the q-axis control or the reactive current, for Vd control and q-axis for Vq control. This control
the proposed design regulates the synchronizing angle structure is completely different from a GFL’s where q-
through the d-axis control or real current while regulates axis is for voltage/var control.
voltage through the q-axis control or reactive current. • The GFM in [1], [11], [12] has cross coupling terms
2) Second, the novel GFM control allows the manufactur- in its outer control and its outputs are the capacitor
ing of a the switchable GFL-GFM control (including filter’s current orders. On the other hand, for the inner
not only controllers for two modes but also switching current control, the converter current measurement is to
logics). The switchable GFL-GFM has been designed, be used. In order to generate the converter current orders,
prototyped, and shown to function as desired in a hard- the conventional GFM requires an additional current
ware testbed. This proves the feasibility of the design, sensor to measure the external current and feedfoward the
implicating a high practical value. current measurements to generate the converter current
3) Third, this research not only shows results from a set orders. This makes the GFM control quite complicated.
of experiments but also reveals the critical difference of
In short, the GFM control implemented in this paper is much
GFL and GFM that has not been captured in the current
simpler in structure and maximumly reuses the blocks of GFL.
literature. This critical difference can be summarized as
How to select GFL control parameters has been well studied
the difference of the relative position between the PCC
in the past, e.g., [1]. Most recently, the per unit-based control
bus voltage vector and the synchronizing frame. It plays
design analysis and parameter selection have been documented
a key role in creating different dynamic performance of
in detail with each parameter given in [27]. Also, a typical
GFL and GFM, as shown in the experiment results.
number for the power-frequency droop parameter m is also
well known, e.g., 3%, 5%. The parameters of the control are
D. Structure of the paper shown in Table I. The only challenging task is to set the
The sections following outline the switchable VSC control parameters for the GFM’s d-axis outer control. This will be
design, analysis, hardware setup, and experiment results. The presented in the following.
switchable control design is presented in Section II. The
TABLE I: Parameters for the circuit and the VSC control.
hardware testbed setup is then presented in Section III. Section
IV presents four sets of experiment results, including basic Description Symbols Values (SI/pu)
control function tests, control mode switching tests, operating Power base Sb 20 VA
Voltage base (LL, rms) Vb 17.32 V
condition switching tests, and grid voltage dip tests, along with System frequency f 60 Hz
detailed insightful analysis. Finally, Section V summarizes this Grid voltage (LL, rms) vg 17.32 V
paper and highlights the contributions. DC voltage VDC 40 V
Switching frequency fSW 4 kHz
Choke filter Rf 0.0094 pu
II. S WITCHABLE VSC C ONTROL Choke filter Lf 0.0754 pu
Shunt C Cf 0.2658 pu
Fig. 1 shows the switchable VSC control. The VSC can Transmission line Rt 0.0029 pu
operate in the GFL mode when the flag is set to 1 and operate Transmission line Lt 0.2155 pu
Load resistance Rl 1.33 pu
in the GFM mode when the flag is set to 0. The figure shows GFM: P-f droop gain m 3% pu
that there are two different outer controls that generate the GFL: PLL PI 40+400/s
reference signals for the inner loop current control. The GFL Inner current control PI 0.3+20/s
GFL power control PI 0.6+60/s
control regulates real power in the d-axis and regulates AC GFL voltage control PI 0.4+40/s
voltage in the q-axis, while the GFM control enforces the GFM angle control PI 2+14/s
alignment of the PCC bus voltage with the synchronizing GFM voltage control PI 3+40/s
frame in the d-axis and regulates the AC voltage in the q-
axis.
Aside from the difference in d-axis outer control, the main
difference between GFL and GFM is the synchronization A. Philosophy of the d-axis angle control for GFM
method. In GFL, voltage-based synchronization or PLL is used The design of the d-axis outer control for GFM is based
to generate a synchronizing angle θPLL while in grid-forming on the grid-operating condition with the grid been treated as a
control, power-based synchronization is used to generate a constant voltage source behind a pure reactance jX, while the
synchronization angle θPS . In this paper, power-frequency VSC is treated as a current source synchronized to the grid
droop control has been adopted as the synchronizing control. through a synchronizing angle δ. This treatment is reasonable
It is also easy to implement a virtual synchronous generator since the inner current control is very fast compared to the
control with inertia emulated. outer control. This circuit topology is illustrated in Fig. 2.

Authorized licensed use limited to: INDIAN INSTITUTE OF TECHNOLOGY GUWAHATI. Downloaded on January 06,2025 at 07:20:08 UTC from IEEE Xplore. Restrictions apply.
© 2024 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.See https://www.ieee.org/publications/rights/index.html for more information.
 This article has been accepted for publication in IEEE Transactions on Sustainable Energy. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/TSTE.2024.3520989

shown in Fig. 3.
The closed-loop transfer function from δ to δPCC becomes:
X Kp + Ksi


∆δPCC ∆θPCC Ts + 1
= = K

= , (6)
∆δ ∆θPS 1 + X Kp + s i T1 s + 1
where T = Kp /Ki and T1 = T + 1/(XKi ), Kp and Ki are
the d-axis outer control parameters.
Fig. 2: Illustrative circuit of an IBR integrated to a grid. The loop gain from the synchronizing angle ∆δ to real
power P and back to itself can be found as follows:
T s + 1 cos δPCC mω0
The PCC bus voltage and the current viewed from the grid Loop Gain = . (7)
dq-frame is as follows. T1 s + 1 X s
This feedback system is always stable under the assumption
V = V ejδPCC = jX(id + jiq )ejδ + Vg , (1) that the PCC bus voltage is well controlled. The closed-loop
where δ is the relative angle of the synchronizing frame system from the power order to the PCC bus voltage angle is
against the grid. Hence δ is either ∆θPLL or ∆θPS . The grid further examined. This system is presented as follows:
impedance is assumed purely reactive and the reactance is X. ∆δPCC T1 s + 1
If the grid dq-frame is aligned with the PCC bus voltage space =   . (8)
∆Pref T1 2 1 cos δPCC cos δPCC
vector at the initial steady state, then δPCC and δ are 0 initially. mω0 s + mω0 + X T s+ X
The small-signal relationship of the above equation is: It can be seen that the closed-loop system is a second-order
0 system. If T (T = Kp /Ki ) is too small, the system may
 
∆V + jV ∆δPCC = jX (∆id + j∆iq ) + jI ∆δ , (2)
subject to oscillations.
0 Fig. 4a shows the Bode diagrams of the closed-loop system
where I = id + jiq is the initial current vector viewed from
for two set of PI controllers (2 + 14/s vs. 0.2 + 14/s) under
the synchronizing frame. The imaginary part of Eq. (2) is:
a varying grid impedance. Fig. 4b shows the time-domain
V ∆δPCC = X∆id − Xiq ∆δ ≈ X∆id , if iq is small. (3) response of ∆δPCC subject to a step change in the power
order. The grid impedance varies: 0.25 pu, 0.50 pu, 0.75 pu,
Eq. (3) shows that ∆id is proportional to the PCC bus voltage
0.95 pu. It can be seen that 0.2 + 14/s may lead to a peak at
angle ∆δPCC . Therefore, id may be adjusted to track ∆δPCC .
about 2 Hz when X is below 0.95 Hz in the Bode diagrams,
Note that Vq is the PCC voltage’s projection to the q-axis
and overshoot due to oscillations at about 2 Hz in the time-
of the synchronizing dq-frame. Hence:
domain responses. Therefore, to achieve robust performance,
Vq = V sin(δPCC − δ) ≈ δPCC − δ, if V ≈ 1. (4) 2 + 14/s is chosen as the PI controller for the d-axis outer
loop of the GFM.
With −Vq or (δ−δPCC ) passing into the d-axis PI controller
to generate id order, id may be adjusted to further influence
B. Analysis of the GFM control for standalone operation
δPCC according to Eq. (3). Also, δPCC influences the real
power exporting. Eq. (5) presents the power and angle rela- While the GFM design is based on the grid-connected
tionship assuming the PCC voltage is kept constant at 1 pu. operating conditions, this control is also analyzed for the
standalone operation when the GFM is serving a resistive
V Vg
P =sin δPCC , load. Fig. 5 illustrates the circuit of a GFM serving a resistive
X load where the GFM is again treated as a controllable current
cos δPCC
=⇒ ∆P = ∆δPCC . (5) source. It can be seen that the dq-frame PCC bus voltage is
X related to the dq-frame IBR current as follows:
∆vd = Rl ∆id , ∆vq = Rl ∆iq . (9)
The effect of the inner current control is modeled as a low-
pass filter 1/(τ s + 1). It can be seen that ∆vq is the input
to the d-axis outer control influencing ∆id , while the ∆vd is
the input to the q-axis outer control influencing ∆iq . These
relationships form a closed-loop system shown in Fig. 6.
In this feedback system, the q-axis voltage control needs to
be tuned to coordinate with the d-axis control (2 + 14/s) for
Fig. 3: The block diagram illustrates the relationship between the d-axis angle stability. Fig. 7 shows the open-loop system Bode diagrams
control and the power-frequency droop when the GFM is in the grid-connected
condition. for two sets of the voltage PI control: 0.4 + 40/s vs. 3 + 40/s.
The integral gain is 40, while the proportional gain of 0.4 and
Combining (3), (4), and (5), the d-axis control and the 3 are compared. The resistive load Rl is at 20 Ω or 1.33 pu and
power-frequency droop control, the block diagram relating the the low-pass filter’s time constant is assumed to be 25 ms. The
synchronizing angle, the PCC bus angle, and real power is blue dotted line (when kp = 0.4) in Fig. 7 shows that at 10 Hz,

Authorized licensed use limited to: INDIAN INSTITUTE OF TECHNOLOGY GUWAHATI. Downloaded on January 06,2025 at 07:20:08 UTC from IEEE Xplore. Restrictions apply.
© 2024 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.See https://www.ieee.org/publications/rights/index.html for more information.
 This article has been accepted for publication in IEEE Transactions on Sustainable Energy. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/TSTE.2024.3520989

Bode Diagram Bode Diagram

80
10 kp = 0.4
60
kp = 3

Magnitude (dB)
0 40
Magnitude (dB)

-10 20

0
-20 X increases
-20
-30
-40
90
-90

Phase (deg)
45
-135
Phase (deg)

0
-180
-45

-225
-90
10-1 100 101 102
10-2 10-1 100 101 102

Final version
Frequency (Hz) Frequency (Hz)

Fig. 7: The Bode diagram of the open-loop system in Fig. 6, illustrating the
(a)
potential oscillatory stability issue if the voltage control’s proportional gain
is too small.
Step Response
3.5

3
𝑣 𝑖 𝑣 𝑖 𝑣
𝑉
2.5
Grid
2 𝑅 𝐿 𝑅 𝐿
Amplitude

1.5
𝐶 𝑅
PWM
1

𝑚
0.5

0 X increases
(a)
-0.5
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 𝑃 , 𝑖 ,
+ 𝑃𝐼
Time (seconds) −

𝑖 , , 𝑉
(b) 𝑃 𝑖 ,
+ 𝑃𝐼 +
+
− −
𝑉 𝑖 𝑑𝑞
, ,
Fig. 4: (a) Bode diagram of ∆δPCC /∆Pref . (b) Step responses of −
+ 𝑃𝐼 𝑖
𝜔 𝐿 𝑚
𝑖 ,
∆δPCC /∆Pref . Blue lines: 2 + 14/s. Red lines: 0.2 + 14/s. GFL-outer loop 𝑉 𝑖
𝜔 𝐿
𝑉 , 𝑖 ,
𝑖 , +
−
𝑃𝐼 +
+ 𝑎𝑏𝑐
0 +
− 𝑃𝐼 +

𝑖 , ,
𝑉 𝐿𝑃𝐹
𝑉 𝜃
𝑉 , 𝑖 ,
− 𝑃𝐼 𝑣 𝑃𝐿𝐿 𝜃
+

𝑉 𝑃 𝑃𝑆 𝜃
GFM-outer loop

𝑎𝑏𝑐
𝑣 𝑉 1 𝜃 𝑃 1 𝜃
𝑃𝐼 + − 𝑚 +
+
𝜔
+
𝑠 +
𝑠
𝑑𝑞
𝜔 1
𝑃 ,

Phase − locked loop (PLL) Power Synchronization (PS)

Fig. 5: The circuit of a GFM serving a resistive load Rl .

(b)

Fig. 8: (a) The testbed’s circuit diagram. (b) The hardware setup of the testbed
at the USF SPS lab.

III. T HE T EST S YSTEM AND I TS H ARDWARE S ETUP

Fig. 6: The block diagram illustrates the relationship between the d-axis As illustrated in Fig. 8a, the test system is a three-phase
control and q-axis control when the GFM is in the standalone operation VSC that can be operated in both islanding and grid-connected
serving a resistive load. conditions. The VSC is connected to the grid via a choke filter,
a shunt capacitor filter, and a transmission line. The choke
filter consists of Rf , Lf while the shunt capacitor is notated as
the phase goes below 180 degrees, while the magnitude has a Cf . The transmission line is represented by an RL impedance
gain greater than 0 dB, implicating instability. Increasing the consisting of resistance Rt and inductance, Lt . The flow of
proportional gain helps mitigate the stability issue. Therefore, power is indicated by the red arrow. The grid is connected to
it can be seen that the voltage control’s proportional gain the VSC through a switch, allowing the system to be operated
should be increased to ensure stability. in both grid-connected and islanding conditions. The PCC bus

Authorized licensed use limited to: INDIAN INSTITUTE OF TECHNOLOGY GUWAHATI. Downloaded on January 06,2025 at 07:20:08 UTC from IEEE Xplore. Restrictions apply.
© 2024 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.See https://www.ieee.org/publications/rights/index.html for more information.
 This article has been accepted for publication in IEEE Transactions on Sustainable Energy. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/TSTE.2024.3520989

is located between the choke filter and the transmission line, the feasibility of the mode switching logic and the
where it is also connected to a resistive load Rl . In islanding performance of the system.
operation, the VSC serves the load only. 3) Category 3: Grid fault tests to examine the performance
The circuit presented in Fig. 8a has been built as a hardware of the IBR and the benefits of mode switching.
testbed, shown in Fig. 8b. The three-phase VSC is a Silicon
Carbide MOSFET module which can handle up to 800 V DC, TABLE II: Lists of hardware experiments and simulation tests.
24 A continuous current, with a maximum switching frequency Cat Cases Description Comments
of 200 kHz. The choke filter and the transmission line are 1
A1 GFL power and voltage reference step hardware
represented by inductors. The shunt filter capacitor is a 400-V changes
A2 GFM power and voltage reference step hardware
47-µF capacitor. The details of the parameters are presented changes
in Table I. 2
B1 Control mode switching & system con- hardware
The grid is represented by the Chroma regenerative grid dition changes
B2 Sudden islanding & control mode hardware &
simulator. This device is a programmable bidirectional power switching simulation
supply that can operate at up to 45 kVA power and can provide C1 Three-phase grid voltage magnitude dip hardware &
a constant voltage output of 400 V per-phase with an operating 3 simulation
C2 Weak grid event simulation
frequency between 30-100 Hz. The device can be used to C3 Grid voltage dip, line tripping following simulation
generate both single-phase and three-phase voltages. To switch by mode switching
between grid-connected and islanding conditions, we use a
miniature circuit breaker (MCB), shown in both Fig. 8a and It is noteworthy to point out that majority of the research
Fig. 8b. papers on control and operation strategy tests rely on computer
In order to design the VSC control, we need to measure simulation only, e.g., [17], [21], [28], [29], while this research
the voltage of the PCC bus and the current exporting from the has conducted control hardware prototyping and testing in
VSC. The OPAL-RT voltage and current sensor, OP-8662 is a hardware experiment testbed. Additionally, this hardware
used to collect measurements. This device has eight voltage testbed has been modeled for computer simulation. The hard-
sensors and eight current sensors. The voltage sensor can ware and the electromagnetic transient (EMT) simulation
measure up to 600 V and the current sensor can measure up testbeds have been benchmarked.
to 15 A at a sampling frequency of 100 kHz.
The most important part of this setup is the IO modules
A. Category 1: Basic control function tests
which connects the sensor to the controller. We use National
Instrument (NI) 9205 as an analog input module, which is The control strategy built in LabVIEW is capable of op-
connected to the NI cRIO 9049 as the controller hardware. erating in both GFL and GFM controls by the toggle of a
The voltage and current sensor modules have the ability to switch. In this section, we validate the successful operation
reduce the measured quantities to an acceptable range for the of the individual control mode by giving a step change in the
analog input module. control commands.
The cRIO 9049 is a real-time embedded industrial controller 1) A1: GFL: When the system is operating in the GFL
which consists of a real-time controller, reconfigurable IO mode, a PLL is used to synchronize the inverter with the grid.
Modules (RIO), FPGA module and an Ethernet expansion In addition, real power and PCC bus voltage are regulated
chassis. It is capable of real-time signal communication and through the inverter’s outer control. To run the testbed in the
control. This device runs on a 32-bit Linux real-time OS and grid-following control mode, we set the flag as 1 or GFL
has an FPGA module that can operate at a sampling rate of control mode. The inner current control’s commands idq,ref
40 MHz. This device serves as a platform for communication are generated from the real-power and AC voltage regulators.
with the remote PC running LabVIEW software, allowing a Two consecutive experiments were conducted to test the
developer to visually examine the data through NI’s input basic control functions of GFL. Initially, the inverter is sending
and output modules and design the controller in LabVIEW. out 0.5 pu real power and the PCC voltage is 1 pu.
The control signal generated in LabVIEW is fed back to • At t = 4 s, Pref,GFL is subject to a step change to increase
the inverter digital port through a digital out module. For its value from 0.5 pu to 1.0 pu.
monitoring purposes and to save the data from the testbed, • At t = 10.1 s, Vd,ref is subject to a step change to
we use an analog out module and feed the signal to 4 decrease its order from 1 pu to 0.95 pu.
oscilloscopes. A series of measurements are taken, including the real
power and reactive power from the converter to the grid
IV. E XPERIMENT R ESULTS measured at the PCC bus, the frequency and angle output of
Overall, three categories of tests have been conducted with the PLL, the dq voltage and current components in the PLL
the details of cases documented in Table II. frame. Note that the PLL’s angle is a relative angle against a
1) Category 1: Basic control function tests to confirm the synchronous rotating frame. This frame aligns with the PLL
VSC control can fulfill its simple control functions, e.g., frame before 4 s.
power/voltage order following. Fig. 9a shows experiment results of the testbed. It can be
2) Category 2: GFL/GFM mode switching tests during seen that the real power measurement follows the real power
steady-state operating or sudden islanding to examine command to increase from 0.5 pu to 1.0 pu after 4 s; the AC

Authorized licensed use limited to: INDIAN INSTITUTE OF TECHNOLOGY GUWAHATI. Downloaded on January 06,2025 at 07:20:08 UTC from IEEE Xplore. Restrictions apply.
© 2024 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.See https://www.ieee.org/publications/rights/index.html for more information.
 This article has been accepted for publication in IEEE Transactions on Sustainable Energy. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/TSTE.2024.3520989

1 as follows:
P
0.5
Q
0
0 2 4 6 8 10 12 14 P = Vd id , Q = −Vd iq . (10)
-0.4
It can be seen that id is indeed 0.5 pu before 4 s and 1.0
62
0 2 4 6 8 10 12 14 pu after 4 s. After 10.1 s, since the PCC voltage reduces, to
f (Hz)

60
maintain the same level of real power, id should increase. This
0 2 4 6 8 10 12 14
is also reflected in the experiment results of id measurement.
(deg)

10 Research conducted in [30] has shown that the PCC bus’

5 PLL
0 voltage phase angle is proportional to id . Since PLL’s angle
0 2 4 6 8 10 12 14 tracks the PCC bus’ voltage phase angle, hence, the PLL angle
1.2
increases if id increases. It can be seen that the trajectory of
Vd

1
0.8
0 2 4 6 8 10 12 14
the PLL angle is similar as that of id .
0.2
Additionally, based on the change in the angle (about 7
Vq

0
-0.2 degrees due to 0.5 pu power increase), we may estimate the
0 2 4 6 8 10 12 14
total transmission line reactance X.
1
π
1 7 × 180
id

0.5
∆P = ∆θ, =⇒ X = = 0.2443. (11)
0.8
0 2 4 6 8 10 12 14 X 0.5
0.4 We can see that this number is slightly larger than the name
iq

0
0 2 4 6 8 10 12 14 plate number from the inductor. Similarly, the change in the
Time (s)
voltage and the change iq has the following relationship:
(a)
∆V ≈ −X∆iq . (12)
When the voltage reduces by 0.05 pu, iq increase by 0.2 pu.
1
P

0.5
This implicates that X is approximately 0.25 pu.
0 2 4 6 8 10 12 14 Overall, the basic control functions can be fulfilled and the
0
IBR behaves as expected.
Q

-0.4
2) A2: GFM: Fig. 9b shows the test results of the GFM
0 2 4 6 8 10 12 14
62 control. GFM adopts power-based synchronization. Eq. (13)
f (Hz)

60 shows the droop relationship used to generate frequency,

0 2 4 6 8 10 12 14 where m is the droop gain notating the change of frequency
(deg)

10
5 PS
in pu vs. the change of power in pu.
0

1.2
0 2 4 6 8 10 12 14 ω − ω0 = m ω0 (Pref,GFM − P ) (13)
Vd

1
0.8
The GFM control can be used for both grid-connected and
0.2
0 2 4 6 8 10 12 14 islanding modes.Under grid-connected mode, the frequency
will be synchronized with the grid ω = ω0 , hence the power
Vq

0
-0.2
0 2 4 6 8 10 12 14
measurement P should track its order Pref,GFM at steady state.
1 To test the basic control functions of the GFM, we con-
id

0.5 ducted two consecutive experiments.

0 2 4 6 8 10 12 14
0.8 • At t = 4 s, Pref,GFM increases from 0.5 pu to 1.0 pu.
0.4
• At t = 10.1 s, Vd,ref decreases from 1.0 pu to 0.95 pu.
iq

0
0 2 4 6 8 10 12 14
The dq voltage and current components are all based on
Time (s)
the GFM’s synchronization frame. It can be seen that the
(b)
d-axis outer control functions properly by keeping Vq at 0
Fig. 9: Cases A1 and A2: Hardware experiment results. Basic control function for all the time. The P-f droop control functions properly to
tests for a step change in real power order from 0.5 to 1 pu at 4 s and a step
change in Vd order from 1 to 0.95 pu at 10 s. (a) GFL; (b) GFM. keep the difference between the power order and the power
measurement at 0. The power measurement can track the
power order at 4 s. Vd is kept at 1 pu before 10.1 s and
voltage measurement (Vd based on the PLL frame) follows the follows the order to be 0.95 pu.
command to decrease from 1.0 pu to 0.95 pu after 10.1 s. 3) Analysis: Comparison of GFL and GFM: While all the
Since the control uses PLL-based voltage synchronization, rest of signals behave very similar as their counterparts in the
the q-axis voltage of the PCC bus is forced to be zero. It can GFL control case, Vq is shown to have different transients upon
be seen that Vq based on the PLL frame is 0 all the time, power order increase. In the GFL case, Vq has a negligible
except small transients during the command change instants. upward spike at t = 4 s while in the GFM case, Vq has
Before 10.1 s, the PCC bus voltage is kept at 1 pu. The real an obvious downward spike. The two types of spikes are
power and the reactive are associated with the dq-axis currents explained in Fig. 10 and also as follows.

Authorized licensed use limited to: INDIAN INSTITUTE OF TECHNOLOGY GUWAHATI. Downloaded on January 06,2025 at 07:20:08 UTC from IEEE Xplore. Restrictions apply.
© 2024 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.See https://www.ieee.org/publications/rights/index.html for more information.
 This article has been accepted for publication in IEEE Transactions on Sustainable Energy. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/TSTE.2024.3520989

behaves as expected.

B. Category 2: GFL/GFM mode switching

It has to be noted that control mode switching during
operation is manually conducted by the system operators.
It is not the intention of this paper to develop an auto-
matically triggered control mode switching strategy. Control
mode switching will be executed by system operators. They
will make necessary observation before executing control
mode switching to avoid large transients. For example, while
switching for the first time between GFL and GFM, it is
necessary to observe the difference in angle between voltage
synchronization and power synchronization. The switching has
to be triggered at a minimum angle difference, to avoid a big
transient or even unstable condition.
1) Case B1: Control mode switching and system condition
switching: GFL and GFM mode switching and operational
condition switching between grid-connected and islanding
have been tested. The event sequence is shown in Fig. 12a,
while the experiment results shown in Figs. 11 and 12b. Note
that the dq components are based on the synchronizing frame.
When GFL is enabled, this frame is the PLL frame. When
Fig. 10: The positions of the PCC bus voltage vector vs. the synchronization GFM is enabled, this frame is the power synchronization angle
frame upon a step increase in the power order. Red color notates the variables
subject to immediate change. (a) GFL. (b) GFM. based frame.
Fig. 11 shows the experiment results from the four oscillo-
scopes where each measurement has a distinct scale and it is
In the GFL case, upon the power order increase, the d-axis difficult to compare a current order vs. a current measurement.
current order increases due to the power control logic and in Hence, the signals are replotted in Fig. 12 to group similar
turn the d-axis current measurement increases. This increase signals together using the same scale. The operation panel
is also reflected in the increase in the PCC bus voltage angle, designed in LabVIEW. GFL and GFM mode switching can
since id influences the angle according to Eq. (3). Vq in the be done by pushing a button in the front panel, while the
PLL frame then experiences an increase since the PCC bus operational condition can be changed by physically switching
voltage vector leads the PLL frame. The PLL’s PI controller the breaker in the testbed.
follows up and enforces Vq to be 0. Hence, an upward spike a) Case B1 stage 1: control mode switching: From 0
is seen. s to 5 s, the VSC is operating to be connected to a grid.
In the GFM case, upon the power order change, the syn- When the control mode toggles between GFL and GFM, the
chronization angle increases immediately due to the power- d-axis current order and the synchronization angle have to be
frequency droop based synchronization control. The PCC bus switched from GFL’s to GFM’s. Additionally, when either one
voltage vector is initially aligned with the synchronization control block in GFL (or GFM) is enabled, it is desirable to
frame. Upon the sudden power order increase it is now lagging keep its counter part in GFM (or GFL) disabled, or the output
the synchronizing frame, resulting in its q-axis projection to of the block remains intact. Therefore, the input to the PI
the synchronizing frame becoming negative while its d-axis controller of the outer loop is subject to the mode change. If
projection decreases. The d-axis outer control then brings the GFL is enabled, the real power control has the power error
Vq back to 0, or realign the PCC bus voltage to the new as the input while the GFM’s angle control or Vq control has
synchronization frame. This is achieved by increasing id . its input set to 0. If GFM is enabled, the GFL’s real power
Therefore, Vq shows a downward spike upon the power order control then has its input switched to 0 while the GFM’s Vq
step change. control has its input as the Vq error.
Remarks: The difference of the synchronization methods of During the control mode switching, if the system’s operating
GFL and GFM leads to the difference in the relative position point (e.g., real power, voltage) does not change, a very smooth
of the PCC bus voltage vector vs. the synchronization frame. transition can be achieved.
Upon the power order increase, in GFL the PCC bus voltage Fig. 11 and Fig. 12b present the experiments results from
vector has an immediate increase in its angle while in GFM the the hardware testbed, showing the switching between different
synchronization frame has its angle immediately increased. In modes at the grid-connected condition from 0 s- 5 s. Initially
turn, in GFL the synchronization frame lags the voltage vector the VSC is in GFL control mode. At t = 2 s, the mode is
while in GFM the synchronization frame leads the voltage switched from GFL to GFM and there are some transients.
vector. This is due to the initial setting. When the test bed is energized,
Overall, the GFM can fulfill its control functions and the VSC is in the GFL control mode with the GFM control

Authorized licensed use limited to: INDIAN INSTITUTE OF TECHNOLOGY GUWAHATI. Downloaded on January 06,2025 at 07:20:08 UTC from IEEE Xplore. Restrictions apply.
© 2024 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.See https://www.ieee.org/publications/rights/index.html for more information.
 This article has been accepted for publication in IEEE Transactions on Sustainable Energy. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/TSTE.2024.3520989

measurement makes the power synchronization unit to have

an increased frequency and a synchronizing angle. Hence, in
Fig. 11b, ∆θPS is seen to have an upward spike. This then
leads to a decrease in Vq since now the PCC bus voltage
vector is lagging the synchronizing frame based on Fig. 10.
In turn, id,ref,GFM increases based on the d-axis outer control
logic and id follows the order and becomes 1 pu. Before
t = 2 s, the VSC is absorbing 0.35 pu reactive power. At the
moment of switching from GFL to GFM, the reactive power
increases, leading to an increased voltage magnitude Vd . The
(a) Reactive Power (blue), Switch indicator (Cyan) and Real Power (purple).
q-axis outer control works to bring the voltage back to 1 pu
by increasing the q-axis current reference id,ref,GFM and in
turn iq . In about 0.5 seconds, a new steady state is achieved
for all measurements.
In the next two mode switching events at 2.8 s and 4 s, since
the operating condition does not change at all, the switching
transients are negligible. It is also interesting to point out
at t = 2 s, when the new synchronization angle based on
power measurement increased, in turn the PCC bus voltage
vector also experiences an increase in angle. Viewed from
the PLL frame, the voltage vector has its q-axis projection
decreases and hence the PLL shows a downward spike in its
(b) fGFM (blue), fGFL (cyan), θGFL (purple) and θGFM (green). frequency. Hence, in Fig. 12b’s frequency plot, the dynamics
of the frequency measurements from the GFL and the GFM
have opposite trends.
Fig. 13 is the zoomed-in version of Fig. 12b from 2.5 s to
4.5 s. It can be seen that the current order transients are very
smooth, while there is discernible difference in ripples from
one mode to another mode.
b) Case B1 stage 2: operation condition switching: At
t = 5 s, while the VSC is running in GFM control, the
system is switched to the islanding condition by switching
off the breaker manually thereby disconnecting the grid. The
experiment results from 5 s to 7.25 s shown in Fig. 11 and
(c) id,ref,GFL (blue), iq,ref,GFL (cyan), Vd (purple), Vq (green).
Fig. 12b are for the islanding condition. From 7.25 s to 10 s,
the operating condition is back to the grid-connected condition.
The system is operating smoothly for the condition switching.
In islanding operation from 5 s to 7.25 s, the power reference is
1 pu while the load consumes 0.75 pu. In turn, the microgrid’s
frequency rises to 60.45 Hz based on 3% P-f droop parameter:
∆f = −(0.75 − 1) × 3% × 60 Hz = 0.45 Hz. (14)
The relative angle against the grid should keep increasing.
Both the frequency and the angle responses in Fig. 12b show
the expected responses.
(d) id (blue), iq (cyan), id,ref,GFM (purple), iq,ref,GFM (green).
At t = 7.25 s, the breaker is again manually switched on
and thereby the VSC is reconnecting to the grid. It has to be
Fig. 11: Case B1: display in four oscilloscopes. The system runs initially
in GFL control. It switches from GFL to GFM at 2 s, back to GFL at 2.8
mentioned that the angle between the synchronizing frame and
s, and finally GFM after 4 s. The testbed is operating in the grid connected the grid (∆θPS ) is monitored closely. Only when this angle
mode before 5 s while in the islanding mode after 5 s. Note each signal has is close to 0, the switching on action is initiated. This makes
a distinct scale making comparison slightly difficult.
sure that at that moment, the PCC voltage space vector almost
aligns with the grid voltage space vector even their rotating
speeds (frequencies) are different. The resulting transient is
in the off mode. Hence, when the GFM is enabled, its dq-axis minimized. It can be seen upon the breaker switching on, the
current order id,ref,GFM is 0 and this results in the dq current angle immediate stops increasing and stays at a constant value
and the real and reactive power measurements moving towards after a quick transient. The dq-axis voltages also show very
0 at the instant of switching. The reduction of the real power smooth transients.

Authorized licensed use limited to: INDIAN INSTITUTE OF TECHNOLOGY GUWAHATI. Downloaded on January 06,2025 at 07:20:08 UTC from IEEE Xplore. Restrictions apply.
© 2024 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.See https://www.ieee.org/publications/rights/index.html for more information.
 This article has been accepted for publication in IEEE Transactions on Sustainable Energy. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/TSTE.2024.3520989

10

(a)

(b)

Fig. 12: Case B1. (a) Sequence of the events. (b) The experiment results based on data extracted from the oscilloscope results shown in Fig. 11.

Hz. Therefore, the control system will be notified to switch

1
to GFM control mode with a delay of 6 cycles. This delay
Mode

0.5
1: GFL 0: GFM emulates the communication delay between the sensor and
0 a controller. The hardware experiment results of the system
2.5 3 3.5 4 4.5 with/without control switching protection are shown in Fig. 15.
1.2
id id,ref,GFL id,ref,GFM
1.1
id

2.5 3 3.5 4 4.5

0.5 iq iq,ref,GFL iq,ref,GFM

0.4
iq

0.3
Fig. 14: Case B2: Event sequence.
2.5 3 3.5 4 4.5
Time (s)
It can be seen that when the IBR sticks with the GFL control
Fig. 13: Smooth switching from GFL to GFM and vice versa. mode, a sudden islanding event breaks the system (shown in
Fig. 15b. On the other hand, if we are able to switch the IBR
from GFL to GFM in 6 cycles after the islanding event, the
In short, the GFM control is capable to operate in both IBR can serve loads as a standalone microgrid Fig. 15a. The
grid-connected and islanding conditions. load in the islanding testbed is 0.75 pu. So after islanding, the
2) Case B2: Sudden Islanding: Fig. 14 shows the event IBR’s power drops from 1 pu to 0.75 pu, and the frequency
sequence. Initially, the grid-connected system was operating increases from 60 Hz to 60.45 Hz.
in the GFL control mode. The grid was disconnected at 0.5
seconds. The frequency sensed by the PLL of the IBR will be C. Category 3: Grid Fault Tests
increase rapidly and exceed the normal range of 59 Hz to 61 In this section, three different kinds of grid disturbances

Authorized licensed use limited to: INDIAN INSTITUTE OF TECHNOLOGY GUWAHATI. Downloaded on January 06,2025 at 07:20:08 UTC from IEEE Xplore. Restrictions apply.
© 2024 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.See https://www.ieee.org/publications/rights/index.html for more information.
 This article has been accepted for publication in IEEE Transactions on Sustainable Energy. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/TSTE.2024.3520989

11

1
Power

0 P Lab Q Lab
P Sim Q Sim
-1
0 0.5 1 1.5
70
f Sim
f (Hz)

65 f Lab Fig. 16: Case C1: Event sequence.

60
0 0.5 1 1.5
200
in GFL or GFM for the grid voltage dip event. Both hardware
(deg)

0 Lab experiment and the computer simulation results are presented.

Sim
-200
0 0.5 1 1.5
The results from the two environment show good agreement.
In the grid-following control, PLL is used for synchro-
Voltage

1
0.5 V d Lab V q Lab nization, which forces Vq to 0 at steady state. Upon a grid
V d Sim V q Sim
0 voltage dip, viewed from a synchronously rotating frame
0 0.5 1 1.5
aligned with the PCC bus voltage space vector at steady
1
Current

state, Vd experiences a decrease while Vq experiences an

0.5 id Lab iq Lab
increase immediately. This means that the PCC voltage vector
0 id Sim iq Sim
0 0.5 1 1.5 experiences an immediate increase in angle. PLL’s angle then
Time (s)
follows this angle. Therefore the PLL frequency and angle will
(a)
also increase. The detailed analysis can be found in [4].
Comparing Figs. 17a and 17b, we can see the clear differ-
ence between GFL and GFM. Particularly, the variations of
frequency and angle of GFL are much greater than those of
Power

1 P
0 Q
GFM. Indeed, large angle deviation has triggered 389 MW
-1
0 0.5 1 1.5 solar PV tripping in 2021 Texas Odessa event [3]. Analysis
75
70 in [4] has shown that the PLL-based synchronization is the
f (Hz)

f GFL
65 f GFM main contributor. In the GFM case, the synchronizing frame’s
60
0 0.5 1 1.5
angle increases upon voltage dip due to the reduction in real
200 power measurement. Based on the experiment and simulation
(deg)

0 results, it can be seen that this frame almost aligns with the
-200 PCC voltage vector, leading to Vq maintaining 0 in the first 6
0 0.5 1 1.5
cycles.
Voltage

1 Vd
0.5
Also notably, the dynamic response of real power of GFL is
Vq
0 shown to have less variation compared to that of GFM. This is
0 0.5 1 1.5 due to the constant real power control applied in the GFL. On
1 the other hand, the GFM provides frequency support through
Current

id
0.8
0.6 iq
0.4 real power variation.
0.2
0 0.5 1 1.5 The dynamic responses of reactive power Q and PCC bus
Time (s) voltage magnitude Vd are comparable for the GFL and GFM
cases due to the same q-axis AC voltage control. The GFM’s
(b)
control is 3 + 40/s while the GFL’s AC voltage control is
Fig. 15: Case B2: Hardware experiment results of a sudden islanding event 0.4 + 40/s. This leads to slightly less variation in Vd for the
at t = 0.5 s. (a) GFL to GFM switch occurs at about 0.6 s. (b) No control
mode switching. GFM case.
2) Case C2: Weak grid tests: The weak grid tests have
been conducted in computer simulation and the results are
are presented and discussed. The first two cases focus on shown in Fig. 19a. The testbed was initially operated in GFL
individual GFL or GFM’s performance, while the last case control mode before 2 seconds. At 1 second, there is a line
demonstrates control mode switching upon a grid fault. tripping event changing the grid strength of the testbed from
1) Case C1: Three-phase grid voltage magnitude dip: short circuit ration (SCR) at 3 to SCR at 1.5. Fig. 19a shows
It is important to test the responses of an IBR towards grid the system performed very well in the GFL mode when the
disturbances. In this experiment (the event sequence is shown grid strength is 3. However, when the line-tripping event
in Fig. 16), the responses of the VSC towards a balanced happened, there are sustained 12.5 Hz oscillations observed
voltage dip when operating in either GFL or GFM control from all measured components. After switching the control
are compared side by side. model to GFM at t = 2 s, the oscillations were damped-out
A balanced voltage magnitude dip is applied at the grid within 0.4 seconds. This case shows the tolerance of weak
bus. Fig. 18 shows the instantaneous grid voltage measured grid condition regarding the GFM control is higher than GFL
using an oscilloscope. The voltage dip was triggered for 6 control. Thus, the switchable control design can be used to deal
cycles from 1 to 0.3 pu. Fig. 17 shows the response of IBR with weak grid condition if the system was initially operated

Authorized licensed use limited to: INDIAN INSTITUTE OF TECHNOLOGY GUWAHATI. Downloaded on January 06,2025 at 07:20:08 UTC from IEEE Xplore. Restrictions apply.
© 2024 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.See https://www.ieee.org/publications/rights/index.html for more information.
 This article has been accepted for publication in IEEE Transactions on Sustainable Energy. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/TSTE.2024.3520989

12

1.5 1.5

P
1 1
0.5 0.5
0.7 0.8 0.9 1 1.1 1.2 1.3 1.4 1.5 1.6 1.7 0.7 0.8 0.9 1 1.1 1.2 1.3 1.4 1.5 1.6 1.7

0.5 0.5
Q

Q
0 0
-0.5 -0.5
0.7 0.8 0.9 1 1.1 1.2 1.3 1.4 1.5 1.6 1.7 0.7 0.8 0.9 1 1.1 1.2 1.3 1.4 1.5 1.6 1.7
62 62
f (Hz)

f (Hz)
60 60
58 58
0.7 0.8 0.9 1 1.1 1.2 1.3 1.4 1.5 1.6 1.7 0.7 0.8 0.9 1 1.1 1.2 1.3 1.4 1.5 1.6 1.7
(deg)

(deg)
40 40
PLL

20 20

PS
0 0
0.7 0.8 0.9 1 1.1 1.2 1.3 1.4 1.5 1.6 1.7 0.7 0.8 0.9 1 1.1 1.2 1.3 1.4 1.5 1.6 1.7

1.2 1.2
Vd

Vd
1 1
0.8 0.8
0.7 0.8 0.9 1 1.1 1.2 1.3 1.4 1.5 1.6 1.7 0.7 0.8 0.9 1 1.1 1.2 1.3 1.4 1.5 1.6 1.7
0.2 0.2
0 0
Vq

Vq
-0.2 -0.2

0.7 0.8 0.9 1 1.1 1.2 1.3 1.4 1.5 1.6 1.7 0.7 0.8 0.9 1 1.1 1.2 1.3 1.4 1.5 1.6 1.7
2 2
id

id
1 1

0.7 0.8 0.9 1 1.1 1.2 1.3 1.4 1.5 1.6 1.7 0.7 0.8 0.9 1 1.1 1.2 1.3 1.4 1.5 1.6 1.7
0.5 0.5
0 0
iq

-0.5 iq -0.5

0.7 0.8 0.9 1 1.1 1.2 1.3 1.4 1.5 1.6 1.7 0.7 0.8 0.9 1 1.1 1.2 1.3 1.4 1.5 1.6 1.7

Time (s) Time (s)

(a) (b)

Fig. 17: Case C1: hardware experiment and computer simulation results of the grid voltage dip tests for GFL and GFM control. The blue line indicates the
hardware result and the red dotted line indicate the simulation result. (a) GFL. (b) GFM.

simulation results are shown in Fig. 21a. As a comparison,

simulation results without control mode switching are also
shown in Fig. 21b.
It can be seen that the system sticking with the GFL control
goes unstable after the event, while the system comes back to
steady state with the GFL to GFM control switch.

V. C ONCLUSION
This paper presents the design and implementation of a
switchable GFL and GFM control for a VSC in hardware.
Fig. 18: Case C1: three-phase voltage drip applied to the grid voltage. This control enables the VSC to operate in the grid-connected
condition as a GFL controller or a GFM controller. When the
in GFL control mode. Fig. 19b shows the GFM control mode’s VSC is operating in the islanded condition, it operates in the
boundary grid strength condition. The grid strength of the GFM control mode. The switchable GFL and GFM utilize
system changed from SCR = 0.95 to SCR = 0.8 at 1 second. the same sensors and has control logics designed to minimize
It can be seen that the system loses synchronizing stability. transients during mode transients. The performance of the VSC
The boundary condition result shows that the GFM can be has been tested in a hardware testbed. And the hardware ex-
operated under very low grid strength condition compared to periment results show satisfactory performance, demonstrating
a GFL, as remarked in [8]. the effectiveness of the VSC control design and the switching
3) Case C3: Line tripping and grid voltage dip: In this logics. In addition to offering the feasible and robust control
case, we demonstrate the benefit of control mode switching design that has been successfully prototyped in hardware, this
upon a grid fault. The grid fault occurs in a remote location. paper also contributes to in-depth dynamical analysis of GFL
According to [31], a remote grid fault makes the equivalent and GFM by associating responses observed in measurements
Thévenin voltage viewed from the IBR drop. Therefore, an with control logics. Particularly, this paper captures the subtle
event timeline is designed and shown in Fig. 20. The grid difference caused by different synchronization methods.
strength measured by SCR and the grid voltage drop simul-
taneously at 0.5 seconds emulating a balanced 3-phase fault R EFERENCES
followed up by a line tripping event. The IBR switches from [1] A. Yazdani and R. Iravani, Voltage-sourced converters in power systems:
GFL to GFM after the fault with a delay of 6 cycles. The modeling, control, and applications. John Wiley & Sons, 2010.

Authorized licensed use limited to: INDIAN INSTITUTE OF TECHNOLOGY GUWAHATI. Downloaded on January 06,2025 at 07:20:08 UTC from IEEE Xplore. Restrictions apply.
© 2024 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.See https://www.ieee.org/publications/rights/index.html for more information.
 This article has been accepted for publication in IEEE Transactions on Sustainable Energy. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/TSTE.2024.3520989

13

GFL GFM
1.5 1.5
Power
1 P 1 P

Power
0.5 Q 0.5 Q
0
-0.5 0
0 0.5 1 1.5 2 2.5 3 -0.5
61 0 0.5 1 1.5
f (Hz)

60.5 62
f GFL

f (Hz)
60
60 f GFM
59.5
0 0.5 1 1.5 2 2.5 3
(deg)

58
60 0 0.5 1 1.5
40 200
PCC

(deg)
20
0
0 0.5 1 1.5 2 2.5 3
1.5
Voltage

Vd -200
1 0 0.5 1 1.5
0.5 Vq
0

Voltage
-0.5 1 Vd
0 0.5 1 1.5 2 2.5 3 0.5 Vq
0
Current

1 id
0.5
0 iq 0 0.5 1 1.5
-0.5

Current
1 id
0 0.5 1 1.5 2 2.5 3 0.5 iq
Time (s) 0
-0.5
0 0.5 1 1.5
(a) Time (s)

(a)
1
Power

P
0 Q
-1 1.5
1 P

Power
0 0.5 1 1.5 2 2.5 3
0.5 Q
0
f (Hz)

62 -0.5

60 0 0.5 1 1.5
0 0.5 1 1.5 2 2.5 3 65
f GFL
(deg)

f (Hz)

200 60 f GFM
0
PCC

55
-200 0 0.5 1 1.5
0 0.5 1 1.5 2 2.5 3 200
(deg)
Voltage

1 Vd 0
0.5
0 Vq
-200
0 0.5 1 1.5 2 2.5 3 0 0.5 1 1.5
1.5
Voltage

1 Vd
Current

id 1
0 0.5 Vq
iq 0
-1 -0.5
0 0.5 1 1.5 2 2.5 3 0 0.5 1 1.5
Time (s) 1
Current

id
0.5
0 iq
(b) -0.5
0 0.5 1 1.5
Fig. 19: Case C2: Weak grid condition tests. (a) Grid strength changes from
Time (s)
SCR = 3 to SCR = 1.5 at t = 1 s, (b) GFM control. Grid strength changed
from SCR = 0.95 to SCR = 0.85 at t = 1 s. (b)

Fig. 21: Case C3: Computer simulation results for a grid voltage dip and
line tripping event at t = 0.5 s. (a) GFL to GFM switching at 0.6 s; (b) no
control mode switching.

pp. 1–10, 2023.

[5] Y. Cheng, L. Fan, J. Rose, S.-H. Huang, J. Schmall, X. Wang, X. Xie,
J. Shair, J. R. Ramamurthy, N. Modi, C. Li, C. Wang, S. Shah,
B. Pal, Z. Miao, A. Isaacs, J. Mahseredjian, and J. Zhou, “Real-world
subsynchronous oscillation events in power grids with high penetrations
of inverter-based resources,” IEEE Transactions on Power Systems,
vol. 38, no. 1, pp. 316–330, 2023.
Fig. 20: Case C3: Event sequence of a grid voltage dip and line tripping [6] L. Fan et al., “Real-world 20-Hz IBR subsynchronous oscillations:
case. Signatures and mechanism analysis,” IEEE Transactions on Energy
Conversion, vol. 37, no. 4, pp. 2863–2873, 2022.
[7] D. Jovcic, L. Lamont, and L. Xu, “Vsc transmission model for analytical
studies,” in 2003 IEEE Power Engineering Society General Meeting
[2] R. Teodorescu, M. Liserre, and P. Rodriguez, Grid converters for (IEEE Cat. No.03CH37491), vol. 3, 2003, pp. 1737–1742 Vol. 3.
photovoltaic and wind power systems. John Wiley & Sons, 2011. [8] NERC White Paper. (2021, December) Grid Forming Technology: Bulk
[3] Joint NERC and Texas RE Staff Report. (2021, September) Odessa Power System Reliability Considerations.
Disturbance. [9] L. Zhang, L. Harnefors, and H.-P. Nee, “Power-synchronization control
[4] L. Fan, Z. Wang, and Z. Miao, “Large angle deviation in grid-following of grid-connected voltage-source converters,” IEEE Transactions on
ibrs upon grid voltage dip,” IEEE Transactions on Energy Conversion, Power Systems, vol. 25, no. 2, pp. 809–820, 2010.

Authorized licensed use limited to: INDIAN INSTITUTE OF TECHNOLOGY GUWAHATI. Downloaded on January 06,2025 at 07:20:08 UTC from IEEE Xplore. Restrictions apply.
© 2024 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.See https://www.ieee.org/publications/rights/index.html for more information.
 This article has been accepted for publication in IEEE Transactions on Sustainable Energy. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/TSTE.2024.3520989

14

[10] T. Liu and X. Wang, “Transient stability of single-loop voltage- HUAZHAO DING (Graduate Student Member,
magnitude controlled grid-forming converters,” IEEE Transactions on IEEE) received the B.S. degree in electrical engi-
Power Electronics, vol. 36, no. 6, pp. 6158–6162, 2020. neering from the University of Jinan, Shandong,
[11] W. Du, Z. Chen, K. P. Schneider, R. H. Lasseter, S. P. Nandanoori, China, in 2018, and the M.S.E.E. degree from the
F. K. Tuffner, and S. Kundu, “A comparative study of two widely used University of South Florida, Tampa, FL, USA, in
grid-forming droop controls on microgrid small-signal stability,” IEEE 2020, where he is currently pursuing the Ph.D.
Journal of Emerging and Selected Topics in Power Electronics, vol. 8, degree. His research interests include transient elec-
no. 2, pp. 963–975, 2019. tromagnetic simulation, modeling of inverter-based
[12] T. Qoria, Q. Cossart, C. Li, X. Guillaud, F. Colas, F. Gruson, and resources, and dynamic analysis of power systems.
X. Kestelyn, “Deliverable 3.2: Local control and simulation tools for
large transmission systems,” MIGRATE project, 2018.
[13] R. Rosso, X. Wang, M. Liserre, X. Lu, and S. Engelken, “Grid-
forming converters: Control approaches, grid-synchronization, and future
trends—a review,” IEEE Open Journal of Industry Applications, vol. 2,
pp. 93–109, 2021.
[14] Y. Li, Y. Gu, and T. C. Green, “Revisiting grid-forming and grid-
following inverters: A duality theory,” IEEE Transactions on Power
Systems, vol. 37, no. 6, pp. 4541–4554, 2022.
[15] Y. Lin, J. H. Eto, B. B. Johnson, J. D. Flicker, R. H. Lasseter, H. N. Rabi Kar (Graduate Student Member, IEEE) re-
Villegas Pico, G.-S. Seo, B. J. Pierre, and A. Ellis, “Research roadmap ceived the B.S degree in electrical engineering from
on grid-forming inverters,” National Renewable Energy Lab.(NREL), the Biju Patnaik University of technology, Odisha,
Golden, CO (United States), Tech. Rep., 2020. India, in 2012, and the M.S.E.E degree from the
[16] J. Driesen and K. Visscher, “Virtual synchronous generators,” in 2008 University of South Florida, Tampa, FL, USA, in
IEEE Power and Energy Society General Meeting - Conversion and 2017. He is currently working toward Ph.D degree at
Delivery of Electrical Energy in the 21st Century, 2008, pp. 1–3. the Smartgrid Power System laboratory in University
[17] L. A. M. Lima and E. H. Watanabe, “Hybrid control scheme for of South Florida. His research interests include mod-
vsc presenting both grid-forming and grid-following capabilities,” IEEE eling and control, harmonic analysis, and protection
Transactions on Power Delivery, vol. 37, no. 6, pp. 4570–4581, 2022. studies for inverter-based resources.
[18] F. Katiraei, M. R. Iravani, and P. W. Lehn, “Micro-grid autonomous op-
eration during and subsequent to islanding process,” IEEE Transactions
on power delivery, vol. 20, no. 1, pp. 248–257, 2005.
[19] C.-L. Chen, Y. Wang, J.-S. Lai, Y.-S. Lee, and D. Martin, “Design
of parallel inverters for smooth mode transfer microgrid applications,”
IEEE Transactions on Power Electronics, vol. 25, no. 1, pp. 6–15, 2009.
[20] H. Karimi, A. Yazdani, and R. Iravani, “Robust control of an au-
tonomous four-wire electronically-coupled distributed generation unit,”
IEEE Transactions on Power Delivery, vol. 26, no. 1, pp. 455–466, 2010. Zhixin Miao (Senior Member, IEEE) received the
[21] M. B. Delghavi and A. Yazdani, “A unified control strategy for elec- B.S.E.E. degree from the Huazhong University of
tronically interfaced distributed energy resources,” IEEE Transactions Science and Technology,Wuhan, China, in 1992, the
on Power Delivery, vol. 27, no. 2, pp. 803–812, 2012. M.S.E.E. degree from the Graduate School, Nanjing
[22] Z. Wang, Z. Miao, and L. Fan, “Practical start-up process of multiple Automation Research Institute (Nanjing, China) in
grid-tied voltage-sourced inverters in laboratory,” in 2021 North Amer- 1997, and the Ph.D. degree in electrical engineer-
ican Power Symposium (NAPS). IEEE, 2021, pp. 1–6. ing from West Virginia University, Morgantown, in
[23] L. Bao, L. Fan, Z. Miao, and Z. Wang, “Hardware Demonstration 2002.
of Weak Grid Oscillations in Grid-Following Converters,” 2021 North Currently, he is a full professor at the University
American Power Symposium, NAPS 2021, 2021. of South Florida (USF), Tampa. Prior to joining
[24] R. Mittal, Z. Miao, and L. Fan, “Grid Forming Inverter: Laboratory- USF in 2009, he was with the Transmission Asset
Scale Hardware Test Bed Setup and Weak Grid Operation,” 2021 North Management Department with Midwest ISO, St. Paul, MN, from 2002 to
American Power Symposium, NAPS 2021, 2021. 2009. His research interests include power system stability, microgrids, and
[25] Y. Li, L. Fan, Y. Zhou, and Z. Miao, “Stability enhancement module renewable energy. Dr. Miao serves as an associate editor for IEEE trans.
for grid-following converters: Hardware implementation and validation,” Sustainable Energy.
International Transactions on Electrical Energy Systems, vol. 31, no. 11,
p. e13115, 2021.
[26] L. Fan and Z. Miao, “Modeling and Stability Analysis of Inverter-Based
Resources,” CRC Press. https://doi.org/10.1201/9781003323655, 2023.
[27] L. Fan, Z. Miao, D. Ramasubramanian, and H. Ding, “Operational
challenges of solar pv plus storage power plants and modeling recom-
mendations,” IEEE Open Access Journal of Power and Energy, vol. 10,
pp. 477–489, 2023.
Lingling Fan (Fellow, IEEE) received the B.S. and
[28] S. Mishra, D. Ramasubramanian, and P. C. Sekhar, “A seamless control
M.S. degrees in electrical engineering from South-
methodology for a grid connected and isolated PV-diesel microgrid,”
east University, Nanjing, China, in 1994 and 1997,
IEEE Transactions on Power Systems, vol. 28, no. 4, pp. 4393–4404,
respectively, and the Ph.D. degree in electrical engi-
2013.
neering from West Virginia University, Morgantown,
[29] S. M. Ashabani and Y. A. R. I. Mohamed, “A flexible control strategy
in 2001.
for grid-connected and islanded microgrids with enhanced stability using
Currently, she is a full professor with the Uni-
nonlinear microgrid stabilizer,” IEEE Transactions on Smart Grid, vol. 3,
versity of South Florida, Tampa. She was a Senior
no. 3, pp. 1291–1301, 2012.
Engineer in the Transmission Asset Management
[30] L. Fan, “Modeling type-4 wind in weak grids,” IEEE trans. Sustainable
Department, Midwest ISO, St. Paul, MN, from 2001
Energy, vol. 10, no. 2, pp. 853–864, 2019.
to 2007, and an Assistant Professor with North
[31] R. Kar, T. Banerjee, Z. Miao, and L. Fan, “Thévenin equivalent repre-
Dakota State University, Fargo, from 2007 to 2009. Her research interests
sentation of meshed grids for ibr dynamic phenomena replication,” in
include power systems and power electronics. Dr. Fan serves as the Editor-
2024 IEEE Power & Energy Society General Meeting (PESGM). IEEE,
in-Chief for IEEE Electrification Magazine and an associate editor for IEEE
2024, pp. 1–5.
trans. Energy Conversion.

Authorized licensed use limited to: INDIAN INSTITUTE OF TECHNOLOGY GUWAHATI. Downloaded on January 06,2025 at 07:20:08 UTC from IEEE Xplore. Restrictions apply.
© 2024 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.See https://www.ieee.org/publications/rights/index.html for more information.