Partial Grid Forming Concept for 100%

Inverter-Based Transmission Systems

Uros Markovic∗ , Ognjen Stanojev∗ , Petros Aristidou§ , Gabriela Hug∗
∗ EEH - Power Systems Laboratory, ETH Zurich, Physikstrasse 3, 8092 Zurich, Switzerland
§School of Electronic and Electrical Engineering, University of Leeds, Leeds LS2 9JT, UK
Emails: {markovic, hug}@eeh.ee.ethz.ch, [email protected], [email protected]

Abstract—With the current trends in renewable energy in- the voltage magnitude and frequency in the system, and the
tegration, the concept of a 100% inverter-based power system latter providing predefined power to the energized grid. An
is becoming more of a reality. However, the existing Voltage extension to this categorization was presented in [8], where
Source Converter (VSC) control schemes for such systems focus
mostly on the operation of low-voltage microgrids, which have a new class of VSC control mode was defined and named
different requirements from the transmission system perspective. grid-supporting. Essentially, it is a modification of the first
This paper proposes a new classification of VSC control strategies two modes, with incorporation of additional high-level control
depending on their mode of operation. Then, the concept of loops that enable regulation of an AC voltage vector via
partial grid forming VSC is introduced and it is shown that power output. Since a grid-supporting VSC can be represented
a system with zero rotational inertia can operate without a
dedicated grid-forming VSC unit, but rather with partial forming either as a voltage or a current source, it possesses standalone
of key system characteristics distributed across different VSC capabilities only under certain control configurations.
units. The performance of this approach is tested on detailed However, this somewhat general classification is quite re-
VSC models developed in both MATLAB Simulink and virtual strictive for the multifaceted nature of the problem, and
Hardware-In-the-Loop (vHIL) platforms. Furthermore, an in- occasionally unclear, as similar terms are used across the
vestigation towards necessary converter and network criteria for
providing a stable system under the proposed control concepts literature for different VSC concepts, as in case of a grid-
is presented. supporting model presented in [9]. Furthermore, all of the
proposed configurations assume that a converter is either
Index Terms—voltage source converter (VSC), grid-forming, forming both the voltage and frequency in the system, or
grid-following, grid-supporting, partial system forming measuring them via a Phase-Locked Loop (PLL) unit. While
this might prove sufficient for low-voltage microgrids, it is not
I. I NTRODUCTION necessarily optimal for VSC control on a transmission system
Power systems are currently facing a major transition from level, as shown in [10].
large Synchronous Machines (SMs) to smaller generation The concept of a large-scale power system with high PE
units, interfaced via Voltage Source Converters (VSCs). The penetration is now becoming a reality [11], which makes the
presence of existing SMs still allows for the majority of inadequacy of current VSC mode classification even a greater
inverter-based generation to be controlled as grid-following problem. Several studies have addressed the requirements for
(also termed grid-feeding) units [1], [2]. Nevertheless, this 100% inverter-based power systems, with a focus mostly on
mode of operation relies heavily on the assumption of a stiff new control architecture and ancillary services [12], [13],
AC grid and accurate tracking of the already formed frequency as well as the feasibility criteria under various operational
and voltage; this assumption collapses in the case of sys- scenarios [14]. However, all of the conclusions have been
tems with 100% Power Electronics (PE) penetration. Hence, drawn under the premise of a converter operating either as a
several grid-forming control strategies have been proposed grid-forming or grid-following unit. While a need for a more
that provide certain SM-like properties to the VSCs, such as versatile control categorization has been indicated in [13], no
standalone and black-start mode of operation, frequency and solution has been suggested thus far.
voltage support, and synchronization capabilities [3], [4]. The The contribution of this work is three-fold. First, we propose
forming aspect of VSC control has been mostly addressed a new, more comprehensive, classification for control strategies
in the context of microgrids, where frequent configuration of VSCs. The concept of partial grid-forming control strategy
changes can result in switching between the grid-connected is introduced and it is shown that a 100% PE-interfaced
and islanded mode of operation [5]–[7]. system can operate without a dedicated grid-forming unit, but
Due to the nature of the problem, all of the proposed rather with partial forming of individual system parameters
approaches distinguish between a forming and a following distributed across different converter units. Second, we test
mode of the VSC control; with the first one solely establishing the performance of this approach using a detailed VSC model
with a state-of-the-art control structure, developed in both
This project has received funding from the European Union’s Horizon 2020 MATLAB Simulink and virtual Hardware-In-the-Loop (vHIL)
research and innovation programme under grant agreement No 691800. This
paper reflects only the authors’ views and the European Commission is not platforms. Finally, we investigate the necessary conditions and
responsible for any use that may be made of the information it contains. grid configuration criteria for providing a stable system under

978-1-5386-7703-2/18/$31.00 ©2018 IEEE

the proposed approach. ω v
The remainder of the paper is structured as follows. In
Section II, the concept of partial grid forming is introduced, ω∗ ∆ω v∗ ∆v
together with the new converter mode classification. Section III
describes the VSC model used in this paper. Section IV ∆p ∆q
compares the transient response of different partial forming
configurations and investigates the necessary stability require- p∗ p q∗ q
ments, whereas Section V discusses the outlook of the study (a) (b)
and concludes the paper.
Fig. 1: Impact of a time-varying reference on droop charac-
II. PARTIAL G RID F ORMING teristics: (a) active power droop; (b) reactive power droop.

The proposed classification in [8] distinguishes between III. VSC C ONTROL S CHEME
VSCs controlled as a voltage or a current source, with grid-
An overview of the VSC model used in this work is shown
forming being the first and grid-following the latter type. The
in Fig. 2, consisting of an ideal DC voltage source, interfaced
grid-supporting mode can fall into both categories, as it is
through a DC/AC converter and an RLC filter to the grid. The
conceptually based on the previous two schemes with incor-
control scheme contains an outer loop which uses the voltage
poration of the droop control. According to the capability of
and current measurements to compute the desired voltage
individually establishing the voltage vector, all three proposed
magnitude and frequency by means of active and reactive
control modes may either use both the voltage magnitude and
power controllers. These signals are then passed through the
frequency setpoints, or none. However, from the perspective
inner control loop consisting of cascaded voltage and current
of the power system operation, these two characteristics are
controllers. The model also includes a grid synchronization
controlled independently. Furthermore, in order to obtain a
unit that provides the frequency reference for the outer control.
system-level categorization of the VSC operation modes, it is
important to make them independent of the particular device- A. Power Controllers
level implementation. Having in mind that an inverter-based transmission system is
One way to classify the system-level specifications is by dif- under investigation, it is justified to assume a full decoupling
ferentiating between which of the two voltage vector variables of active power and frequency from the reactive power and
(magnitude v and frequency ω) are regulated to constant values voltage terms. Hence, we can employ a standard droop char-
and which ones vary according to locally measured signals. acteristic for regulating the active and reactive power output
In other words, the voltage reference signals can be defined of the converter in the form of
either as constant setpoints (v0 , ω0 ) or as measurements inputs
(ṽ, ω̃), which yields four possible converter operation modes: ω = ω ∗ + Dp (p∗ − λ(s)p) (1)
1) Grid-forming (g-form): Establishes a complete voltage v = v ∗ + Dq (q ∗ − λ(s)q) (2)
vector, similarly to the grid-forming model in [8]: [v ∗ , ω ∗ ].
2) Frequency-forming (f -form): Independently forms the where p and q denote the active and reactive power mea-
frequency, while the voltage magnitude follows the measured surement, while ω and v refer to the frequency and voltage
reference: [ṽ, ω ∗ ]. output of the active and reactive power controller, respectively;
superscript ∗ indicates a respective reference value. The droop
3) Voltage-forming (v-form): Forms the magnitude of the
slopes Dp and Dq are imposed on the mismatch between a
voltage at the Point of Common Coupling (PCC), and syn-
predefined reference and an actual power measurement, passed
chronizes it accordingly via a PLL unit: [v ∗ , ω̃].
through a first-order Low-Pass Filter (LPF)
4) Grid-feeding (g-feed): Voltage vector is completely de-
pendent on the local measurements: [ṽ, ω̃]. ωc
λ(s) = (3)
In order to make the categorization completely independent ωc + s
of the local control, we consider a unified VSC configuration with ωc being the cutoff frequency. In the case of mea-
for all proposed unit types, with active and reactive power surements being used as reference inputs, the droop curve
being regulated through means of droop control. This is also might oscillate as the respective variable varies, providing an
another distinction from [8], where a grid-forming unit is not operating range depicted as shaded region in Fig. 1. Based
controlling the power output. on the computed frequency ω, a corresponding phase angle θ
This classification allows to describe the situation where needed for the (dq)-transformation is determined.
partial forming VSCs are used to operate the system, with
voltage magnitude and frequency being independently formed B. Synchronization Unit
at different locations in the grid. Such approach indicates that a The synchronization unit provides an adequate frequency
100% inverter-based network can operate without a dedicated reference to the outer control loop. In case of a g-form and
grid-forming unit, which has not been considered feasible thus f -form VSC this is just a constant setpoint ω0 , whereas a
far. PLL measurement ωpll is used otherwise. For this purpose,
 ...
idq
g iabc
g
Power dq
dq
Calculation edq
g eabc
g
Unit abc
abc
p q Cf
ω0

Reactive
v Sync. Lf
q∗ Power
Unit
v ∗ Controller
ω ∗ = [ω0 , ωpll ] Rf
h i
dq
edq
g , ig [v m , is ]
θ

Active SRF i∗s SRF v ∗m m vdc

Power Voltage Current PWM Cdc
p ∗ Controller Controller Controller
ω
vdc
Outer control loop Inner control loop

Fig. 2: General configuration of the proposed VSC control structure.

a Type-2 PLL operating in a Synchronous Reference Frame IV. R ESULTS

(SRF) has been implemented [15], which is based on the (dq)- In this section, feasibility of the proposed VSC operation
transformation of a balanced three-phase voltage signal eabc g modes is investigated through transient responses. Addition-
with a magnitude êg and a frequency ω̂: ally, different system conditions and network parameters are
 
dq abc cos (θ̂ − θpll ) considered in order to reflect important properties of the
eg = T p T c eg = êg (4) proposed configurations. For this purpose, a converter model
sin (θ̂ − θpll )
R R was implemented in two different platforms: (i) an averaged
with θ̂ = ω̂dt and θpll = ωpll dt, as well as T c and T p VSC model in MATLAB Simulink with the use of the Sim-
denoting the Clarke and Park transformation matrices. The PowerSystems toolbox; and (ii) a detailed three-phase inverter
synchronization is achieved by initially aligning the d-axis with full switching in vHIL platform from Typhoon HIL [16].
of SRF with the voltage vector, hence diminishing the q- The latter one is a software toolbox within a HIL toolchain
component. Reasonably assuming êg ≈ 1, this would equate that enables HIL models to run on a local computer instead
to sin (θ̂ − θpll ) ≈ 0, i.e. θ̂ ≈ θpll . The PLL is implemented of a HIL device. It is not a simulator, but rather a true HIL
as a PI controller of the phase angle difference, treating it as emulator that runs the same code of the HIL processor and
an error signal and driving it to zero: communicates with the same HIL toolchain. Therefore, it
! provides us with a higher degree of accuracy regarding the
pll Kipll performance of the developed control strategies.
ωpll = ω0 + Kp + eqg (5)
s The nominal parameters of the VSCs are as follows: AC
voltage Vn = 320 kV; DC voltage Vdc = 640 kV; active power
C. Inner Control Loop and Modulation
Pn = 1 GW; and frequency fn = 50 Hz.
The structure of the inner loop controllers follows the same
principles as in [8], [10], and can be described through a A. VSC Interactions
cascade computation of the (dq) reference vectors for the In this section, we investigate the 2-bus system depicted
switching current i∗s and modulation voltage v ∗m as: in Fig. 3, under all possible configuration scenarios listed in
 
∗ i v Kiv Table I; the green and red fields indicate if a scenario is stable
is = Kf eg + (v − eg ) Kp + + ωCf êg (6) or not. Each converter is connected through a transformer
s
  (Rtr , Ltr ), with a transmission line in between the two nodes
Ki
v ∗m = Kfv eg + (i∗s − ig ) Kpi + i + ωLf îg (7) (Rt , Lt , Ct ), and a resistive load (Rl ) supplied at the first bus.
s
As expected, it is confirmed that only scenarios involving
 T  T the forming of both voltage magnitude and frequency are sta-
where îg = −iqg , idg and êg = −eqg , edg , while Kp ,
Ki and Kf are the proportional, integral and feed-forward ble. This notion is straightforward for scenarios S11 , S12 and
gains, respectively; superscripts v and i denote the voltage S21 , as they include at least one g-form unit, which is similar
and current SRF controllers. The generated voltage reference
is combined with the DC-side voltage in order to determine the Rtr1 Ltr1 Rt Lt Rtr2 Ltr2

final (abc) vector of the modulation signal mabc as follows:

Rl Ct Ct

abc −1 dq −1 v ∗m VSC 1 VSC 2

m = (T p T c ) m = (T p T c ) (8)
vdc Fig. 3: Configuration of the investigated 2-bus system.
 eg 1 1.1
1.05 eg 1
1.02 eg 2 eg 2
vl vl
v [p.u.]

v [p.u.]

v [p.u.]
1.05
1 1
eg 1
eg 2
0.98 1 vl
0.95
0 5 10 15 20 25 0 5 10 15 20 25 0 5 10 15 20 25
t [s] t [s] t [s]

50.05 50.05 f1 50.05 f1

f1
f2 f2 f2
f [Hz]

f [Hz]

f [Hz]
50 50 50

49.95 49.95 49.95

0 5 10 15 20 25 0 5 10 15 20 25 0 5 10 15 20 25
t [s] t [s] t [s]

(a) (b) (c)

Fig. 4: Interactions between VSCs under different configuration scenarios: (a) S23 ; (b) S32 ; (c) collapse in S22 and S33 .

to the configuration of existing, fully operational microgrids. 1 p1

Furthermore, the simulations also show that the partial forming p2

p [p.u.]
of individual network parameters (scenarios S23 and S32 ) 0.5
results in a sustainable system, as presented through system
dynamics in Fig. 4a and Fig. 4b. For simplicity and clarity of
0
the results, only MATLAB responses are considered. 0 1 2 3 4 5 6
In both cases, the converter aligns the non-formed vari- t [s]
able according to the output provided by the other unit; f2 (a)
follows f1 , while eg1 follows eg2 in scenario S23 , and vice
versa in scenario S32 . Since frequency is a global variable, 50.5
the synchronization of VSCs is achieved with an adequate
f [Hz]

50
accuracy. However, the voltage mismatch is more drastic, as
it incorporates a voltage drop between the bus with a fixed 49.5 f1
voltage reference, and the one where it is being measured. f2
49
The oscillatory nature of the response is a consequence of the 0 1 2 3 4 5 6
droop characteristic elaborated in Section III. t [s]

System dynamics in scenarios S22 and S33 also indicate that (b)
the non-formed variables tend to align with the measurement, Fig. 5: System collapse in case of insufficient active power
as shown in Fig. 4c for the voltage magnitude in S22 and fre- reserves from an f -form VSC: (a) active power; (b) frequency.
quency in S33 , respectively. However, in the absence of a fixed
setpoint, none of these quantities converge to a steady state. forms the frequency, i.e. falls under g-form or f -form category.
The system instability under high load demand conditions also Similarly, a reactive power regulation is only provided by g-
prevails in certain scenarios, denoted with light green color in form and v-form units. On that note, if the cumulative available
Table I, as will be further elaborated. power reserves of all g-form and f -form VSCs in the system
B. Provision of Power Reserves do not meet the active load requirements, the system will
collapse. An example of such incident for scenario S32 is
In order for the system to be stable, the load demand must shown in Fig. 5, where an f -form unit tries to meet the
be met at all times. However, depending on the VSC mode, increase in demand, but fails due to insufficient capacity; v-
some units might not be responsive to sudden load changes, form VSC is unresponsive.
but rather follow the predefined power setpoints. Essentially, a
converter reacts to a change in active power demand only if it C. Sensitivity Analysis
As previously suggested, the length of a transmission line
TABLE I: Scenario Stability Results has a direct impact on the voltage drop, and hence the
VSC 1 \VSC 2 g-form f -form v-form g-feed
respective oscillations. For that reason, we have conducted a
g-form S11 S12 S13 S14 sensitivity analysis for scenarios S23 and S32 , with line lengths
f -form S21 S22 S23 S24 varying from 25 km to 300 km. The results are showcased in
v-form S31 S32 S33 S34 Fig. 6, where the voltage response of a f -form VSC in each
g-feed S41 S42 S43 S44 case has been depicted. It is clear that the maximum distance
 1
25 km
sured voltage with slight oscillations, whereas the v-form one
150 km synchronizes according to the setpoint of the first VSC.

eg1 [p.u.]
300 km

0.99 V. C ONCLUSION
In this paper, a new classification of VSC control strategies
0.98
depending on their mode of operation was presented. A
0 5 10 15 20 25 concept of partial grid forming is introduced, which indicates
t [s]
that a 100% inverter-based system can be sustained without a
(a)
dedicated grid-forming unit. Unlike the existing conventions,
25 km the concept is based on partial forming of individual sys-
1.1 150 km
tem parameters, distributed across different converter units.
eg2 [p.u.]

300 km
1.05 Subsequently, a detailed model of VSC was developed in
both MATLAB Simulink and vHIL platforms, with a goal
1
of accurately capturing the system dynamics. Finally, using
0 5 10 15 20 25 time-domain simulations, the stability of various configuration
t [s] scenarios was analyzed. The necessary converter and network
(b) criteria for providing a stable system under the proposed
control concepts have been investigated.
Fig. 6: Sensitivity of voltage oscillations to the length of a
transmission line: (a) S23 ; (b) S32 . R EFERENCES
[1] C. K. Sao and P. W. Lehn, “Control and power management of converter
between the f -form and v-form unit must be limited, so that fed microgrids,” IEEE Trans. Power Syst., vol. 23, no. 3, pp. 1088–1098,
the system stability is preserved. However, it is observed that Aug 2008.
[2] F. Katiraei and M. R. Iravani, “Power management strategies for a
the location of the loads connection in the network can have microgrid with multiple distributed generation units,” IEEE Trans. Power
a significant impact on the overall response, since the voltage Syst., vol. 21, no. 4, pp. 1821–1831, Nov 2006.
sensitivity is drastically lower for scenario S23 . This is due [3] G. Diaz, C. Gonzalez-Moran, J. Gomez-Aleixandre, and A. Diez,
“Scheduling of droop coefficients for frequency and voltage regulation
to line dynamics of the shunt admittance, which stabilizes the in isolated microgrids,” IEEE Trans. Power Syst., vol. 25, no. 1, pp.
magnitude of voltage across the load. Since it occurs in the 489–496, Feb 2010.
vicinity of the f -form unit in scenario S23 , the oscillatory [4] C. L. Moreira, F. O. Resende, and J. A. P. Lopes, “Using low voltage
microgrids for service restoration,” IEEE Trans. Power Syst., vol. 22,
characteristic of the converter voltage is drastically reduced, no. 1, pp. 395–403, Feb 2007.
hence diminishing the effect of the transmission line length. [5] T. Green and M. Prodanović, “Control of inverter-based micro-grids,”
Electric Power Systems Research, vol. 77, no. 9, pp. 1204 – 1213, 2007.
D. vHIL Results [6] Q.-C. Zhong and T. Hornik, Control of power inverters in renewable
The presented results are verified through vHIL, by ana- energy and smart grid integration. John Wiley & Sons, 2012, vol. 97.
[7] Q. Jiang, M. Xue, and G. Geng, “Energy management of microgrid
lyzing the system stability for scenario S32 . The response of in grid-connected and stand-alone modes,” IEEE Trans. Power Syst.,
both converters matches the previous analysis, as shown in vol. 28, no. 3, pp. 3380–3389, Aug 2013.
Fig. 7 through balance of the respective voltage magnitudes [8] J. Rocabert, A. Luna, F. Blaabjerg, and P. Rodrguez, “Control of power
converters in ac microgrids,” IEEE Trans. Power Electron., vol. 27,
and active power outputs. The f -form unit tracks the mea- no. 11, pp. 4734–4749, Nov 2012.
[9] G. M. Foglia, L. Frosio, M. F. Iacchetti, and R. Perini, “Control loops
design in a grid supporting mode inverter connected to a microgrid,” in
eg1
1.2
eg2 2015 17th European Conference on Power Electronics and Applications
(EPE’15 ECCE-Europe), Sep 2015.
v [p.u.]

1
[10] U. Markovic, P. Aristidou, and G. Hug, “Virtual Induction Machine
Strategy for Converters in Power Systems with Low Rotational Inertia,”
in IREP Bulk Power System Dynamics & Control Symposium, Aug 2017.
0.8
[11] W. Winter, K. Elkington, G. Bareux, and J. Kostevc, “Pushing the limits:
0 5 10 15 20 25 Europe’s new grid: Innovative tools to combat transmission bottlenecks
t [s] and reduced inertia,” IEEE Power and Energy Magazine, vol. 13, no. 1,
pp. 60–74, Jan 2015.
(a) [12] G. Denis, T. Prevost, P. Panciatici, X. Kestelyn, F. Colas, and X. Guil-
laud, “Review on potential strategies for transmission grid operations
0.35 p1 based on power electronics interfaced voltage sources,” in 2015 IEEE
p2
0.3 Power Energy Society General Meeting, Jul 2015.
p [p.u.]

0.25
[13] M.-S. Debry, G. Denis, T. Prevost, F. Xavier, and A. Menze, “Maximiz-
ing the penetration of inverter-based generation on large transmission
0.2 systems: the migrate project,” in 6th Solar Integration Workshop, 2017.
0.15 [14] J. A. Taylor, S. V. Dhople, and D. S. Callaway, “Power systems without
0 5 10 15 20 25
fuel,” Renewable and Sustainable Energy Reviews, vol. 57, pp. 1322–
t [s] 1336, 2016.
[15] S.-K. Chung, “A phase tracking system for three phase utility interface
(b) inverters,” vol. 15, no. 3, pp. 431–438.
[16] Typhoon hil — hardware in the loop testing software and hardware.
Fig. 7: Transient response of the converters in vHIL platform [Online]. Available: https://www.typhoon-hil.com/
for scenario S32 : (a) voltage magnitude; (b) active power.