2024 Promises and Challenges of Grid Forming Transmission System Operator Manufacture and Academic View Points

Promises and Challenges of Grid Forming:
Transmission System Operator, Manufacturer and

Academic View Points
Carmen Cardozo∗ , Thibault Prevost∗ , Shun-Hsien Huang† , Jingwei Lu‡ , Nilesh Modi‡ ,
Masaya Hishida§ , Xiaoming Li§ , Adil Abdalrahman¶ , Pär Samuelsson¶ , Thierry Van Cutsem∥ ,
Yorgo Laba∗∗ , Yahya Lamrani∗∗ , Frederic Colas∗∗ and Xavier Guillaud∗∗
∗
R&D, Réseau de Transport de Electricité (RTE), 92073 La Defense, France
†
Electric Reliability Council of Texas (ERCOT), Austin, Texas, 76574, USA
‡ Australian Energy Market Operator (AEMO), Brisbane, Qld. 4000 Australia
§ Zenobē Energy, WC2N 6DU London, U.K.
¶ Hitachi Energy - HVDC, Lyviksvägen 3, 771 80, Ludvika, Sweden
∥ independent consultant, Liège, Belgium
∗∗ Univ. Lille, Arts et Metiers Institute of Technology, Centrale Lille, Junia ULR 2697 - L2EP 59000 Lille, France
Corresponding author: Xavier Guillaud (xavier.guillaud@centralelille.fr)
Abstract—With the increasing penetration of power electronic From the Original Equipment Manufacturer (OEM) perspec-
converters in the power system induced by the energy transi- tive, requirements related to desired performance rather than
tion, Grid Forming (GFM) technology emerges as crucial for requirements on the detailed control structure are preferred.
complementing traditional synchronous generators in fulfilling
system needs. All over the world, TSOs have started introducing Consequently, in recent years, various Transmission System
performance-based requirements to define the desired behaviour Operators (TSOs) worldwide have introduced performance-
of GFM units without prescribing specific technical solutions. based requirements to define the desired behaviour of GFM
Based on these specifications, manufacturers design their grid- units without prescribing specific technical solutions, particu-
connected equipment. However, depending on requirements, chal- larly on the control strategy and the associated resource [5]–
lenges may arise in optimizing control strategies without hard-
ware modifications, potentially becoming cost-driving factors. [7]. This task involves projecting system-wide needs into
Intellectual property protection limits information disclosure, requirements for individual assets. Key challenges encompass
restricting the guidance available to TSOs during cost-benefit avoiding unnecessary constraints that could prevent OEM from
assessments. Academic contributions on GFM control and generic delivering cost-effective solutions while ensuring that they
models can bridge the gap, providing a fair portrayal of the ultimately comply with the proposed requirement set and that
general behaviour and then facilitates an open discussion on their
ability to meet the requirements and contribute to fulfil system they can effectively address system-wide needs.
needs. This survey paper provides a comprehensive overview of System developers and OEM design technical solutions
the perspectives offered by these diverse stakeholders. based on detailed specifications and a profound understand-
Index Terms—BESS, Grid forming, HVDC, system needs. ing of equipment constraints. As requirements accumulate,
challenges may arise in optimizing control strategies with-
out resorting to hardware modifications, potentially becoming
I. I NTRODUCTION cost-driving factors for the installation design. The details of
As we progress through the energy transition and the these technical solutions are often subjected to Intellectual
deregulation of the electricity sector, a growing imperative Property (IP) protection, limiting the information these parties
arises to align system needs with available resources capable are prepared to disclose and, consequently, restricting the
of meeting them. Their timely deployment and optimal real- guidance available to TSOs during cost-benefit assessments.
time utilization are crucial to ensure an affordable and reliable To bridge the gap, academic contributions on GFM offer a
power supply over time. In this context, Grid Forming (GFM) consistent framework of study for the various types of grid
can be seen as a technology with the potential to be deployed forming control, starting from basic concepts. This requires
in various resources, including Power Park Modules (PPM), to clarify the model of the system under control, propose a
Battery Energy Storage Systems (BESS) and Flexible AC rigorous procedure leading to various topologies of controls,
Transmission Systems (FACTS), to complement Synchronous some few variants and eventually design the parameters of
Generators (SG) in fulfilling specific system needs [1]–[4]. the controllers. The proposed generic models can facilitate
an open discussion on their ability to meet the requirements
and contribute to fulfil system needs, while shedding lights on
Submitted to the 23rd Power Systems Computation Conference (PSCC 2024). the burden new requirements may impose on specific devices.
23rd Power Systems Computation Conference Paris, France — June 4 – 7, 2024

PSCC 2024
While these models offer valuable insights into the overall Similar lists have been proposed in [1], [3]. Here, we present
system behaviour, more precise vendor models are always an overview of the diverse factors involved in their quantifica-
required for detailed design studies. tion from the perspective of three TSOs, operating systems of
The paper is organised as follows: Section II examines various characteristics, namely the Australian Energy Market
the motivations behind fostering GFM adoption. Subsequently, Operator (AEMO), the Electric Reliability Council of Texas
Section III outlines the resources that can actively contribute to (ERCOT) and Réseau de Transport d’Electricité (RTE).
the satisfaction of the identified system needs and discusses the 1) AC fundamental frequency voltage: TSOs deliver AC
subtleties of GFM technical specifications in different regions voltage to customers. The frequency and depth of perturbations
to meet local needs. Afterwards, Section IV brings forward such as low/high voltage events, as well as the harmonic
the challenges of implementing this solution on High Volt- content, serve as quality KPI. Contractually binding thresholds
age Direct Current (HVDC) transmission systems. Following may apply to prevent malfunctions or tripping of grid users.
that, Section V provides an academic perspective on defining Some of these events could jeopardize system security if they
generic models and testing them in various situations. Finally, lead to large power imbalances. These indicators are often
conclusion are drawn in Section VI. monitored with specific devices at sensitive location [8].
2) Inertia: refers to the power system’s ability to resist
II. F ROM THE TSO S ’ LENS : QUANTIFYING AND changes in frequency following power imbalances. It is often
MONITORING SYSTEM NEEDS
quantified as an energy buffer, usually measured in MWs.
The responsibilities of different TSOs may vary slightly Another classical indicator used in this context is the Rate
from one country to another, depending, for instance, on of Change of the Frequency (RoCoF), expressed in Hz/s.
whether they own the grid infrastructure or not. Nevertheless, Limiting its maximal value, hence maintaining a minimal level
these responsibilities generally include ensuring cost-effective of inertia, is important for various reasons:
security of power supply — an endeavor easier said than • To allow sufficient time for primary frequency control to
done, requiring coordinated actions across a wide range of restore power balance. This need may be reduced by the
time frames. To underscore the need for GFM technology, development of Fast Frequency Response (FFR) services.
this section revisits the prerequisites essential for the proper • To avoid disconnection of RoCoF sensitive units.
functioning of power systems. Section II-A first recalls the • To ensure that the frequency can be accurately measured
comprehensive nature of system needs and elaborates on Key by any device on the grid that relies on this signal.
Performance Indicators (KPI) allowing their quantification. • In particular, to provide time for the execution of defense
Subsequently, Section II-B presents ongoing developments plans, namely Under Frequency Load Shedding (UFLS)
toward the consideration of these KPI in decision-making schemes, in the event of disturbances exceeding the
processes pertaining to both system planning and operation. In system’s Frequency Containment Reserve (FCR).
particular, we focus on the aspects in which GFM technologies The state-wide blackout experienced by South Australia
will play a pivotal role. Challenges related to the transition to (SA) on September 28, 2016, is a well-known example of
variable energy sources and the increasing need for flexibility the risk associated with an inertia shortfall. Approximately
resources to maintain power balance across different time 850,000 SA customers lost electricity supply, including house-
scales are topics beyond the scope of this work, although they holds, businesses, transport, community services, and major
are acknowledged to be significant concerns for TSOs. industries. This event occurred after the SA power system was
A. Defining and quantifying system needs separated from the rest of the Australian National Electricity
Market (NEM) power systems, which was triggered by mul-
In this section, we discuss the system needs related to
tiple factors, including severe weather conditions [9]. After
power system security. Beyond power balance, load sharing,
the SA separation, there was a deficiency in the electricity
and classical ancillary services such as frequency and voltage
supply within the SA island, which caused a rapid frequency
regulation, these needs include:
reduction with a RoCoF at 6 Hz/s, beyond the capability of the
1) AC voltage (maintain fundamental frequency), local UFLS scheme for arresting the frequency in-time, and
2) inertia (providing time for regulation and defense actions an eventual system-wide blackout. Based on extensive power
based on system observation), system simulations conducted by AEMO during subsequent
3) fast fault current (for protection and voltage support event investigations, it was suggested that keeping the RoCoF
during short-circuits), below 3 Hz/s could have reliably preserved the frequency in
4) system strength (in the sense of low ∆V ∆I ), the SA island [10]. However, achieving a lower RoCoF was
5) robustness (maintain performance across a wide range not possible due to the very low level of inertia available
of grid operating conditions), in the SA island on September 28, 2016. Following these
6) stability: findings, recommendations by AEMO were implemented in
• transient stability, collaboration with the local Network Service Providers (NSP)
• small-signal stability (positive damping), in SA, including the design of Special Protection Schemes
7) withstand capabilities (according to risk policies), and (SPS) to manage the risk of SA separation and a RoCoF-based
8) restoration capabilities. UFLS system.

PSCC 2024
3) Fast Fault Current: historically, Synchronous Machines As for quantifying the system strength, another indicator is
(SM) have provided high and fast current in response to grid typically used: the Short-Circuit Ratio (SCR). It normalizes
event. The system has been designed to handle such high the available SCP at a specific location with respect to the
current and has also taken benefit of it to design simple protec- nominal power of a certain facility. Different variants of this
tion schemes. This property facilitates selectivity, allowing for indicator have been defined to account for PEIR, including
quick fault detection and clearance while disconnecting the the Effective SCR (ESCR), the Weighted SCR (WSCR), the
minimum amount of devices. Additionally, it helps limiting Generalized SCR (gSCR) among others [18]–[23]. Similar
voltage drop along the network, thereby improving quality to the arguments presented for the AC voltage need, we
for customers by reducing flicker and limiting the spread of could say that TSOs have been expected to provide a certain
short undervoltage events. Understanding the injected current level of SCP to their customers. Technical specifications and
by Power-Electronic Interfaced Resources (PEIR) is essential connection agreements typically precise minimal and maximal
for deducing fault information from the current waveform. In values, thereby defining the range for which the installation
general, PEIR are required to inject fault current up to their is designed to operate correctly. However, similar to inertia,
rated capacity [11], with specific details about the expected this service has been taken for granted in SM-dominated
dynamic behaviour outlined in different standards, such as systems, and the means for TSOs to marginally adjust it are
the IEEE Std 2800-2022 [12] and the EN 50549-2 [13]. Due often limited to topological changes. Additionally, defining
to the limited current capability of PEIR, priority rules have scenarios for calculating available SCP and compiling relevant
traditionally favoured reactive current during faults to support data representative of the system’s evolution over decades
voltage. However, with their growing share, concerns about has always been challenging [24], leading to the adoption of
potential frequency-related events resulting from substantial different approaches by various TSOs:
reductions in active current are prompting a reassessment of a) AEMO experience: in Australia, there have been a
this practice in certain regions [14]. Moreover, the notion few PEIR driven sub-synchronous oscillation incidents due to
of system strength, detailed below, has been traditionally the lack of system strength at the vicinity of these PEIR. One
correlated with the fault current due to the inherent properties notable incident involved five nearby solar farms, ranging from
of SM. In the Australian NEM for instance, the minimum three 50MW to 90MW, connected to 66 kV distribution systems
phase fault currents are used as a proxy for system strength in a remote area of the Australian NEM power system. The
at the fault level nodes which are selected as per the system loss of a nearby 220kV transmission circuit resulted in a low
strength requirements methodology [15]. In a broader context, system strength condition for this area, and caused the five
as PEIR respond distinctly to system events, particularly solar farms to contribute to a 7 – 8 Hz sub-synchronous voltage
for small and large disturbances, it becomes imperative to oscillation with a magnitude of 0.8% to 1.0% of the nominal
differentiate fault behaviour from system strength. value, observable at the 220kV transmission substations in this
4) System strength: more generally AEMO defines system area [25]. The oscillation was eliminated after upgrading the
strength as the ability of the power system to maintain and con- inverter control systems of the five solar farms with a solution
trol the voltage waveform at any given location in the power developed in collaboration between AEMO and the OEM.
system, both during steady state operation and following a b) ERCOT experience: similarly, the integration of PEIR
disturbance [16]. Although conventionally associated with the has led to the reduction of system strength, resulting in
concept of Short-Circuit Power (SCP) in SM-dominated power undesired oscillatory responses and unexpected unit trips in
systems, it is more related to voltage stiffness, represented by recent years. The Odessa I and II events, occurred in 2021 and
the voltage excursions resulting from injected current varia- 2022, are examples of this phenomenon [26], [27]. In addition,
tions. For this reason, it is often assimilated to an impedance. various localized oscillations among PEIR or interactions with
We refer to strong and weak grid conditions for scenarios with grid elements have also been observed.
high and low SCP, translating to low and high grid connection c) RTE experience: in France, historically, large gener-
impedance, respectively. This characteristic naturally affects ators (starting from 800 MW) are required to operate with a
power quality, but also small signal stability (SSS) as follows: grid-connection impedance of up to 0.6 pu. [28], equivalent
• The association of power quality with system strength is to a SCR of 1.67. This requirement emerged due to the
occasionally emphasized by traditional customers using development of larger Nuclear Power Plants (NPP) in remote
induction machines, which can cause significant voltage locations with radial networks. Conversely, for smaller units
drops during the startup process, potentially leading to (below 50 MW), a minimum SCR of 5 (corresponding to
max
equipment malfunctions of other users in weak grids. Xsc = 0.2 pu.) has been defined. Current efforts are focused
• While the relationship between system strength and SSS on evaluating the necessity of revisiting these requirements.
has gained attention with the massive connection of 5) Robustness: the electrical system undergoes continu-
PEIR [17], the SSS of remote generation based on large ous changes. Consumption and distributed generation patterns
SM has also been a subject of historical scrutiny. These shape residual load profiles while dispatchable generation is
concerns nonetheless faded as power systems became constantly being adjusted based on market and regulation
increasingly interconnected, and the widespread adoption processes. Simultaneously, operators modify the grid topology
of Power System Stabilizers (PSS) took precedence. to manage security constraints. These factors contribute to an

PSCC 2024
ever-increasing volatility in line flows and voltage profiles, in uphold system stability in the long term through performance-
Europe accentuated by the expanding cross-border exchange based grid connection requirements. In between, adjustments
capacity. Moreover, preventive actions may be replaced by cor- of protection settings may provide operators with additional
rective measures to reduce operational costs, placing increased flexibility to adapt to changing conditions [33].
demand on the regulating and protective layers [29]. Conse- a) Transient stability: technical requirements commonly
quently, the system may experience more frequent temporary applied to generation units focus on ensuring stability during
excursions into exceptional operating conditions. Although faults of specified types, depth, durations, clearance and topol-
timely handled by suitable remedial actions, RTE has, for in- ogy under the most challenging operating conditions. The fault
stance, witnessed an increase in the overall number of one-off duration to be considered for a particular installation is dictated
cases of 225 and 400 kV overload, particularly at the French- by the system overall protection scheme, and must be included
Spanish border and for the N-1 scenario [30]. The increasing in the withstand capabilities. In France, for instance, units
voltage and frequency volatility, especially in the transient time connected to the high voltage grid are required to withstand
frame during and right after the disturbance, could trigger the faults lasting 150 ms, while the requirement is reduced to
protection and/or cause the PEIR to temporarily reduce the 85 ms for the extra high network (above 110 kV), and it
support or disconnect from the grid. Therefore, now, more might be extended up to 250 ms in other EU countries [11].
than ever, all grid-connected devices must demonstrate desired In anticipation, the impact of PEIR on the transient stability
performance across a wider range of operating conditions. of the remaining SG is assessed. In this context, accelerating
ERCOT has, for instance, proposed more robust technical real-time evaluations of Critical Clearing Time (CCT) to cope
requirements, essentially to align with IEEE 2800-2022. with the growing number of possible states is an active area of
6) Stability: TSOs ensure large-scale power system sta- research [34], along with the optimisation of remedial actions.
bility by implementing actions across various time frames, b) Small-signal stability (positive damping): RTE has
from planning to real-time operation. In the short term, the traditionally applied stability margin requirements based on
acceptability of system-wide operating points, in accordance frequency domain studies to ensure the proper damping of
with a given risk policy, is often determined through contin- SG natural (local) modes [28], which have often defined PSS
gency analyses supported by time-domain simulations. There- design. In Continental Europe (CE), concerns about interarea
fore, TSOs rely on accurate models and suitable simulation oscillation phenomena have led to the requirement of Power
tools to assess the dynamic behaviour of the power system. Oscillation Damping (POD) functions to PEIR in the latest
Predefined mitigation measures, including topology or even- Connection Network Codes (CNC) update [6]. For HVDC
tually more costly dispatch changes, are implemented when systems, inherent damping and Sub Synchronous Damping
needed based on the monitoring of specific KPI. However, Control (SSDC) requirements were introduced in 2016 in
the integration of distributed PEIR complicates the task of addition to POD specifications [35]. Damping thresholds are
maintaining a relevant system representation and identifying enforced by some TSOs, such as ERCOT, which sets a 3%
effective remedial actions. Moreover, the growing variability damping requirement to all resources. If non-compliance is
and uncertainty in operating conditions make traditional offline identified in studies, adjustments to PEIR control are required,
stability assessments increasingly inadequate. Therefore, the often in collaboration with vendors. In real-time, if oscillations
incorporation of more conservative assumptions becomes nec- occur, PEIR may be directed to curtail or disconnect until
essary to accommodate diverse system conditions, generating the issue is resolved. With faster dynamics and the high
in turn a potential risk of infeasible situations. To cope with proliferation of PEIR across the grid, the SSS phenomenon
these challenges, various TSOs, are in the process of deploying now also extends to higher frequency events and might require
online Dynamic Security Assessment (DSA) tools [29]. On the complementary methods for study and monitoring [36], [37].
Australian side, in accordance with the National Electricity c) Modelling requirements for stability studies: signif-
Rules (NER), AEMO is responsible for managing power icant efforts have been undertaken by TSOs worldwide to
system security on the eastern coast of the NEM power system enhance requirements for Root-Mean Square (RMS) and Elec-
following the AEMO’s Power System Security and Power tromagnetic Transient (EMT) time-domain models. There is
System Stability Guidelines [31], [32]. Practical implemen- a particular focus on their validation processes to ensure
tation involves integrating linear constraints into the market accuracy and robustness under various system strength con-
dispatch engine. They represent static and dynamic limitations, ditions. In France, for instance, new requirements for RMS
covering voltage collapse as well as transient and oscillatory model validation align with the IEC 60400-27 standard [38],
stability considerations. These constraints are derived from while scenarios are consistently defined with respect to EMT
limit advice provided by regional NSP and their effective validation cases [39], [40]. On the Australian side, AEMO
enforcement is verified through simulation. Updates occur as updated in 2018 its Power System Model Guidelines [41],
needed, typically when connecting a new generator. introducing, for the first time, relevant requirements for EMT
Finally, during the planning stage, assessing instability risk models of generators seeking connection in the NEM systems.
proves exceptionally challenging due to the unavailability of These guidelines were further updated in 2023 to include
accurate dynamic models for future assets, and uncertainty inverter-based loads [42]. Moreover, ERCOT perform periodic
surrounds their specific locations in the grid. In practice, TSOs EMT studies for the regions with high penetration of PEIR in

PSCC 2024
a weak grid condition to ensure the RMS tool adequacy [43], This performance includes, but is not limited to, the max-
[44]. Finally, obtaining PEIR linearised models suitable for imal RoCoF following power imbalances (tied to inertia)
SSS studies remains an open question [45], [46]. and the propagation of voltage disturbances (linked to the
Other power system stability phenomena, such as voltage system strength). Concurrently, these characteristics depend
collapse, are excluded from the discussion since their evolution on the properties of the very same grid-connected devices.
is not expected to be significantly influenced by the adoption Complicating this scenario is the recognition that system needs
of GFM solutions. The changes in these relatively slower are inherently local and highly sensitive to various factors,
events are more likely to be dictated by the shift in generation such as the relative size of installations and system operating
location, power flow patterns, and the modification of the conditions — encompassing grid topology, instantaneous load,
dynamic behaviour of the load, along with the deployment generation mix, and, in particular, their geographical location.
of (in)sufficient regulation capabilities. Consequently, these needs, along with the availability of
7) Withstand capabilities: in addition to the stability re- resources to fulfill them, practically become almost as volatile
quirements defined for specific events and discussed in para- as the system itself.
graph II-A6a, grid-connected devices and their associated Moreover, the limitations of physical components and the
auxiliary equipment must ride through a broader range of complex dynamics of the closed-loop system make it difficult
disturbances. They extend beyond short-circuits, to cover to accurately quantify and, even more so, predict system needs.
for potentially more constraining scenarios like evolving CNC addresses this challenge from two perspectives. On one
faults, generator disconnections, and changes in grid topology. hand, they aim to improve immunity of various technologies
These events lead to fast transients in electrical quantities to alleviate constraints on system performance. On the other
such as voltage dips/swells, frequency excursions, and phase hand, they establish a baseline of responses required at the
shifts. Up-to-date requirements in European CNC include at installation level to contribute to this overall system perfor-
least minimal RoCoF and Fault Ride Through (FRT) time- mance. Exhaustively defining these requirements necessitates
dependent profiles. Withstand capability involves stable func- a thorough understanding of the security management process,
tioning and prompt recovery to the previous operational point, which is precisely the focus of the upcoming section.
but it does not necessarily fully specify performance criteria.
B. Sizing and monitoring system needs
Nonetheless, upper bounds on the active power recovery time
of generating units have been enforced to prevent subsequent Over time, TSOs have diligently worked on defining suit-
events, notably a frequency incident. Presently set to 2 seconds able security thresholds for emerging system needs. This, in
for PPM in France, it should soon be reduced to 1 second [6]. principle, would enable the optimal deployment and utilization
For HVDC systems, the required rise time (up to 90% of of the resource pool. Building upon the overview previously
the pre-fault active power) is usually below 500 ms, set to presented, in this section, we initiate a critical examination of
200 ms in France to limit transient stability constraints in the practical implications of optimizing fundamental resources
nearby SG. Analogously, ERCOT initially established FRT for the existence of the power system. We aim to underscore
capability in 2013. The requirement has evolved over the the complexity of extending DSA concepts to new phenomena
years to accommodate the growth and advancement of PEIR. associated to the integration of PEIR. For this purpose, we
Following IEEE 2800-2022, ERCOT has proposed revised take a detour from the service-oriented framework, adopting
FRT requirements for existing and new PEIR to ensure their a power system security management perspective.
ability to remain connected and provide necessary reliability 1) On the security management of power systems: it relies
support, including active and reactive support, during events. on design, preventive and corrective actions to meet specific
8) Restoration capabilities: system restoration has histor- reliability criteria. This task unfolds in at least three key stages:
ically relied on SG. For this purpose, islanding capability is • In planning: system infrastructure design, often founded
usually mandatory for large SG connected to the transmission on prospective scenarios and simplified models.
network [47], [48], and is regularly tested [30]. To some extent, • In anticipation: determination of preventive actions, rely-
this requirement has been extended to a specific category of ing on system state forecast and contingency analysis.
PEIR, with the subtlety that at least a GFM source (a virtual • In real-time: deployment of corrective actions when an
SM), external to the installation, is supposed to be part of the unforeseen event occurs, based on the system observation.
island [11], [28]. For HVDC systems, black start capability is Consequently, at the operational level, the security manage-
often required and tested during commissioning [49]. Although ment depends on our proficiency in:
the participation of PEIR in system restoration is beyond the • For preventive actions, precisely forecasting the system
scope of this work, it is recognized as a relevant research topic state within the time horizon determined by the delay in
for ensuring the resilient operation of future power systems. deploying countermeasures.
In conclusion, the task of safeguarding power system secu- • For corrective actions, accurately observing the system
rity extends beyond the act of balancing power. In this context, state and having levers that can be fully deployed before
system needs find definition in the complex equilibrium be- protective actions are triggered.
tween the diverse immunity of grid-connected devices and the • In both cases, making informed decisions regarding the
overall system performance. necessity, effectiveness, and optimality of these actions.

PSCC 2024
These tasks become particularly challenging in systems c) General considerations on screening indicators: as-
facing increasing variability and uncertainty stemming from sessing stability in systems with high share of PEIR neces-
both the energy transition and the deregulation of the electric- sitates substantial data and computing power. When these
ity sector [29]. Here, we focus only on the decision-making resources are limited, high-fidelity time-domain simulations
part, which embeds different dimensions. In particular, we are may be replaced by simplified ones, to obtain (supposedly
interested in three of them, which will be illustrated with the conservative) screening indicators, also referred to as proxies.
congestion management example: They have proven useful both in the planning phase with
• Choosing the KPI under scrutiny (e.g. line current). a broad range of scenarios and in real-time operation to
• Defining the security threshold(s), which determine the quickly deploy remedial actions as required. In the following
need for action (e.g. maximum allowable values). paragraphs, we discuss the use of different inertia and system
• Establishing the reliability criteria (or risk policy) that strength KPI in the TSOs’ decision-making process.
guides the aforementioned decisions (e.g. the N-1 rule). 3) Inertia screening indicators: the risk of inertia shortfall
The primary goal is to ensure that the KPI consistently is under investigation on a years-ahead basis in several systems
remains below the security threshold in the event of any single in Europe, US and Australia. Furthermore, the implementation
outage. One could assert that during the planning stage, the of real-time monitoring has been explored in select systems,
feasibility space is expanded, allowing for optimization as we including the Nordic region [51] and the UK [52].
approach real-time. In practice, this is the result of a trade-off a) AEMO experience: on 19 September 2017, the Aus-
between investments and operational costs. It is nonetheless tralian Energy Market Committee (AEMC) published a new
worth highlighting that if the phenomenon of interest is neither rule [53] which requires AEMO to develop the inertia require-
predictable nor observable, or if available remedial actions ments methodology [54], and calculate the minimum level of
cannot be timely deployed (controllable), the concept of its inertia and the secure level of inertia for each inertia sub-
security management in operation loses its meaning. networks based on the methodology, on an annual basis. An
Despite advancements in the DSA field, security manage- inertia shortfall forecast for at least a five-year window must
ment strategies and tools may still be distant from accounting be included, with the latest 2023 inertia report published in
for dynamic phenomena beyond classical stability (frequency, December 2023 [55]. Once a shortfall is declared by AEMO,
voltage and transient stability), especially for considerably the inertia service providers are responsible for providing the
large interconnected systems. In this scenario, the fallback requirement levels of inertia services to fill the shortfall.
solution remains the incorporation of appropriate margins b) ERCOT experience: along with the integration of
during the design phase. This aims to prevent the security PEIR, the system inertia, focusing on the Synchronous Inertia
criteria associated with the relevant KPI from becoming the (SI) provided by the online SM, has been closely monitored.
binding constraint in real-time operation. In addition, critical inertia level needs to be maintained to
2) Assessing new security risks: specifically those arising manage the risk of UFLS activation under potential credible
from the transition from SG to PEIR. This refers back to the generation outages and prevent potential blackout. Considering
system needs listed in section II-A, extending the discussion that ERCOT peak load is 85 GW and the loss of 2800 MW
to their planning, prediction, and observation. must not trigger UFLS (first stage at 59.3 Hz), a threshold of
a) Reliability criteria: similar to how spinning reserves 100 GW-s of SI has been identified through detailed system
address the potential outage of generation and/or load to ensure dynamic analysis. In order to maintain sufficient system SI,
power system security, this principle should be extended to real-time and near term future inertia levels are displayed
all resources needed to maintain minimal system-wide per- and operator procedures to commit the additional resources
formance at all times. Moreover, in addition to covering for as needed have been implemented.
the loss of predefined assets, other factors such as structural c) European Network of Transmission System Operators
modifications in the grid topology, changes in the generation for Electricity (ENTSO-E) experience: CE has experienced a
mix, or even the loss of the last synchronous link between two few system splits in the last decade resulting in load shed-
areas may prove crucial depending on the specific phenomenon ding [56], [57]. A dedicated working group conducted a study
under consideration. Furthermore, identifying pertinent Worst on the CE system [58], characterizing the RoCoF of different
Case Scenarios (WCS) involves not only defining critical areas in the case of various system split scenarios within
contingencies but also plausible initial states. Europe. The combined effects of reduced inertia, associated
b) Time frame: system needs must be met by resources to the increasing number of installed PEIR, and increased
across every time horizon. For instance, to ensure power power exchanges lead to RoCoF values exceeding 1 Hz/s
balance close to real-time, operational considerations include in both resulting areas for many scenarios and reaching up
maintaining margins for various anticipation periods, depend- to 4 Hz/s. This situation poses a risk of defense mechanism
ing on the activation time of associated flexibility levers [50]. misbehaviour, potentially leading to a Europe-wide blackout.
Looking years ahead, this goal translates into ensuring that Even lower RoCoF values may cause localized load shedding
enough generation capacity is built; a task accomplished misoperation with reduced impact. The new version of the
through load forecasting and adequacy studies. Scenario and European CNC will ensure that PEIR will be able to ride
risk-based approaches allow us to cope with uncertainty. through such events [6].

PSCC 2024
The potential need for specific operational procedures may
be explored in the coming years, while thoughtfully consider-
ing the nature of security constraints before extending service
provision mechanisms developed for the balancing problem.
4) System strength screening indicators: in spite of the
challenges associated with the modelling of PEIR in SCP
calculation tools [59], different indicators have been employed
worldwide for both planning [21], [60] and operational studies.
a) AEMO experience: there are currently two categories
of regulations that are used to ensure sufficient system strength
is maintained in the NEM:
• System strength requirements: AEMO calculates mini-
mum three phase fault currents at representative locations,
called the fault level nodes, across the NEM using the
System Strength Requirements Methodology [15]. They
are reviewed and updated annually, and published in
AEMO’s annual system strength report (latest report
at [61]), along with projected system strength shortfalls. Fig. 1: Example of ESCR results on a French 2030 scenario
Fault level shortfalls are addressed by Transmission NSP.
• System strength impact assessment: generators seeking
connection to the NEM power systems must undergo the validity of screening indicators are crucial for planning and
assessment as per the System Strength Impact Assess- real-time decision-making processes. Detailed time-domain
ment Guidelines [62], to ensure they do not jeopardize the simulations will remain necessary for assessing remedial ac-
stable operation of the NEM under low system strength tions and conducting project-specific grid connection studies.
conditions. Otherwise, they must provide necessary mit- In conclusion, the evolution of system-wide security indi-
igation measures as determined by the assessment. cators has necessitated the establishment of minimal require-
In both cases, positive sequence fault current calculations ments — security limits — to ensure ongoing compliance with
are used to determine the required fault levels, and detailed desired reliability criteria. To address system complexity and
assessment are conducted with large scale EMT simulations. data unavailability, screening indicators have been adopted to
b) ERCOT experience: while a minimum system distinguish acceptable situations from those requiring remedial
strength criterion has not yet been established, various pro- actions, both in the planning and operational time horizons.
cesses and requirements are in place to secure the reliable Presently, these indicators primarily include inertia, as well as
connection and operation of PEIR. For instance, ERCOT system strength and/or fault current. Finally, the deployment
employs the WSCR in real-time to manage wind generation in of the resource pool capable of fulfilling system needs must
the Panhandle area, imposing limits on the maximum output be carefully and robustly planned years in advance to ensure
that wind generation can export to ensure stable operation. that enough capacity is always available. In the following,
c) RTE experience: ESCR from [18] has been evaluated we examine how specific resources can (or cannot) provide
for 2030 French high PEIR scenario as illustrated in Fig. 1. services to satisfy the outlined system needs.
Red and orange dots are nodes considered at risk, indicating
III. F ULFILLING SYSTEM NEEDS
areas where more detailed studies might be necessary. A
concentration of very low ESCR is observed at the 90/63 Most of the services previously discussed, including AC
kV level, primarily due to the consideration of renewable fundamental frequency voltage, system strength, inertia, and
connection at lower voltages. high and fast fault current, have been historically and inher-
SCR-based KPI show promise for long-term cost-benefit ently provided by SG as a by-product of electricity generation.
analyses and real-time system strength monitoring. However,
there is no a singular and universally accepted definition A. Limitations of conventional Grid Following (GFL) PEIR
of system strength and establishing security thresholds for As PEIR displace the conventional SG fleet, the availability
selected indicators in PEIR-dominated grids requires careful of these resource decreases, posing challenges for TSOs to
consideration of the varying system conditions. On one hand, maintain system stability under certain scenarios. This is re-
concerns persist about the conservatism of these KPI, which lated to the dynamic response of conventional GFL converters:
could lead to unnecessary costs as weak grid situations may 1) Inertia: it is recalled that, the virtual inertia, typically
not always pose stability risks. On the other hand, limited provided by GFL converters, can be categorized as a fast
anticipation may result in resource scarcity and potentially ex- frequency control. After a sudden event, the change of active
pensive mitigation measures (the cost of inaction). Therefore, power needs to be immediately handled by some unit with
refining modelling hypotheses and gaining confidence in the true inertia (grey area in Fig. 2 from [63]).

PSCC 2024
Under these circumstances, other grid users, such as PPM
and certain loads, or service providers, through tenders [67],
may contribute to providing these critical services. When they
do, they will be referred to as GFM resources.
Fig. 2: Time frame difference for FFR and inertia B. Introducing GFM resources in the grid
Recent global development indicates that the GFM Inverters
2) System strength: analogously, GFL converters might (GFMI) have emerged as a key technology which could assist
be capable of providing fast voltage support but still rely TSOs in tackling challenges in the energy transition, including
on system observation. Transiently, their so-called current the reduction of SI and system strength. In particular, BESSs
source behaviour or, more precisely, the potential interactions have been identified as the low-hanging fruits of the GFM
between the current loop and the Phase Locked loop (PLL) capability tree [68] with Great Britain (GB) and Australia
prevent them from contributing to the system strength and to leading the harvest [69]. Finland may follow after requiring
the stability of nearby sensitive devices. GFM capability for all BESS connected in areas with a high
3) Fast Fault Current Injection (FFCI): when provided by penetration of PEIR [70]. Additionally, TSOs can resort to
GFL converters, this function relies on voltage drop mea- the installation of dedicated devices to meet stability needs,
surement to achieve a desired response IQ = f (V ), which namely Synchronous Condensers (SC) and specific FACTS
introduces a delay in the reaction including measurement time with GFM capability, including HVDC systems.
and control dynamics. Moreover, the injected current does 1) AEMO experience: NSP may choose to procure or
not adapt to the grid impedance seen from the device. As a contract SC as means to ensure the required fault levels. In
consequence, the impact of PEIR on the Short-Circuit Current addition, generators identified with adverse system strength
(SCC) shape is notable. While SM exhibit an exponential impact may choose to remediate by installing SC. Moreover,
decay in current, GFL converters with IQ control showcase a Voluntary Specification for GFMI [7], has been developed
a fast ramp followed by steady current. To the best of the by AEMO in collaboration with industry partners and market
authors’ knowledge, there is no current record of protection participants, where the provision of inertia is expected to be
mis-operation resulting from the reduced fault current in- a core capability. Furthermore, AEMO, in collaboration with
feed associated with the integration of PEIR. This absence of a GFMI OEM, published a study demonstrating that GFMI
issues is likely attributed to the implementation of appropriate can effectively suppress a real sub-synchronous oscillation
precautionary measures, indicating that current procedures and phenomenon observed in a remote area of the system [71],
available technical solutions have proven reasonably effective thereby presenting an alternative to the solution documented
so far. However, it’s important to note that the concern in [25]. Based on these findings and similar analyses from
′′
persists [64]–[66] and that traditional indicators like Ik and Australia [72], the provision of positive damping for oscilla-
Ib may no longer be relevant, as illustrated in Fig. 3. tions in power systems that are sensitive to system strength
has been also deemed as a core capability for GFMI [7].
2) ERCOT experience: periodic EMT studies for the re-
gions with high penetration of PEIR in a weak grid condition
are performed to assess the need for grid enhancement such as
new circuit or SC, to mitigate the undesired and unexpected
instability and generation outages. To maintain reliable PEIR
operations under low system strength conditions, two SC (175
MVA each) were implemented in the Panhandle region in
2018. In response to the substantial growth of PEIR (≈40 GW)
in west Texas, approval was granted in 2023 for additional
SC at six locations, providing a total capacity of 2100 MVA.
Ongoing studies are focusing on identifying the most viable
Fig. 3: SCC characteristics: Typical SM (blue), SM + Moder- and effective enhancements for the ERCOT grid, including
ate GFL (purple), SM + High GFL (green) the integration of GFMI. An ERCOT-wide assessment has
highlighted several benefits of GFM BESS on the grid [73].
4) Stability, robustness and withstand capability: maintain- Consequently, GFM performance requirements are currently
ing a stable system at all times requires controls that are under development and are expected to be released by 2024.
robust across a wide range of system conditions, which include 3) RTE experience: RTE has actively participated in mul-
coping with various system strengths. While GFL converters, tiple research initiatives on GFM over the last decade. After
equipped with a PLL-based synchronisation mechanisms, can the MIGRATE H2020 project, RTE coordinated OSMOSE.
operate effectively in various system strength conditions with Within the WP3, two 1-MVA-sized GFM BESS demonstrators
proper tuning, they may be prone to instability under specific were deployed in collaboration with an OEM and the École
tuning and varying system strength conditions. Polytechnique Fédérale de Lausanne (EPFL) [63]. Factory

PSCC 2024
Acceptance Tests (FAT) based on a Power Hardware-In-The- 1) Project description: the site is directly connected to the
Loop (PHIL) test bench have provided valuable insights into 400 kV transmission system and consists of 102 GFMI and
the ongoing development of GFM requirements. Four types MV Power Stations, 445 modular energy storage solutions,
of GFM capability, each with increasing requirements, were and step up to 400 kV through the HV transformers. Each
defined to account for the physical constraints of different re- inverter specifies its GFM capability, and the power plant
sources, including overcurrent capability and available energy controller aggregates each response. The site is designed to
buffer. Currently, RTE is involved in the AGISTIN project to maximize the SCL and inertia provisions by utilizing the short-
demonstrate the provision of GFM capability by loads [74]. time current boost capability of various elements, including
In general, the variety of control solutions found in the inverter and battery modules. The site has achieved 2.1 pu of
literature complicates quantifying the effective contribution of SCL (corresponding to 420 MVA) and 0.3 pu of inertia power
GFM PEIR to system stability. As discussed in Section V, (60 MW, base 200 MW, equivalent to 1,500 MVAs) for all
various GFM solutions may offer similar functionalities and operational conditions, as demonstrated by EMT simulations.
performances. Therefore, within a predefined validity domain, 2) Contribution to inertia: Fig. 4 shows EMT simulation
GFM contribution to stability can be evaluated using simplified demonstrating the system inertial response which is consis-
generic models. This is particularly relevant for TSOs during tently provided to different RoCoF events under all operational
the planning stage when project-specific models are unavail- conditions (full discharge, full charge, idling). Inertia constant
able. For design and interaction studies, especially during and damping factor are adjusted accordingly.
compliance verification for grid connection, detailed EMT
vendor models are necessary. Finally, FAT based on Hardware-
In-The-Loop (HIL) platforms are also valuable tools.
C. An example of GFM capability tender: the GB case
In GB, National Grid (NG) Electricity System Operator
(ESO) initiated in 2019 a competitive tender to procure long-
term stability services from service providers, spanning from
the stability pathfinder phase 1 (SP1) to phase 3 (SP3).
Through this process, NG ESO has secured: Fig. 4: Frequency drop/rise of 0.8 Hz at 1.0Hz/s
• 12.5 GVAs of inertia in SP1 for 5-year contracts,
3) Contribution to SCL: EMT simulations demonstrated
• 11.55 GVA of Short-Circuit Level (SCL) and 6.75 GVAs
more than 2.0 pu fault current at the required PCC retained
of inertia in SP2 for 10-year contract in Scotland, and
voltage, and enhanced fault current characteristics for remote
• 8.7 GVA of SCL and 17.08 GVAs of inertia in SP3 for
faults due to the prolonged current injection. While a SC
10-year contract in the England region.
exhibits higher peak fault current, it may also experience a
The inertia, in MVAs, is defined by multiplying the Inertia faster decay. Given that the SCL contribution is assessed at
constant H as per Eq. (1) and the solution rating Sr , where 100 ms after the fault, the response offered by the GFMI can
∆P is the designed active power change during a frequency be made comparable to the SC for a moderated rating increase
event and f0 is the pre-fault system frequency. as shown in Fig. 5.
∆P f0
H= (1)
2Sr RoCoF
In order to measure the inertia, Active RoCoF Response Power
is used for a RoCoF of less than 1 Hz/s, assessed under
multiple operational conditions prior to a disturbance. The
SCL is measured at 100 ms of FFCI after the transmission
fault and assessed based on the multiple fault impedance and Fig. 5: 3ph-gnd fault with PCC retained voltage of 0.26 pu.
effectiveness between the Point of Common Coupling (PCC)
and target nodes in the network. GFMI-based technologies The Blackhillock project is 200 MW BESS with GFMI
were invited to bid for the first time in the SP2 and two project, designed by less SCL and Inertia capability, targeting
developers have secured contracts using GFM BESS solutions. to achieve Commercial Operation Date (COD) in 3Q 2024.
Zenobē Energy is currently developing three GFM BESS 4) FAT: towards COD, real power model-based tests with
projects in Blackhillock, Kilmarnock South, and Eccles. These individual GFMI and HIL tests on multiple GFMI have
sites provide the stability services addition to active and reac- validated the provision of GFMI capabilities including:
tive power services, as one of multiple revenue streams [75]. • Active RoCoF Response Power stacking with other active
In particular, the Kilmarnock South project is a 300 MW/600 power services while charging/discharging.
MWh BESS with a GFMI solution, originally designed by • Active Phase Jump Power Response to phase jumps of 5
200MW plant and demonstrated high stability capabilities in degrees and 60 degrees injected at GFMI controller and
the following. PCC in the HIL simulator without trip.

PSCC 2024
• Fault Ride Through/Fast Fault Current Injection (FRT, Additionally, it is designed to contribute to future devel-
FFCI): Demonstrate the contracted SCL during the fault opments concerning technical prerequisites, standards, service
conditions, and the active phase jump power, FRT and specifications, and procurement procedures. This initiative was
FFCI capabilities. mandated by the AEMO’s Engineering Framework Priority
• POD: Respond to the voltage and frequency oscillations Actions publication [83] after the need was identified in [84],
with the oscillation frequency of 5 Hz, 7 Hz and 9 Hz. and is the result of a collaborative effort.
AEMO’ Voluntary Specification defines GFMI as the ca-
D. Specifying GFM capability
pability to maintain a constant internal voltage phasor in
Different sets of technical requirements specifying the ex- a short time frame, with the magnitude and frequency set
pected behaviour of GFM resources have been proposed by locally by the inverter, thereby allowing immediate response
various TSOs. NG ESO was the first to formalize them in to a change in the external grid. Then, this document divides
the grid code framework as a non-mandatory requirement and technical capabilities into two categories: Core Capabilities
has recently shared lessons learned from this exercise [5] and Additional Capabilities, as follows:
and implemented an update. Similarly, AEMO published a
Voluntary Specification for GFMI in 2023 [7]. These efforts • Core Capabilities are fundamental requirements for
have more recently been complemented by implementation GFMI, ensuring basic operational effectiveness and grid
guidelines with more details about compliance verification, stability. These include essential functions such as voltage
including testing benchmarks [76], [77]. source behaviour and frequency domain response.
In CE, a first standard was released in Germany [78], • Additional Capabilities enhance the PEIR performance
and an update of the Requirements for Generators (RfG) and adaptability in various grid conditions and needs.
regulation has been submitted by Agency for the Cooperation These advanced features focus on aspects such as im-
of Energy Regulators (ACER) to the European Commission proved interoperability and control functionalities.
(EC), including mandatory GFM capability for PPM [6]. Other The primary distinction lies in the level of necessity: Core
relevant contributions worth mentioning include [79], [80]. Capabilities are crucial for fundamental operation, while Ad-
Finally, a dedicated ENTSO-E working group is addressing ditional Capabilities provide extended functionalities for more
the particular case of HVDC-connected PPM [81], and the complex grid scenarios. Moreover, achieving core capabilities
InterOPERA project has extended the analyses to Multi- is expected with minimal to no modification to plant hardware
terminal Multi-vendor DC grids [82]. In this section, we recall and operational processes compared to a GFL design, whereas
the main features of three of these examples while highlighting additional capabilities may involve hardware upgrades. How-
their similarities and specificities. ever, GFM devices are expected to inherently possess some
1) GB GFM code (GC013): in 2022 NG ESO implemented form of small energy buffer to achieve core capabilities, even
the Minimum Specifications Required for Provision of GB if this energy buffer is not always available.
GFM capability including: AEMO’s voluntary specification also outlines the core ca-
• Active RoCoF Response Power: the transient injection or pabilities that PEIR must possess to qualify as GFMI and,
absorption of active power to the total system as a result whenever feasible, presents anticipated performance descrip-
of the RoCoF value at the PCC. tions for such inverters, in a qualitative manner. They include:
• Active Phase Jump Power: the transient injection or • Voltage Source Behaviour: GFMI should behave like
absorption of active power to the total system as a result a voltage source behind an impedance, maintaining a
of changes in the phase angle between the internal voltage near-constant voltage magnitude and phase angle during
source and the PCC. transients (instant power response during disturbances).
• Active Damping Power: naturally injected or absorbed • Frequency Domain Response: GFMI should exhibit low
to reduce active power oscillations in the total system. impedance magnitude around the fundamental power
Damped response of a GFM plant to an oscillation system frequency, reflecting voltage source behaviour.
between the internal voltage source and the PCC. • Inertial Response: GFMI should provide a synthetic iner-
• Voltage Jump Reactive Power: the transient reactive tial response, offering immediate active power response
power injected or absorbed from the plant to the total to grid disturbances.
system as a result of either a step or ramp change in the • Surviving Loss of Synchronous Connection: GFMI must
difference between the voltage magnitude and/or phase operate stably in a grid lacking other GFMI or SM,
of the internal voltage source and the PCC. adapting to changes like system splits or loss of SG.
• FFCI: reactive current injection that starts to be delivered • Weak Grid Operation and System Strength Support:
into the system in less than 5 ms when the voltage falls GFMI should operate stably in weak grid areas, improv-
below 90% of the nominal value at PCC. ing nearby GFL PEIR stability.
2) AEMO’ Voluntary Specification for GFMI: developed to • Oscillation Damping: GFMI should adequately damp
provide initial guidance to stakeholders, including developers active and reactive power outputs following disturbances
and OEMs, for the elaboration of GFM offerings considering and contribute positively to damping oscillations in the
the evolving regulatory landscape of GFM technology. sub-synchronous frequency range.

PSCC 2024
AEMO’s voluntary specification illustrates the expected 4) Discussion on different requirement sets: a consistent
response of GFMI under various conditions, emphasizing trend in the different definitions of the GFM capability is
stability and responsiveness associated with core capabilities. the requirement of a Voltage Source Behaviour characterized
The deliberate omission of detailed quantitative requirements by the capability to maintain a constant internal voltage (in
aimed for flexibility to accommodate the evolving indus- amplitude, phase, and frequency) behind a (constant) internal
try expectations and system needs. More recently, AEMO impedance during transients [86]. Whether explicitly stated or
collaborated with industry stakeholders and consultants to not, the internal voltage phasor (and/or impedance) of GFMI
develop a simulation-based test framework to demonstrate is allowed to change on a longer time scale to achieve desired
GFMI capability as per the voluntary specification [77]. performances, typically synchronization, positive damping,
3) ENTSO-E CNC: the proposed RfG draft introduces non- power reference tracking and limitation due to physical con-
exhaustive GFM requirements as follows: within its current straints. Additional requirements on the output active and
and energy limits, the PPM shall be capable of behaving at reactive power and/or current in response to variations in
the terminals of the individual unit(s) as a voltage source voltage amplitude, phase, frequency, or grid-side impedance
behind an internal impedance (Thevenin Source), during nor- at PCC are ways to quantify this behaviour.
mal operating conditions and upon inception of a network To ensure clarity, following the lessons learned from NG, it
disturbance [6]. The system response is further specified for is recommended to explicitly delineate detailed performance
network disturbances, as long as the installation limits are not requirements for scenarios in which resource technical limits
reached: the instantaneous AC voltage characteristics of the are reached and those in which they are not. While current lim-
internal Thevenin source shall be capable of not changing its its are well understood by multiple stakeholders, a challenge
amplitude and voltage phase angle while positive sequence remains in the unambiguous definition of inherent energy
voltage phase angle steps or voltage magnitude steps are limits and how they may evolve with operating conditions,
occurring at the connection point. The current exchanged be- a characteristic closely tied to the underlying technology.
tween the PPM and the network shall flow naturally according Moreover, GFM Core requirements, applied within current
to the main generating plant and converter impedances and and inherent energy limits, have demonstrated no signifi-
the voltage difference between the internal Thevenin source cant impact on installation design. However, the addition of
and the voltage at the connection point. advanced features, such as high FFCI, harmonic filtering,
As for AEMO, detailed dynamic performances are excluded. or black start capability, essential for secure operation in
They will be defined at the national level by relevant System specific conditions, may involve oversizing power elements or
Operators (SO), helped by non binding recommendations from changes in operational practices. Therefore, their deployment
Implementation Guidance Document (IGD) to be issued next necessitates detailed cost-benefit analyses. Pilot projects for
year (2025). Finally, ongoing efforts at the normalization level both use cases exist.
should facilitate compliance verification in the years to come, Importantly, the term GFM includes both generating and
specially for PPM connected to the Low Voltage (LV) and non-generating devices, irrespective of whether they are con-
Medium Voltage (MV) distribution networks [78], [85]. nected to the transmission or distribution networks. Further
It is nonetheless clearly stated that the installation shall be work may be required to assess the impact of GFM behaviour
capable of stable operation when reaching the PPM current on Distribution System Operators (DSOs) operational prac-
limits, without interruption, in a continuous manner and tices, a topic beyond the scope of this paper. Finally, it must be
returning to the behaviour described before as soon as the noted that GFM capability is an addition to existing resource-
limitations are no longer active. If reaching the current limit, specific technical requirements. In some cases, updates may
the GFM behaviour must be maintained for responses as be needed to ensure compatibility among them.
specified before for disturbances that require the current to 5) Brief consideration on system restoration: GFM does
vary in the opposite direction of the active limitation. not imply black start capability. The latter requires the unit to
Moreover, for type A units (in France below 1 MW) the self-energize (automatic startup of the auxiliary loads), initiate
requirement is optional and is defined up to the PPM inherent a blacked-out grid, and sustain its operation for a specific dura-
energy storage, which means an energy reserve available in tion. It also necessitates storage or an available primary source,
physical components of a PPM, which has not necessarily along with several additional control layers. However, GFM
been designed to suit the grid forming requirements, but may does entail islanding capability, which facilitates restoration.
be used for such purposes, without affecting the design of the Units with GFM could remain energized and ready to take on
physical components of individual units. From type B units load, contributing to voltage support, while only dispatchable
(Pmax >1 MW) the requirement will become mandatory and sources are expected to participate in the balancing task.
include an inertia consideration as contributing to limiting In conclusion, addressing the urgent need for GFM devices
the transient frequency deviation for over frequency events, requires a combination of measures. This includes updating
extended to low frequency only for electricity storage modules, connection codes to require technical capabilities from grid
but also for Type C and D PPM (Pmax >18 MW in France). In users, developing new services when applicable, and pro-
those cases, the relevant SO is allowed to request the provision gramming investments in TSOs-owned assets when necessary.
of additional energy beyond the inherent storage. Considering the time required for grid code updates to take

PSCC 2024
effect, the latter two solutions have been more immediately
available. Notably, regions such as Texas, Australia, Italy, and
Finland have opted for SC to ensure power system security.
Meanwhile, solutions based on GFM BESS have proven to
be more cost-effective in specific scenarios [69], effectively
providing SCL and Inertia by optimizing the plant design to Fig. 6: Simulation setup.
the system needs. In addition, by stacking multiple revenue
streams on top of stability services, BESS may provide greater
overall benefits to both SO and consumers by being able to the internal voltage angle during disturbances. Compared
provide all these services in parallel at lower cost. to the synchronous machine with a physical inertia, some
Lastly, Germany has embraced the installation of a new fundamental differences exist. For instance, a large inertia
family of FACTS, capable of providing GFM capability, such may be desirable from a frequency support point of view
as the E-STATCOM (where the E stands for Energy) [87], but may severely contradict other requirements posed on
as well as a new generation of HVDC interconnectors. In the HVDC. The larger the emulated inertia, the larger the
the following, Section IV delves into the challenges and required energy requirement for grid frequency support. For
opportunities of deploying GFM capability in HVDC systems. the synchronous generator, the inertia stems from a large
spinning energy. In a VSC HVDC, no such energy is available
IV. I MPACT OF GFM REQUIREMENTS ON HVDC C ONTROL
since the energy stored on the DC side is very limited and
While providing a great increased flexibility in control, must be available for AC voltage generation, [89]. Thus,
GFM via HVDC also introduces a new feature in the power the remote station AC network must provide the energy for
system. It is in this context important to recall that the HVDC grid support. As a consequence, the remote AC grid must in
is a transmission system and not a generating unit nor a this case be able to handle transient power variations. Even
storage. The sections below will discuss the physical and op- so, the energy transmission capability of the link itself is
erational constraints of the HVDC and when it will inherently limited. In principle, a separate external energy storage could
give a different behavior than a synchronous machine. If the also be used for grid support. However, such storage will
constraints of the HVDC are not reflected in the requirement increase the design cost and is at the time of publication
specifications, it may be impossible to fulfill the requirements. not commercially available. If the energy can be provided
Also, there is a risk for contradictory requirements that need to from the other side or via a separate storage, it is possible to
be considered, see [88]. This section aims to show how some provide a large inertia at least in the active power controlling
of the system requirements discussed in Section III-D will station. The larger the inertia, the larger the energy transfer
affect the HVDC, and also to point out the need for modeling on the link during large disturbances will be. A case where
the DC side of the HVDC in sufficient detail. The following active power will be largely affected directly by inertia is
items are discussed and illustrated with simulations: RoCoF. An example is given in Figure 7 where the grid
• Impact of virtual inertia frequency in the active power controlling station 1 operating
• DC side and GFM control as inverter is ramped down from 50 to 49 Hz with the rate
• Transient current capability 2.5 Hz/s. The upper plot in Figure 7 shows the response in
• Virtual impedance impact on large power disturbances active power in the active power controlling station 1 for
• Synchronization different inertia constants H. Also, a frequency controller
In the simulations, a generic controller was used for balancing with dead-band was employed. It can be seen that due to the
the arm voltages. This means that although realistic, the perfor- slower change of voltage angle following with high inertia,
mance will differ from that of a fully optimized vendor model. the active power response in station 1 is larger with higher
However, the level of detail in the model is the same, and the inertia and also that the overshoot is larger with higher H.
same general tendencies can be observed. The simulations are As also further discussed below, the larger the inertia, the
performed using a symmetric monopole point to point HVDC larger the impact on the DC voltage will be due to this. The
setup with power directions according to Figure 6. Station 1 lower plot in Figure 7 shows the corresponding response in
is inverter in active power control (grid forming) while station active power in the DC voltage controlling station 2 which is
2 is rectifier in DC voltage control mode in all simulations in GFL control. From Figure 7 it is also clear that in order
except for in an offshore case. In all simulations, negative to be able to provide the frequency support, some headroom
sign of active power means rectifier operation and positive sign is needed in the design. The larger the inertia in station 1,
means inverter operation. In the figure, each cell represents a the larger the corresponding active power change response
half-bridge submodule. by the DC voltage controller in station 2 will be in order to
balance the DC voltage on the link. This emphasizes the need
A. Impact of virtual inertia for a sufficiently fast DC voltage control when the inertia is
To resemble the behavior of the synchronous machine, it high in the other station. The AC network connected to DC
is common that GFM methods implement a virtual inertia voltage controlling station must also be able to accommodate
via, for instance the swing equations. This will slow down this rather large power variation.

PSCC 2024
response in station 2 will need to be to counteract that. To
be able to handle this, as also further discussed in section
IV-B, the DC voltage control in the remote station must then
be sufficiently fast to respond not to compromise DC voltage
stability. Thus, an inertia requirement in the active power
controlling station will also stress the DC voltage control in
the other station.
0.6
0.5
0.4
H=0.1s
P1 (pu)
0.3
H=0.5s
H=1.0s
0.2
H=2.5s
H=5.0s
0.1
-0.1
1.9 2 2.1 2.2 2.3 2.4 2.5 2.6 2.7 2.8
1.08
1.06
Udc1(pu) 1.04
1.02
1
Fig. 7: Response in active power to a frequency ramp 2.5 0.98
Hz/s for different inertia in active power controlling station 1.
0.96
Upper plot shows the active power in station 1. Lower plot
0.94
shows active power in station 2 1.9 2 2.1 2.2 2.3 2.4 2.5 2.6 2.7 2.8 2.9
Time (s)
An example of a severe power disturbance is a large voltage Fig. 8a: Station 1, upper plot active power P1 , lower plot DC
phase jump in the grid. Assuming that the grid and internal voltage, station 1 in active power control
converter voltages are 1 pu, the instantaneous change in active
power caused by a phase jump is given by (2). Here, ∆Pδ is Since a virtual inertia slows down the voltage angle change
the change in active power, Xconv is the converter impedance of the converter, it will also slow down the active power
(including converter transformer), SCR is the short-circuit response time resulting in slow step responses in active power.
current ratio. It can be seen that the instantaneous change This is illustrated in Figure 9. For the same reason, slow
in power is proportional to the angular voltage change, thus recovery time after faults in the AC networks in the active
inertia has a negligible impact on the magnitude of power power controlling station results. The impact on fault ride
change. The level of inertia will however be reflected in the through behavior and some fast-acting control will increase
recovery phase after the disturbance. with higher inertia.
1 B. DC side and GFM control
∆Pδ ≈ 1 sin(∆δ) (2)
Xconv + SCR The implementation of GFM control as a voltage source
This is illustrated in Figure 8a-8b. Here, a positive phase behind impedance is also possible in the DC voltage
jump of 10 degrees is applied for five different virtual inertia controlling station. This can be beneficial for weak grids as
values in station 1 which is in active power control. Figure 8a shown in [90]. It should be noted that ”inertia” for dc voltage
shows the response of station 1. The corresponding response of control in this case does not resemble a true inertia as in
Station 2 which is in DC voltage control is shown in Figure 8b. active power control station i.e. it is just a measure of stiffness
The initial impact on the active power in station 1 is roughly so that a less stiff response (fast dc voltage control) resembles
the same for all cases. In the recovery phase, recovery to pre- low H and a more stiff response (slow dc voltage control)
fault value is affected significantly by the inertia. A higher resembles high H. Just as in the active power controlling
inertia gives a longer lasting active power deviation in station station, see Figure 9, an inertia requirement in the DC voltage
1 which is also reflected in the DC voltage controlling station controlling station will slow down the response of the voltage
2. The higher the inertia in station 1, the larger is the impact angle and thus result in a slower change in active power flow.
on the DC voltage of the link and the larger the active power This in turn will slow down and thus possibly deteriorate

PSCC 2024
0 of the link depending on the severity of the disturbance and
-0.1 the HVDC topology. This also further emphasizes the need
-0.2 for sufficiently fast DC voltage control. Another disturbance
that could be more critical for GFM schemes for the same
P 2 (pu)
-0.3
reason is large temporary temporary over voltage. To illustrate
-0.4
the discussion in this section, Figure 10 shows an example
-0.5
of an HVDC-link experiencing a 3-phase fault near station
-0.6 1 which is in active power control operating as inverter.
1.9 2 2.1 2.2 2.3 2.4 2.5 2.6 2.7 2.8 The connected AC grids on both stations are relatively
weak. In the DC voltage controlling station 2 operating as
rectifier in GFM operation, one fast DC voltage controller
1.03
corresponding to lower inertia and one slow DC voltage
1.02 controller corresponding to a higher inertia is considered. It
is seen that in the slow response case corresponding to higher
U dc2(pu)
1.01
inertia, the DC voltage controlling station is more sensitive
1 to the fault in the other station due to the slower DC voltage
control. Thus, it will not be possible for the DC link to
0.99
recover after this disturbance if the DC voltage controlling
0.98 station has higher inertia than feasible, as the control cannot
1.9 2 2.1 2.2 2.3 2.4 2.5 2.6 2.7 2.8 react fast enough to keep the voltage below admissible levels.
Time (s)
As explained above, this impact is due to slowing down the
Fig. 8b: Station 2, upper plot active power P2 , lower plot DC
change in AC voltage angle. The very same behaviour can
voltage, station 2 in DC voltage control
be expected if the same fault is here applied at the terminal
of DC voltage control station instead if active power control
0.58
is slow due to high inertia constants. For faults close to the
0.56
rectifier, the converter may instead go into over-modulation,
i.e., lose control due to low DC voltage if the DC voltage
0.54 can not be controlled fast enough by the remote station.
P1 (pu)
0.52
0.5
H=0.1s
H=0.5s
H=1.0s
0.48 H=2.5s
H=5.0s
0.46
1.9 2 2.1 2.2 2.3 2.4 2.5 2.6 2.7 2.8 2.9
Time (s)
Fig. 9: Step response in active power for different inertia, H
the DC voltage control, see also [91]. Any such requirement

must thus be thoroughly investigated and handled with great
care. A simultaneous inertia requirement in the active power
controlling station will also further increase the strain on DC
voltage stability since as discussed above in the previous
section, a disturbance in the active power controlling station
may then create an energy transfer requiring a fast response
from the DC voltage controlling station. This emphasizes that
an important limiting factor is related to the relatively low
energy capability of the link and the bandwidth limitations
of the DC voltage controller. This is often overlooked since
it is not directly related to AC side behavior but rather to the Fig. 10: Example of response to three phase fault in active
internal HVDC control and design. For certain disturbances, power controlling station 1 for different speed of DC voltage
for instance large phase jump or severe inverter AC faults, control in the DC voltage controlling station 2.
energy may accumulate on the DC side, which for an multi
modular converter will also transiently increase the submodule An aspect of modelling which is often not accounted
capacitor voltages. This could affect the continued operation for, since the focus is on the AC system response, is the

PSCC 2024
internal dynamics on the DC side of the HVDC link. A large the onshore grid until pre-fault values are reached. When grid
disturbance on the AC side will cause a disturbance also on phase angle at t = 0.9 s experiences a positive phase jump
the DC side. To investigate the impact, the DC side and the back to 0 degrees, the DC voltage instead increases meaning
corresponding control system consisting of DC voltage control the energy in submodule capacitance and cable capacitance
as well as energy balancing controllers for the converter arms increase. Had the DC control reacted slower to the first
would need to be modeled. For GFM converters, this may be disturbance (as would have been the case with a higher inertia),
of extra importance since an inertia requirement as discussed the DC voltage would have decreased more and there would
in the previous section will slow down the converter output have been a risk to compromise the DC voltage stability. This
voltage angle and thus the DC voltage dynamics. For this will always be the case if the other converter is in control of
reason, it could also be important to model the other station frequency and voltage not contributing to DC voltage control.
in the same detailed manner, especially if there is an inertia In principle, similar behavior is expected also if the remote
requirement in both stations, since the dynamics of both station is in active power control mode.
stations as discussed will affect the DC voltage stability.
1.2
As shown in section IV-D events like large voltage phase
P 1 (pu)
1
jumps (depending on jump magnitude and grid strength) may
0.8
even pose a challenge in controlling multi modular converter
0.6
energy or submodule capacitor voltages, which is overlooked 0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6
-0.6
if a simplified model is used. In the same way, a situation as
P 2 (pu)
-0.8
depicted in Figure 10 would be difficult to evaluate without -1
a detailed representation of the DC side. There is a direct -1.2
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6
connection between the different requirements on an HVDC 1.05
U dc2
link and the physical design, and thus to the design cost. In
Udc (pu)
U dc1
a case where the purpose is to determine that a design is not 1
only technically, but also economically feasible a detailed 0.95

0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6
model representation of the DC side as described above is 0
needed.
Ph 1 (deg)
-5
-10
The following example considers GFM capability of DC 0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6
voltage control in an HVDC connecting Power Park Modules Time (s)
(PPM’s). An HVDC converter can have islanded operation Fig. 11: Example of response to 10 degree phase jumps in
capability only if there is another supporting station that onshore grid.
controls the DC voltage in a stable manner. Figure 11 shows
a case where GFL wind turbines are connected to an offshore
converter controlling the voltage and frequency offshore as C. Transient current capability
is traditionally done, thus acting as a slack bus for offshore A VSC cannot tolerate transient overcurrent levels com-
power. This means that the offshore converter can be con- parable to those of a synchronous machine. It normally has
sidered as a constant power source not contributing to DC an impedance smaller than that of the synchronous machine
voltage control. As a consequence, DC voltage stability must and less current capacity. Further, during the fault and the
be ensured via DC voltage control in the onshore station. If following fault quasi steady state, the fault current provision is
GFM behavior in terms of inertia is required by the onshore limited to one per unit or slightly above that. It is important to
converter to supply/absorb the active power required by the note that this is the total current, including both positive and
AC grid depending on the type of transient events, there will negative sequence. It could thus be necessary to determine
be a conflict between maintaining the DC voltage stability how to prioritize between positive and negative sequence
and GFM behavior since the DC voltage control will be fault current provision from the converter. When the transient
slowed down. In the example in Figure 11, P1 is the active current is limited in the VSC, it will affect the behavior for
power in onshore converter, P2 is the active power in offshore large disturbances which will then deviate from that of a
converter, UDC1 and UDC2 are the DC voltages of onshore synchronous machine. Any GFM implementation must thus
and offshore converter respectively and δ1 is the grid phase be equipped with a scheme to limit transient overcurrent in
angle of the onshore grid. At t =0.1 s, a 10 degree negative order to guarantee continued operation. If the implementation
phase jump occurs. The active power infeed from offshore includes an inherent current controller, this is trivial. If a pure
converter remains constant throughout the disturbance period. voltage source emulation (VSE) scheme without underlying
This means that the energy fed to onshore in the phase jump current control is implemented, this must be performed by
comes from the converter submodule capacitance and cable other means. One possibility is to do a hard mode shift to
capacitance and the DC voltage starts to fall quickly. Soon current control when the disturbance is detected. Other possi-
after that, the DC voltage control in the onshore converter bilities include current limiters in parallel to the VSE scheme.
reacts to restore the DC voltage by taking the energy from Note that with a properly implemented current limitation

PSCC 2024
scheme, the current level for a grid fault should not be critical
itself. Still, the transient current during a disturbance could be
critical in the sense that the DC side voltage or submodule
capacitor voltages are charged to a critical level. To further
improve current limitation, a virtual impedance or admittance
via introducing an artificial voltage drop in the converter
may be implemented. This could be crucial to increase the
withstand capability, see IV-D, and also to improve damping.
D. Virtual impedance impact on large power disturbances

For large power disturbances (e.g., large phase jumps), a
GFM converter will behave differently from a GFL converter
since the voltage angle of the GFM converter will not follow
the grid voltage angle as quickly. The instantaneous change in
active power is given by (2). It is seen from the equation that
the larger the total converter impedance Xconv including the
virtual impedance, the less the corresponding instantaneous
change in active power will be. Thus, in the GFM converter,
implementation of a virtual impedance or admittance as men-
tioned in the previous section could be necessary for withstand.
Fig. 12a: Positive phase jump of 30 degrees in active power
Figure 12a shows such an example for a 30 degree positive
controlling station 1, with and without virtual impedance (VI).
phasejump in a grid with SCR=20. The phase jump is applied
in active power controlling station in cases with and without
a virtual impedance implemented. The case without added
virtual impedance trips while the case with virtual impedance
withstands the large disturbance. A large positive phase jump
during inverter operation when running close to full power and
the grid is strong is anticipated to be the most critical case. In
that case, the issue would be related to the DC over-voltage or
submodule capacitor voltage caused by the rapid change of AC
side active power. In the case without virtual impedance shown
here, the power exceeds 1.5 pu. The virtual impedance helps in
limiting the difference between DC and AC power and hence
the charging of DC side energy and submodule capacitors.
This is an example showing the importance of modeling the
DC side dynamics in an multi modular converter properly.
Virtual impedance will help with improving performance for
phase jump in both directions. Figure 12b shows the similar
cases but for a 30 degree negative phasejump. Here, DC side
charging is not as critical as in the positive phasejump case in
Figure 12a. Current is hard limited in both cases, avoiding a
trip. However, due to the less aggressive current limitation in
the case utilizing a virtual impedance, the frequency support
properties (visible in active power response) are better in that Fig. 12b: Negative phase jump of 30 degrees in active power
case in the sense that the frequency support is given faster. controlling station 1, with and without virtual impedance (VI).
E. Synchronization
For a converter connected to a grid where other units can af- power. The conflict of the synchronization stability and the
fect frequency and voltage, some grid synchronization method inertia response of the GFM converter is discussed in [92],
is necessary to handle disturbances where the converter goes where a transient damping method (in which grid frequency
into limitation. The literature describes both schemes with and is estimated using a PLL) is added in the active power control
without a phase-locked loop (PLL). A relatively fast contin- loop.
uous synchronization to the grid may improve the transient An important aspect related rather to steady state behaviour
stability and possibility to resynchronize to the grid after large without any grid synchronization which may not be obvious
disturbances. Examples of such disturbances are large voltage is the power order tracking when frequency deviates from
angle jump or high RoCoF when converter operates at rated nominal, [93]. The deviation from the active power order

PSCC 2024
may then not be acceptable for an HVDC. The deviation forming control and establish a coherent theoretical framework
from the active power order for VSM in steady state without around this concept. This is the aim of this section which is
frequency synchronization can be understood from Figure organised as follows:
24 and the following discussion. With a reasonable choice • Bibliographic review of a selection of key papers address-
of parameters, the power deviation for a 0.5 Hz frequency ing grid forming control
deviation from nominal frequency could be around 0.5 pu. • Identification of the dynamic model of the system
which is large considering for instance a 1 GW link. In that • Design of the active power control
case, an estimation of grid frequency via a PLL/FLL would be • Enhancement of large-disturbance stability
necessary to compensate for the frequency deviation to achieve • Small-signal stability analysis in a wide range of grid
proper power tracking. strength
Some simulation examples are shown in Figure 13. Here, • Proposal of a 4 VSCs benchmark to illustrate previous
a change in frequency from 50 to 50.2 Hz is applied. The insights on a more complex system.
different cases show synchronization with different time con-
The proposed material will mainly focus on transmission
stants of a PLL used for synchronization. In the case with
grid connected converters.
no synchronization a large (0.1 pu.) deviation in active power
remains in steady state after the frequency change. Although
A. Short bibliographic review on grid forming control
coherent with GFM properties, such a large deviation from
the power reference might not be desirable. It should be noted Different solutions have been proposed to control the active
that, besides the use for reference following and resynchro- power exchanged between the converter and the AC grid
nization after disturbances, a PLL is traditionally also used with controlled voltage angle which is the fundamental prin-
for synchronization to grid before deblocking the converter. ciple of the grid forming control. Among them, the droop
In some papers, e.g. [94], it is claimed that it is possible to control-based [95] is one of the most well-known solution.
synchronize to grid before connection to grid without using a The frequency of the modulated voltage is calculated from
PLL. However, in many applications (HVDC/FACTS) such a a difference between the power reference and the power
solution is not feasible as it assumes that there is already an measurement. It can be demonstrated that a frequency droop
established DC voltage before energizing the converter. control is naturally integrated in this type of control. When
the participation to primary frequency regulation service is
1.05
T1 not required, it is possible to incorporate an estimation of
1
5T1 the grid frequency into this control scheme [96] to ensure
25T 1
reference tracking in steady state. Various control structures
P(pu)
No synch.
0.95
aim to emulate the synchronous generator like the Virtual syn-
0.9
chronous machine [97] [98], Synchronverter [99], Inducverter
0.85
0 0.5 1 1.5 2 2.5 3
[100], Virtual induction machine [100]. With a Proportional
Integrator (PI) controller [101], the grid forming control is also
providing some inertial effect without using any information
50.2
on the grid frequency. The matching control [102] is based on
an analogy between the capacitor of the DC bus and the inertia
f(Hz)
50.1
of the synchronous machine. The virtual oscillators has been
50 introduced in order to facilitate the self synchronisation [103].
0 0.5 1 1.5 2 2.5 3 In [104], the authors improve the power reference tracking
Time(s) capability of a grid forming control thanks to a feed-forward
Fig. 13: Example of active power response with different action on the angle.
synchronization speed to a grid frequency change from 50 In addition, the regular presence of an LC or LCL filter
to 50.2 Hz. on the AC side of a VSC also requires to be managed by
the control scheme as it brings additional dynamics to the
converter. A possible solution is to implement a cascaded
V. T OWARDS A COMPREHENSIVE MODELING OF GRID
voltage-current loops [105], [106]. The optimal design of
FORMING CONTROL
the current and voltage controllers are not straightforward.
The evolution of the power system has been outlined in In [105] a tuning approach based on the optimization of the
previous sections. New rules are being incorporated to grid placement of eigenvalues is used. In [107], a specific Model
codes and, among them, notably those dealing with the GFM Predictive Control to design the voltage and current controller
capability. On the manufacturer side, providing solutions with is used. Alternative options for designing these controllers
such capabilities has important implications, as also described can be found in [108], [109]. Additionally, it is possible to
in previous sections. achieve direct voltage control, which is a simpler solution
In addition to this industrial challenge, a twofold theoretical requiring less tuning effort [96]. Consequently, the multitude
challenge has to be tackled: provide a definition of grid of grid forming control schemes, especially the diverse names

PSCC 2024
ig , vg
assigned to controllers with essentially similar characteristics,
contributes to confusion among engineers and researchers. Lc ig
PCC
Lg
Several reviews on grid-forming control can be found in P∗
the literature [110]–[112]. However, to the best of the authors’ Q∗ VSC ∗

vm
vm vg ve
knowledge, none of these reviews delve into the fundamental V ∗
control
principles required to unify and design the various types of
grid forming control.
Fig. 15: Connection of a VSC to a Thévenin equivalent
B. Definition of the model for the design of the power syn-
chronisation control
Similarly, a phasor can be associated to the current ig flowing
Before starting to design this control, it is highly important
out of the VSC (see Fig. 15). Assuming that the current phasor
to clearly define the model of the system to be controlled.
lags the phasor of the Thévenin voltage source by an angle ϕ,
Firstly the model of the converter must be specified. The
the former can be written as:
main function of the VSCs is to modulate the DC bus
voltage in order to generate a set of three-phase voltages Ig = Ig ej(δe −ϕ) (5)
vm = (vma , vmb , vmc ) as shown in Fig. 14a. This requires
defining a set of three-phase signals in the control vm ∗
= With this static formulation, the system can be represented by
∗ ∗ ∗
(vma , vmb , vmc ). The low-level control, which manages the the electrical circuit shown in Fig. 16a with the corresponding
transistors switching, aims to ensure that the average value phasor in Fig. 16b.
of the phase voltages over the transistor switching time Ts is
Ig jXc PCC jXg
equal to their corresponding control signals:
∗
< vmj >Ts = vm j
, j = {a, b, c} (3)
Vm Vg Ve
Hence, in a first approximation, the VSC can be assimilated
to an ideal three phase voltage source that is driven by the
control as sketched in Fig. 14b. In other words, the converter (a) electrical circuit representation
is represented by a lossless power amplifier between a signal d
∗
generated by the control vm and the actual modulated three- Vm
phase voltages vm . The dynamics of transistor switching can jXc Ig
be approximated by a delay that takes into account the internal Vg

δm
operation of the converter. q δg jXg Ig
Ve
∗
vm abc
δe
ϕ
∗ ∗ ∗ Ig
Low level control vm a
vm a
vm a
(b) phasor representation
vma
Fig. 16: Static model of the system
udc A
B vmb
C
vma vmb vmc
From that circuit, two expressions are easily derived for the
vmc
active power in terms of phase angle differences, namely:
Vm Vg
(a) Full model (b) Average model P = sin (δm − δg ) (6)
Xc
Fig. 14: Basic representation of a Voltage Source Converter
Vm Ve
P = sin (δm − δe ) (7)
Secondly, the grid connected to the converter has to be Xc + Xg
modeled. As shown in Fig. 15, the converter is assumed to From (6), it can be derived a first type of control [113]. In
be connected to a Thévenin equivalent system composed of the sequel, only the second formulation of the power (9) is
a voltage source ve in series with the grid inductance Lg . To considered to design the control.
simplify the analysis, the grid resistance is neglected, without Getting back to time domain, a time-varying angle is
loss of generality. On the converter side, a lossless connection associated to each angle in the phasor, according to:
is also considered in the form of an inductance Lc . With
a proper choice of per unit system, the transformer can be θx = δx + ωb t (8)
represented by merely its leakage inductance.
Substituting θm (resp. θe ) to δm (resp. δe ), Eq. (7) can be
In steady state, for a given grid frequency, it is possible to
rewritten as:
associate a phasor to each AC voltage according to:
Vm Ve
V x = Vx ej(δx ) (4) P = sin (θm − θe ) (9)
Xc + Xg

PSCC 2024
The VSC control is implemented in the well-known dq C. From a static model to a dynamic one
reference frame. The d axis is aligned with the phasor V m In order to design a control, a dynamic model is needed.
of the modulated voltage vm . Thus, in steady state: The aim of this section is to explain why the static model
vmd = V ∗ , vmq = 0 derived in the previous section can be also used as a dynamic
model for two different types of grid forming control including
Where V ∗ represents the voltage magnitude reference. a current loop or not.
Since vmd and vmq are generated by a controller, it is
possible to modify these components in order to incorporate Vg∗ Q P
a “virtual reactance” Xv , according to: V∗ AVR
vg
Vg RMS
dq
vmd = V ∗ − Xv igq vmq = Xv igd (10) Calculation P,Q

which are nothing but the steady-state equations of an reac- P∗ P ∗
∆vm d
Calculation
+ - s
Rv s+ω + jXv
tance in series with the VSC and the Thévenin inductances. f
∗
Its purpose has been presented in Section IV-D to mitigate the Active power ∆vm q
Control
effect on a phase shift on the power variation in the converter.
∗ ∗ igdq vgdq
With the latter, the model under consideration is hybrid, since vm d
vm q
it refers to both the physical components as in Fig. 15, and Inverse Park
θ∗ Park
Transformation Transformation
the virtual reactance, as shown in Fig. 17a. A virtual voltage ∗
vm
V is also added in the model. Introducing δ as the phasor abc
vg ig
angle associated to this voltage, the active power expression
delay
is easily rewritten to take Xv into account: Lc ig Lg
V Ve
P = sin(δ − δe ) (11)
Xv + Xc + Xg vm vg ve
It is also possible to associate a time domain angle θ to δ,
it yields: Fig. 18: Overall organisation of a Direct Voltage grid forming
V Ve control
P = sin(ρ) (12)
Xv + Xc + Xg
with the angle ρ defined as: 1) Direct voltage control: Fig. 18 shows the overall organ-
isation of the first type of Grid Forming control. The symbol
ρ = θ − θe = δ − δe (13) ∗
is employed to designate a control reference, corresponding
to its equivalent variable in the system.
V is now taken as reference of the dq frame (see Fig. 17b). The core element is the active power control which gen-
The active power control will be based on the phase angle erate the angle θ∗ of the time varying virtual voltage v ∗
difference δ − δe . associated to V . Doing so, it can be understood that this
loop is also synchronising the converter to the grid. This
jXv jXc PCC jXg
Ig double functionality is summarized in a unique expression:
the power synchronisation control. An Automatic Voltage
V Vm Vg Ve
Regulator (AVR) is controlling the voltage Vg at the Point
Control
of Common Coupling (PCC) (see Fig. 17a). More precisely,
it generates the magnitude V ∗ of that virtual voltage. As
(a) electrical circuit representation ∗ ∗
d previously explained, vm d
and vm q
are derived from V ∗ by
possibly adding a virtual inductance. Since there is direct link
∗ ∗
and V ∗ , this control can be called a Direct
V
jXv Ig between vm d
, vm q
Control Grid Forming. It is also referred in the literature as a
Vm
Voltage Control Grid ForMing (VC-GFM) control. This is the
jXc Ig name that will be used in the following sections.
Vg
δm
ρ In ordre to identify the dynamic model, an open loop control
q δ δg jXg Ig is firstly considered to check its validity in dynamic conditions.
Ve
Note that this type of control could not be used in a real
δe
ϕ application, only in simulation. By inverting Eq. (12) ρ∗ can
Ig
be obtained as a function of P ∗ . Assuming that both RMS
(b) phasor representation voltages Ve and V are equal to 1 pu, and linearizing Eq. (12)
Fig. 17: Static model extended to the virtual inductance yields:
ρ∗ = (Xc + Xv + X̃g )P ∗ (14)

PSCC 2024
Using this equation requires to have an estimate (X̃g ) of the 2) Introducing a current loop in the control: Traditionally,
grid reactance Xg . In simulation, it is possible to set X̃g = Xg VSCs have been driven in grid-following mode, which is cen-
The inversion of Eq. (13) yields θ∗ as a function of ρ∗ : tered around a current control loop. To maintain the advantage
of this control loop in terms of current limitation following a
θ∗ = ρ∗ + θ̃e (15)
significant disturbance, certain grid-forming control schemes
which shows the need to know the voltage phase angle θe . also keep the current control loop. In this case, the active
During this identification stage, it is possible to set this angle power is still controlled by the voltage angle but the incor-
equals to the Thevenin voltage angle. It yields in frequency poration of current loops requires generating d, q reference
domain: currents which are coherent with the voltage angle generated
θ̃e = ωb /s (16) by the active power control.
Those current references can be obtained from an estimation
Xc = 0:15 and Xv = 0
I˜ of the grid current I using the difference between the
1 voltages V and V g :
Rv = 0
0.9
Rv = 0:09
0.8
(V − V g )
I˜g =
Active Power (pu)
0.7 (19)
0.6 j(Xc + Xv )
0.5
0.4 Decomposing this estimation into its d, q components yields:
0.3
0.2 ĩgd = −vgq /(Xc + Xv ) (20)
0.1
0 ĩgq = (V ∗ − vgd )/(Xc + Xv ) (21)
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Time (s) The current references are set to the above estimated values,
Fig. 19: Simulation of an open-loop control of active power i.e.
i∗gd = −vgq /(Xc + Xv ) (22)
θe i∗gd = (V ∗ − vgd )/(Xc + Xv ) (23)
ρ P Fig. 21 shows the corresponding blocks to be added. The

θ∗ - V Ve
+ Xc +Xg +Xv left part deals with the generation of the d, q current references.
It can be proved that a filter (with cut-off frequency ωv ) has to
Power model
be applied to guarantee the stability of the system. The right
Fig. 20: Open-loop control of active power part shows a current control loop similar to the one found
in a grid-following control scheme. A simple proportional
As illustration, Fig. 19 presents simulation results obtained controller has been chosen (with gain Kpcc ) but a PI or even
for a step change of 0.5 pu of P ∗ with a 0.15 pu connection another controller is commonly used.
reactance Xc . The grid is considered strong enough so that
its reactance is negligible. It can be seen that the steady-state Quasi-static model Current Loop
value of P is equal to 0.5 pu as expected, but the dynamic i∗gd vgd
∗
response unveils an oscillation at 50 Hz. The latter is due to vgd wv
wv +s
1 +- Kpcc +
+
-
vm d
Xc +Xv
the poorly damped poles of the grid (corresponding to the
vd∗ +- igd Xc
aperiodic variation of the current in the abc frame). As shown
vq∗ +-
igq Xc
in [96], [114] they can be damped using a damping virtual −1 - +
∗
vm
+ Kpcc
i∗gq
+
q
resistance Rv as follows: vgd wv
wv +s
Xc +Xv
vgq
+
∗ s
vm = V ∗ − Rv ig (17) Fig. 21: Quasi static model and closed loop current control
d
s + ωf d
∗ s
vm = −Rv ig (18) The overall structure of grid forming control with current
q
s + ωf q control loop is shown in Fig. 22. The active power control
in which the washout filter avoids affecting the steady-state scheme is unchanged. No damping virtual resistance is needed
voltage, while ωf can be adjusted to obtain the desired for this control since the closed loop current control is already
dynamics. The response shown with blue curve in Fig. 19 has damping the grid modes.
been obtained for Rv = 0.09 pu. If the required bandwidth By way of illustration, Fig. 23 presents simulation results
on active power control is not too high, the small transient obtained for a step change of 0.5 pu of P ∗ . The gain Kpcc
can be considered negligible and both the static and dynamic has been set to yield a 3000 rad/s bandwidth for the system
models can be considered equivalent. It can be concluded that with current control loop. ωv has been set to 100 rad/s.
the dynamic system can be represented by the simple block The corresponding response time is around 3/ωv = 30 ms.
diagram shown in Fig. 20. The curves show that the responses with and without current

PSCC 2024
Vg∗ position 1). Hence, a simple integral action is enough to obtain
AVR P an accurate control of the power. Generally, a first-order filter
Vg Q ωc
ωc +s is included to avoid the double-frequency oscillations of
P∗ P V∗ RMS P,Q
power in case of unbalanced operation.
calculation Calculation
1 1
ωb θ̃e
θe
Active power Quasi static ω̃g 2 s
SW1
Control model ωc ωb θ∗ - ρ P
P∗ mp + Vm Ve
i∗gdq +− ωc +s
s
+ + Xc +Xg +Xv
Power model
Current
Loop (a) 1st formulation
∗ ∗ 1
vm d
vm q
vgdq igdq
1
θe
P0 ω̃g 2
SW 1
Inverse Park ∗ Park
θ +
P ∗
ωc + ω ωb θ∗ - ρ P
θ∗ Transformation Transformation +− mp Vm Ve
+ ωc +s + +
s Xc +Xg +Xv
∗
vm abc
Power model
vgabc igabc
1 − 1
Fig. 22: Overall structure of grid forming control with current R
+
SW 2
Optional
control loop Droop control
∆Pmax
(b) 2nd formulation

control loop are quite similar; The steady-state value is the Fig. 24: Closed-loop active power control
same, the dynamics are only slightly different. It can be
concluded that the variant including the current control loop A simple reorganisation of that scheme yields the one
and the quasi static model can be represented by the same shown in Fig. 24b. Considering for the time being that the
block diagram shown in Fig. 20. In the rest of the paper, switch SW1 is on position 1 and, SW2 is open, this control
the control scheme in Fig. 22 will be referred to as ”Current corresponds to the so-called “droop control”. The name comes
Control” Grid ForMing, CC-GFM in short. from the variation of the active power proportionally to the
frequency deviation. Indeed, in steady state, the voltage fre-
0.7
Without Current Loop quency ω, is equal to the grid frequency ωg . Hence, assuming
0.6 With Current Loop
that the upper selector is in position 1, it can be written:
Active Power (pu)
0.5
0.4 ωg = ω = 1 + mp (P ∗ − P ) (24)
0.3
or equivalently:
0.2
1
0.1 P = P∗ + (1 − ωg ) (25)
0
mp
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 which is a classical frequency control equation with mp as
Time (s) the droop coefficient. This can be seen akin an ”embedded”
Fig. 23: Simulation of open-loop active power control with a primary frequency support. However, it is not possible to limit
current control loop the contribution of the VSC to this primary frequency support.
If the VSC is not intended to participate in frequency
D. Closed-loop active power control control, it is possible to cancel its contribution by using the
estimate ω˜g of the grid frequency, i.e. (SW1 in position 2). In
We now get to the point of devising a closed-loop active this case, Eq. (25) becomes:
power control. Since the static model in Fig. 20 has been
1
shown to be a valid approximation of the system dynamics, P = P∗ + (ω˜g − ωg ) (26)
it can be used to design the sought controller. A step-by-step mp
approach is followed hereafter to define the various types of Since ω˜g = ωg in steady state, it can be deduced that the
control applied when opting either for VC-GFM or for CC- active power P is equal to its reference P ∗ . In the case of
GFM. The type of controller depends on the requirements for a fast frequency estimation, the embedded frequency support
the closed loop system. contribution mentionned above can be considered as cancelled.
The starting point is the very simple active power controller Then, it is possible to add an external frequency droop
presented in Fig. 24a. In this scheme, the voltage angle θe control of the type used in speed governors synchronous
can be seen as a disturbance for the active power control. The machines (see Fig. 24b with the SW2 closed). It is also
corresponding estimate θ̃e acts as a feed-forward action. It possible to limit the contribution to frequency support to a
is calculated by integration of the nominal voltage (SW1 in predefined value ∆Pmax .

PSCC 2024
Parameters In.uence
Another solution consists of adding an integrator in the loop 0.6
as shown in Fig. 25a. Indeed, in steady state, the input of the
added integrator should be zero, the active power P is then 0.5
equal to its reference P ∗ . In this second-order system, the
parameters could be adjusted for proper placement of the two 0.4
Active Power (pu)

poles [115], [116].
0.3
θe
ω ρ 0.2
ki ωb θ∗ -
P∗ +− +− + V Ve
P
Xc +Xg +Xv
s s
Power model 0.1
kd
SCR = 20 H = 5s
0 SCR = 2 H = 5s
SCR = 20 H = 1s
(a) Adding an integrator in the loop
K - ω̃g 0 0.5 1 1.5 2 2.5 3
+
Time (s)
θe
Fig. 26: Simulation results with a PI controller – 0.5 step on
θ∗ ρ P
P∗ +−
1 ω ωb
+
- V Ve the active power reference
− 2Hs s Xc +Xg +Xv
Power model
consists of specifying the damping. For a desired value ζ̃,

(b) VSM configuration the corresponding value of kd can be obtained by inverting
Eq. (30):
Fig. 25: Closed loop active power control with an inertial effect s
2(Xc + Xg + Xv )
kd = ζ̃ (32)
It is possible to use the added integrator to emulate an Hωb
“inertial” effect. Consider provisionally that kd = 0. The
following equation can be easily derived from Fig. 25a: This equation highlights that an estimate of the grid reactance
Xg is needed to determine kd . In the absence of any informa-
1
P = P∗ − s ω (27) tion, a zero value can be assumed for Xg , which means that
ki kd is adjusted for a strong grid. As (31) highlights, ωn depend
which reminds the link between the active power and the on the value of H as it has been showcased in Section IV-A.
derivative of frequency (classical “swing equation”) in a A well-known alternative is the Virtual Synchronous Ma-
synchronous machine. Referring even more to the synchronous chine (VSM) scheme shown in Fig. 25b. The similarity of
machine, the coefficient 1/ki in Eq. (27) can be redefined as: this scheme with the one in Fig. 25a has been shown in [101].
1 The first can be modified to yield the second, and conversely.
= 2H (28) In particular, the same damping is obtained by setting:
ki
where H plays the role of the inertia coefficient (of the rotating 2Hkd ωb
K= (33)
masses) in a conventional power plant. Typically, the value of Xc
H will be chosen to meet system inertia (i.e. rate of change
in the VSM scheme.
of frequency) requirements, while the choice of kd remains a
Fig. 26 a shows the simulation results for a PI controller in
degree of freedom. It can be chosen to obtain a satisfactory
case of a strong grid (SCR = 20) when the value of ζ̃ is set to
dynamic response, as explained next.
1. The general behaviour is similar to a second order system
Assuming Vm ≈ 1 and Vg ≈ 1, the transfer function
with a 5% overshoot as expected. The simulations have been
between P and P ∗ can be obtained as:
done for VC-GFM and CC-GFM, the results are superimposed
P 1
= (29) so only one result has been shown in this figure. When the
P∗ 1 + 2Hkd s +
2H(Xc +Xg +Xv ) 2
s SCR decreases to 2, the value of the grid reactance increases,
ωb
For this second-order system, the damping ratio ζ and the then the damping ratio is decreasing as shown with (30). More
frequency ωn are given by, respectively: simulation results and the simulation models can be found in
s [117]. The table below is a summary of the control parameters
Hωb which will be used on the following sections.
ζ = kd (30)
2(Xc + Xg + Xv )
E. Current saturation and Large Disturbance Stability
ωb
r
ωn = (31) As it is well-known, the power electronic converters can-
2H(Xc + Xg + Xv ) not handle large currents like synchronous machines. Con-
With the already mentioned single degree of freedom it is sequently, in case of a large disturbance, a current limitation
not possible to control both ωn and ζ. A natural choice must be included in the control system to protect the converter.

PSCC 2024
TABLE I: List of parameters for the control be employed, similar to its application in transient stability
Type of control Parameter Value analysis of synchronous machines. The extended P (δ) equa-
H 5s
ζ 1 tion including the additional virtual reactance XV I is given in
Xv 0.3 the following equation:
VC-GFM Rv 0.09
ωv 100 rad/s V Ve
CC-GFM PV I (δ) = sin δ (36)
current loop dynamic 3000 rad/s Xc + Xg + Xv + XV I
Inverse current control loop Current control loop

In normal operation, XV I = 0. However, in the case of a
vg d igq vg d
current limitation, XV I ̸= 0 due to the use of the virtual
i∗gd i∗gdsat
∗
vm
′
d +- +- Kp−1 ++ +- Kp +- ++
∗
vm d reactance method. Consequently, the maximum value of P (δ)
igd
Lf ωm igq CSA Lf ωm decreases. In this context, Fig. 28 illustrates a scenario that
Lf ωm Lf ωm
igd igq
∗
may arise in the case of a bolted fault. Under normal operating
+- ++ +- ++ vm
′
∗
vm +- Kp−1 Kp ++
conditions, P (δ) intersects with the power reference P ∗ at
q
i∗gq i∗gqsat
q
vg q igq vg q
the point 1. When the fault is applied, P decreases rapidly
Fig. 27: Artificial threshold current loop and its inverse struc- till point 2. At this stage, δ increases while P remains at a
ture. very low level. At point 3, the fault is cleared, leading to
an increase in active power; however, the current limitation
remains activated. In the case of using a virtual reactance, the
When this disturbance is ended, the question of resynchroniza- active power increases until the curve PV I (δ) is reached (point
tion has to be studied carefully in order to recover a stable 4). It will be shown in the dynamic simulation that the angle
post-fault operation. These are the two aims of this section. decreases when the active power P is superior to P ∗ . Due
1) Solutions for Current Limitations in Grid-Forming Con- to the higher damping coefficient of a grid forming compared
trol: In order to limit the current, the virtual reactance-based to a synchronous machine, there is nearly no overshoot on
method stands out as a well-known approach for current the angle when the fault is cleared. Hence, the Equal Area
limitations in the literature [118], [119]. The basic idea is to Criterion cannot be applied to the grid-forming control. After
increase the virtual reactance already presented earlier in this reaching point 4, the angle decreases until it reaches the
article. This transient reactance value is dynamically adjusted original P (δ) at point 5 and then converges to the original
with respect to current limitation criteria, activated when the operating.
current exceeds a specific threshold In . A possible solution is In case of using the CSA method, the PCSA (δ) is given by
described by the following equations: the following equation [123]:
V Ve Imax
(
KpPV I σX/R δI if δI > 0 PCSA = p sin δ (37)
XV I = (34) V 2 + Ve2 − 2V Ve cos δ
0 if δI > 0 ≤ 0
The operation could be the same, but the angle at the
RV I = XV I /σX/R (35)
fault clearance was such that the active power P is lower
Where δI = Ig − In . KpPV I and σX/R denote the virtual than P ∗ after the fault. Consequently, the angle continues
reactance proportional gain and virtual reactance ratio, respec- to increase and the system is unstable. The limit stability
tively. For further insights into the optimal parameter selection, angles (ψmaxCSA , ψmaxV I ) are illustrated for both current
additional details can be found in [120]. limitation methods. It can be deduced that this classical CSA
Another method is called the Current Saturation Algorithm method provides less stability margin than the virtual reactance
(CSA). It entails a current control responsible for determining approach. Further insights provided in [110], [124] underscore
the saturated current [121], [122]. To limit the current, the the coupling of current saturation with active power control,
current references have to be saturated. In the case of a CC- resulting in a reduction in the critical clearing time. Indeed,
GFM, the current loop is already implemented and the CSA Eq. (37) illustrates how active power expression varies with
can be reduced to a limitation of current references. In the the current saturation state, influencing the system’s stability
case of a VC-GFM, current control is artificially introduced limit.
and canceled out by ‘”inverse current control” [122]. Fig. 27 This analysis is based on a static model; however, it is
presents a simple example with a proportional controller, but essential to validate the model’s accuracy through dynamic
it can be extended to other types of controllers. In normal simulations.
operation, the initial voltage references remain unchanged Accordingly, an 80% voltage drop is applied to a Thevenin
∗ ∗′ equivalent voltage source during 200 ms. The VC-GFM of
(vm d
= vm d
). However, when a current limitation is needed,
the ”CSA” block saturates the current references. Fig. 27 is implemented with H = 5s and a 0.3 pu virtual
2) Impact of Current Limitation Type on Large Disturbance reactance. The initial state of active power is set to 0.8 pu.
Stability: While the current has been limited, the question Fig. 29a confirms the general trends given by the static
of restoring a stable operating point when the disturbance is model. Although some dynamic phenomena are noticed, the
canceled needs to be addressed. The classical P (δ) curve can operating points observed are in line with those given by

PSCC 2024
P without current limitation Additionally, recent research [127] proposes enhancing
P PV I with the application of virtual large disturbance stability through modifications to voltage
impedance as current limiter reference control. It aims to improve the angle stability of
PCSA with the application of the CSA large signals by modifying the acceleration and deceleration
zones. Indeed, it leads to an increase in the maximum of the
characteristic P (δ), particularly during the fault. However, it is
5 4 noteworthy that the new characteristic obtained remains lower
1 than the blue curve of Fig. 28 during the fault. Consequently,
P∗
4’ the new characteristic may also be below the P ∗ for a severe
grid fault then leading to instability.
2 3
δ(rad) A method referred to as ”virtual power” was introduced
ψmaxCSA ψmaxV I ψmax for the CC-GFM in [123], [128] in order to enhance the
π/4 π/2 3π/4 π
large disturbance stability. The original version of the virtual
Fig. 28: Schematic representation of P (δ) under normal con- power method is based on the CC-GFM control. It uses the
ditions and during current limitation. unsaturated current output from the quasi-static model (see
Fig. 31) to generate the virtual power Pvirt :
1.4
in normal conditions
1.2
Pvirt = vgd i∗gd + vgq i∗gq (38)
1
P* = 0.8 pu In Fig. 31, i∗gdqsat represents the saturated currents used as
0.8
0.6
references for the current loop. They are derived from the un-
0.4
saturated references i∗gdq , divided by a proportional coefficient
0.2 K such that:
0 1 – Initial
operating
-0.2
point i∗gdq = i∗gdq /K (39)
sat
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 1.1
[rad] Applying Kirchhoff’s equations between PCC and the grid
(a) Power angle curve voltage source yields:
1 1
vgdq = vedq + jXg igdq (40)

0.5 0.5
0 0
Merging these two equations, (22) and (23), leads to the
0 0.5 1 1.5 2 0 0.5 1 1.5 2
virtual system equation:
time [s] time [s]
1.2 1
V ∗ = ved + (Xc + Xv + Xg /K)i∗gq (41)
0.8
1 0 = veq − (Xc + Xv + Xg /K)i∗gd (42)
0.6
0.8
0.4 It results in the following expression for the virtual power
0 0.5 1
time [s]
1.5 2 0 0.5 1
time [s]
1.5 2 Pvirt (δ) curve:
(b) Dynamic response V Ve
Pvirt = sin δ (43)
Fig. 29: System response to 200 ms grid fault during the Xc + Xv + Xg /K
application of the current limitation. During normal operation Pvirt (δ) is similar to P (δ) since
K = 1. In case of current limitation K becomes larger than 1,
then the maximum of Pvirt (δ) is greater than the maximum
the static model and the system remains stable after the of P (δ). The most interesting feature of this method that,
fault clearance. Fig. 29b showcases the evolution of the main with the use of the virtual power, the virtual operating point
electrical variables with respect to the time. It has been can seamlessly converges towards this curve in case of large
checked with dynamic simulation that in the case of employing disturbance without any apprehension of over-current in the
CSA, the system becomes unstable. real current, as this latter is effectively limited by the CSA.
3) Improving Large Disturbance Stability through the Vir- Since this Pvirt (δ) has been enlarged compared with the two
tual Power Method: Numerous approaches have been pub- previous methods, the large disturbance stability is improved
lished to increase the stability margin of GFM control during for the virtual power and also for the real system since it is
large stability disturbances. One approach involved using a closely linked to the virtual one. In other words, it can be
PLL during current saturation and refining synchronization seen that there is a decoupling between the large disturbance
angle control [125], [126]. However, this method requires a stability enhancement principle and the current limitation.
complex control structure and a thorough understanding of Fig. 30 illustrates the theoretical evolution of the virtual
PLL-related dynamics. power following a bolted fault. The same types of operations

PSCC 2024
Pvirt P without current limitation
Part already found Additional part to
Pvirt for Eg = 1 pu
in VCGFM perform the CCGFM
′
4 ∗
vm dq
vgdq
vgdq
Inverse Quasi Static
Current Loop Model
igdq
0 sat ∆i∗g i∗gdq
∗ 1
P
sat
igdq
2 3
δ(rad)
++
ψmax
π/4 π/2 3π/4 π i∗gdq
Fig. 30: Schematic representation of P (δ) with the implemen- Imax sat
tation of the virtual power. CSA
i∗gdq Pvirt = vgd i∗gd + vgq i∗gq

sat
+−
can be observed as those explained in the previous section, igdq P∗ Pvirt
except that the Pvirt (δ) has been enlarged, making it more
Current Active Power
likely that the virtual system will recover its initial operating vgdq Loop Control
∗
point after the transient. vm dq
4) Application of Virtual Power Method to VC-GFM: Ex-

Inverse Park θ∗
panding upon the proven efficacy of the CC-GFM architecture Transformation
in handling large disturbance stability, it is proposed to adapt ∗
vm abc
the virtual power method for the VC-GFM. Fig. 31: Presentation of VC-GFM architecture with the imple-
The key idea is to apply the principle of the inversion control mentation of the virtual power method.
loop to maintain the VC-GFM control scheme in normal
operation, and to switch to CC-GFM in the case of current
saturation, as shown in Fig. 31. Subsequently, the virtual power 5
P virt ( )
4- Virtual system
P( ) characteristic for E g = 1 pu
method, as previously described, can be applied during large resynchronisation
P virt ( ) characteristic for Eg = 1 pu
disturbances. 4
To assess the effectiveness of the proposed method, the same 3

event as in Fig. 29 is simulated with the same initial operating
2
1- Initial
point. The results, depicted in Fig. 32 confirm that even if operating
the virtual power calculated by the simulation model is not 1
point P* = 0.8 pu
following exactly the theoretical curve, the dynamic evolution
is in line with the behavior that could be anticipated with 0
2- Angle divergence 3- Fault clearance
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
the static model. Fig. 33, highlights that the current is always [rad]
under control. There is no hardware limit for the virtual power Fig. 32: System response of the system following a 200 ms
since it is only a control variable. grid fault.
Furthermore, an enhancement to the virtual power method
has been proposed, as detailed in [128]. This modification
improves system dynamics and accelerates the restoration 1
4
process following fault clearance.

2
Other events, such as phase shift, line disconnection , have 0.5
been tested to evaluate the virtual power method within the

0 0
hybrid architecture described in Fig. 31 etc. In all theses cases, 0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
time [s] time [s]
the system recovers the stability after the end of the event.
1.5
1.2
F. Stabilizing effect of the grid forming control 1

1
While there is a consensus that GFM controls are expected

0.8 0.5
to provide services to the grid to counter the destabilizing
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
impact brought by the decrease of synchronous generation time [s] time [s]
and the increase of conventionally controlled PEIR, their study Fig. 33: Dynamic time response of the system following a 200
has been mostly limited to their own stability in weak grids ms grid fault.
and their provision of inertia and frequency control [129].

PSCC 2024
considered, the studies are either tuning oriented [135], [136],
or assume that a GFM is equivalent to an infinite bus, thus
regressing to a study with identical controls in the system
[137]. When the stabilizing impact of GFM controls is directly
targeted by a study [138], different GFM controls are consid-
ered as equally stabilizing, assuming adequate tuning and the
study is based exclusively on time-domain simulations without
any further insights on how GFM controls enhance the grid
strength. A recent study has tackled this same question [139],
but the controls did not account for the presence of a virtual
reactance, which has been introduced in subsections IV-D and
Fig. 34: Description of the test setup with a passive load V-B
In this subsection, the setup proposed in [133] and enhanced
TABLE II: Test setup parameters
in [139] is adopted and its state space model is analyzed.
Circuit Parameter Value
Snom , Sb 1.044 GVA
This setup is retained as it is unimpeded by the constraints
Unom , Ub 400 kV of the SUIB setup, and yet simple enough for a comprehen-
VSC
Pnom 1 GW sive analysis of interaction phenomena behind the instability
Qmax 300 MVAr
mechanisms. As shown in Fig. 34, the setup consists of three
Lc 0.15 pu
Filter converters, where V SC1 and V SC21 are controlled using a
Rc 0.005 pu
OHL
Ll 0.144 pu generic GFL control, as used in [139], and V SC22 is GFM-
Rl 0.0072 pu controlled using the IP scheme in Fig. 25a, with and without
Lg 1 pu
Grid current control shown in Fig. 22 and Fig. 18. The converters
Rg 0.1 pu
are set up such as V SC1 is the same size as the total size of
V SC21 and V SC22 :
This section is focused on the small signal stabilizing effect V SC21 V SC22 V SC1
provided by a GFM converter applied to a small benchmark. Snom + Snom = Snom = Sb (44)
As shown earlier in subsection V-D, the GFM active power Where Snomx
refers to the nominal power of x and Sb refers
control can follow different schemes with a similar level of to the base power used for the per-unit. The parameters of
dynamic performance. Moreover, it has been demonstrated in the setup are detailed in Table II, where the per-unit values of
subsubsection V-E4 that a VC-GFM control can perform just the Overhead Lines (OHL) and the grid are based on Sb and
as well as its CC-GFM counterpart during faults. Therefore, Ub , while each converter filter per-unit (pu) values refer to the
the question to ask is: for an identical active power control, MVA base (Snom V SC1 V SC21
, Snom V SC22
or Snom ) of each converter.
does the presence of a current loop play a role in the stabilizing The operating point studied in this section is such that no
performance of GFM controls? This is investigated in the power flows to the external grid, hence allowing the increase
following. of the electrical distance to the infinite source without any
1) A suitable test setup for small signal studies: Similarly static constraints, while the converters are at their nominal
to the previous sections, control studies are often based on powers:
a single converter connected to a Thévenin equivalent, also
referred to as a Single Unit Infinite Bus (SUIB) setup. Most (
literature studies restrain the study of converter controls to PV∗ SC1 = −(PV∗ SC21 + PV∗ SC22 ) = −Pnom
(45)
this setup to assess their dynamic performance and conclude VV∗SC1 = VV∗SC21 = VV∗SC22 = 1 pu
on their stability in weak grids. Small signal stability studies
utilize this setup to provide tuning recommendations [130], Where PV∗ SCi and VV∗SCi are the active power and voltage
[131], to assess controls stability limits, and to highlight the set points of each converter i ∈ {1, 12, 22}, respectively.
differences between different controls, such as GFM and GFL The setup at this operating point is representative of a study
[132]. case of a remote transmission corridor, where V SC1 is one end
However, it has been shown that, for a properly tuned of an HVDC link evacuating the power generated by a wind
control, a Thévenin equivalent setup is not sufficient for small farm split into a subset of generators in GFL mode (V SC21 )
signal studies as it is limited by reactive power constraints and another subset of generators in GFM mode (V SC22 ).
rather than the dynamic performance of the controls [133]. 2) Impact of GFM controls on the stability of the system:
Furthermore, the use of a SUIB cannot highlight any in- One way to highlight the stabilizing impact of the different
teraction phenomena among PEIR, which is an important GFM controls is by investigating the minimal size of the GFM-
phenomenon that needs to be accounted for [129]. While controlled V SC22 to reach the small signal stability of the
studies considering bigger systems can be found in literature, system. As previously reported in [139], the VC-GFM control
they have been limited to study cases with identical converters provides a stronger stabilizing effect compared to the CC-GFM
controls [134]. When both GFM and GFL converters are alternative as a much smaller VC-GFM-controlled V SC22

PSCC 2024
400-kV overhead lines
225-kV aggregated cables
66-kV bus
˜=
HVDC 1
= WP1
50 km (6 //) E
A1 ˜ A2
A
50 km (6 //) F WP2
B2
32 km
65 km
B
Thévenin
(a) equivalent
=
C
70 km
B1
˜= HVDC 2 ˜
Fig. 36: The Four-VSC benchmark system
GFM alternative.
Further studies using the participation factors of the dom-
inant modes, encircled in Fig. 35, have allowed the confir-
mation of previous findings as well: the VC-GFM-controlled
converter decouples the GFL-controlled converter from the
(b) weak grid, whereas the CC-GFM-controlled converter presents
Fig. 35: System eigenvalues for varying VSC22 proportions additional interactions due to its vulnerability to the PCC
under VC-GFM (a) and CC-GFM (b) voltage (used for the QSEM).
G. Extension of findings to a 4 VSC meshed benchmark

is required to stabilize the setup. This observed difference 1) Introduction of the four-VSC benchmark system: A
can be reduced when considering passive loads but remains four-VSC benchmark system is proposed in this section to
significant. Here, the same investigation is conducted to verify extend the previous findings. It has been designed to offer the
whether these trends stand when considering more complete following features:
GFM controls, including the virtual reactance and the AVR • small enough for its stability margins to be easily assessed
outer loop. In the same manner as [139], the state-space • involving several VSCs electrically rather close to each
model of the full system is constructed, linearized, validated other (to contemplate possible interactions)
against EMT time-domain simulation and then analyzed. The • connected to an external grid of adjustable strength in
grid reactance Zg is chosen to reflect a weak grid, such as terms of short-circuit power
a 100% GFL system would be unstable, here: Zg = 1 pu. • relying on generic models of the type used in the previous
Using the linearized state-space model, a parametric sensitivity sections
study is conducted to assess the minimal size, expressed The one-line diagram is shown in Fig. 36. The system hosts
V SC22
in % of Snom relatively to Sb , of the GFM-controlled two wind parks (WP1 and WP2) and the terminal converters of
converter required to reach small signal stability. The results two HVDC links (HVDC1 and HVDC2). The remote converter
are illustrated in Fig. 35. It can be seen that only a 8% of of each HVDC link is not modelled, the DC voltage being
VC-GFM control is sufficient for the system to regain small assumed constant. WP1 and WP2 are aggregated equivalents
signal stability where 22% of CC-GFM would be needed. If an of a large number of generators. The connection transformers
accepted dynamic performance is set as a damping ratio above of all four VSCs are represented explicitly.
10%, the VC-GFM control complies with such a performance The connection to an external system is considered through
requirement for as little as 13% while the CC-GFM-controlled the Thévenin equivalent attached to bus C. Since the Thévenin
converter can only reach such a performance for a proportion voltage source forces the system frequency to return to its
above 53%. nominal value in steady state, this 100% power-electronic con-
It is worth noting that this gap widens if weaker grids are verter based system is clearly not meant to address frequency
considered. For instance, for Zg = 2 pu, the minimum size issues.
of the GFM-controlled V SC22 is four times bigger if it’s The 400-kV part of the transmission grid is meshed, al-
controlled under CC-GFM (84%) rather than VC-GFM (21%). lowing to simulate the outage of one or two circuits without
Such findings join the previous conclusions found in [139]: disconnection from the external system. The line lengths are
the VC-GFM control allows for stability and acceptable dy- given in Fig. 36. Each wind park is radially connected to the
namic performance with a much smaller volume than its CC- grid through six 225-kV cables and six transformers in parallel

PSCC 2024
TABLE III: 4-VSC system: line and cable data
R X ωC/2 Snom
Ω/km Ω/km µS/km MVA
400-kV∗ 0.016 0.320 1.5 1300
225-kV∗∗ 0.0084 0.017 180.0 2400
pertain to: ∗ each circuit ∗∗ 6 cables in parallel
TABLE IV: 4-VSC system: transformer data

R X n Snom
pu pu MVA
A2 - A 0.005 0.15 1.02 2400
B2 - B 0.005 0.15 1.05 2400
A1 - A 0.005 0.15 1.02 1200
B1 - B 0.005 0.15 1.04 1700
WP1 - E 0.005 0.12 1.05 2400
(a) Active power reaction to a small signal disturbance
WP2 - F 0.005 0.12 1.04 2400
to cope with the maximum production of each park. The cables

are 50 km long and correspond to the AC connection of an
offshore wind farm. The shunt reactors at buses E and F aim
at absorbing the excess reactive power produced by the cables.
The line, cable, transformer and VSC parameters are given in
Table III, IV and V, respectively.
The benchmark has been modelled and is studied in EMTP,
where the 225kV cables were modelled using wideband cable
models based on the physical and geometrical properties of
the cables from [140]. (b) Voltage reaction to a small signal disturbance
By its meshed structure and the nominal power of its Fig. 37: WP2 reaction to a π/40 phase jump at the infinite
components, the present benchmark allows for the study of a voltage source
large spectrum of operating points. For the following studies,
two extreme operating points are retained and are detailed in
Table VI. of reaching the (static) loadability limit of the equivalent (for
H. Small signal stability properties of GFM controls large values of its reactance).
By setting the short circuit power provided by the external
The purpose of the study is the verification of the varying
Thévenin equivalent at bus C to 8 GVA, the system is small
stabilizing effects of the VC-GFM and CC-GFM controls. The
signal-unstable if all converters are controlled in GFL mode.
chosen operating point is the first from Table VI, the wind
However, by controlling the second wind park (WP2) with a
parks inject the same power, most of which is evacuated by
GFM mode, the system regains small signal stability as shown
the HDVC links. This results in a lightly loaded network. At
in Fig. 37a and Fig. 37b, where the power and voltage of
this operating point, the system stability is assessed in terms
WP2 are measured after the system experiences a π/40 phase
of minimal short-circuit power of the external system. The
jump at the Thévenin equivalent voltage source. Furthermore,
corresponding Thévenin reactance is varied. The net power
it can be observed that when WP2 is VC-GFM-controlled,
injection of the VSCs (minus the network losses) is taken
the system displays a better damping, more precisely double
by the load at bus C, which results in no power flowing
the damping ratio, than when it is CC-GFM-controlled. If the
into the Thévenin equivalent. Hence, the initial state remains
external grid is made even weaker, the study case with the
unchanged while the reactance is varied and there is no risk
CC-GFM controlled WP2 is the first to become small signal
unstable, which goes to confirm the previous trends regarding
TABLE V: 4-VSC system: converter data the difference in stabilizing properties of VC-GFM and CC-
Snom Pnom GFM schemes.
MVA MW
WP1 2400 2300
1) Large Disturbance Stability: As confirmed by the previ-
WP2 2400 2300 ous section, the VC-GFM control provides a better stabilizing
HVDC1 1200 1150 effect within the scope of small signal stability. The next
HVDC2 1700 1630
verification to be made is whether this control is compliant
TABLE VI: 4-VSC system: operating points with large disturbance requirements [141], [142]. Here, the
second operating point of Table VI is considered, this results
operating Power (MW) injected by
point # WP1 WP2 HVDC1 HVDC2 in a heavily loaded network, both wind parks and both HVDC
1 1500 1500 -1120 -1600 links injecting active power into the grid. Nevertheless, the
2 1500 1500 1150 1400 system is N-1 secure with respect to the outage of any 400-

PSCC 2024
mode. For the GFL converters, in addition to the structure
shown in [139], a Fault Ride Through (FRT) block is added
to the control to limit the currents in order to protect the
converters, to dynamically support the voltage during the fault
and to allow a slow active power recovery. The detailed control
structure is available in [117]. For the VC-GFM converter, the
control shown in Fig. 31 is used.
The EMT time domain simulations conducted in EMTP,
presented in Fig. 38, show that currents are successfully
limited in both cases. However, when all converters are in GFL
mode, the post fault overvoltage is dangerously high and the
system displays an oscillatory behavior post fault. On the other
(a) HVDC2 Voltage reaction to a large signal disturbance
hand, when HVDC2 in VC-GFM mode, the currents smoothly
transit from the saturated limit to their steady state value,
with the active power response remaining within the TSOs
required response time. Moreover, the post fault overvoltage
is significantly reduced and the oscillations are better damped.
This study case shows that even under the most demanding
large disturbance event, the solution identified as best for small
signal stability (here, VC-GFM) succeeds at complying with
large disturbance stability requirements in a more realistic
system than a SUIB setup.
VI. C ONCLUSION
(b) HVDC2 injected current reaction to a large signal disturbance The promises and challenges of GFM solutions have been
discussed, starting with an overview of the system needs and
the corresponding requirements applied by various TSOs to
grid-connected assets, notably PEIR. Subsequently, we delved
into the development of cost-effective technical solutions from
the perspectives of a BESS developer and a manufacturer of
HVDC systems. Finally, we presented the academic viewpoint
on GFM models and control strategies. This section summa-
rizes the key takeaways of this survey paper.
Power system security relies on a volatile balance between
the varied immunity of grid-connected devices and the overall
system performance. As power system dynamics evolve, this
equilibrium shifts. The greater the sensitivity of grid users,
(c) HVDC2 active power reaction to a large signal disturbance the more demanding the requirements on TSOs to constrain
Fig. 38: HVDC2 reaction to a 200ms three-phase fault fol- electrical quantities under plausible contingencies, and vice
lowed by a line trip versa. Therefore, both withstand capabilities and performance
requirements are systematically addressed through CNC.
By contributing to system strength, inertia, and damping,
kV circuit. No load is present, hence, the whole production GFM resources serve as pillars for the stable operation of
(minus the network losses) is exported to the external system power systems, alleviating constraints on other grid users. The
represented by the Thévenin equivalent at bus C. The short- more, the merrier: as the requirement extends to a greater
circuit power is set higher than the previous case to allow the number of devices, constraints are expected to decrease for all
evacuation of the power injected by the HVDC links and wind grid users. Similar to SG, PEIR can, to some extent, contribute
farms, while the system is small signal stable even for when to system stability as a by-product of their main purpose and
all converters are in GFL mode. limit the deployment of dedicated assets. However, contrary
The most severe large disturbance event considered here is a to SG this behaviour is not inherent; it must be specified.
250 ms three phase fault at the most loaded line of the system, Depending on the specific formulation of a new requirement
here one of the A-C lines transmitting 1.5 GW each, followed applied to a given facility, it may entail either minimal
by the opening of this line. design modifications (limited to control updates) or hardware
Using this operating point and disturbance event, the con- upgrades. Consequently, the description of the GFM capability
trols large disturbance stability is tested. All converters are has evolved with industry understanding of the burden it may
first in GFL mode, then HVDC2 is switched to VC-GFM pose on equipment sizing.

PSCC 2024
In particular, Core requirements has been established within is required to equip the grid with some voltage sources.
current and inherent energy limitations, while advanced fea- However, describing a GFM VSC merely as a voltage source
tures that may impact installation design are considered op- behind a reactance is insufficient for control design purposes.
tional. GFM BESS can effectively provide SCL and Inertia by Considerations on the desired active power control and the
optimizing the plant design to fit system needs. However, for synchronization to the grid are also required. One possible
resources with low to no inherent energy storage, the inertial solution is to combine both functionalities in a single loop:
response may fall into the additional requirements category. this is the concept of power synchronization control, which is
In operation, procedures to reserve headroom capacity may be the basis of grid forming control.
required to ensure enough resource availability at all times. The main aim of the academic part of this paper is to
Although originally referred to a type a control, the GFM propose some generic models for Grid Forming Converters
denomination has now been adopted by TSOs to designate that allow to derive generic conclusion on GFM controls
a specific set of features, performance-based requirements, without digging into specificities of each and every GFM
to guide OEM in the product development. As stakeholders control. This approach can help to bridge the gap between
converge in the technical specification of GFM capability, TSOs, with raw requirements, and OEM with detailed and IP
current efforts are being redirected to defining compliance protected models.
verification procedures, while dealing with technology-specific The genericity of the proposed models is based on the way
attributes and aggregated plant-level responses. the active power controller is designed.
Challenges in quantifying the necessary deployment of • Firstly, some considerations about the model were pro-
GFM resources to maintain system stability across different posed. Is a static model of the system accurate enough
time frames are partially due to the variety of control solutions to design the active power control? It can be answered
in the literature for PEIR. The similar functionalities offered positively, provided that too short a response time is not
may be, to some extent, captured by simplified generic models. requested for the closed-loop system. Indeed, it has been
Widespread KPI characterizing the decreasing voltage stiff- shown that the static model can be used for designing
ness, and thereby, increasing instability risks, include system both types of control, i.e. including a current loop (CC
(or regional) inertia and system strength. To inform decision- GFM) or not (VC GFM).
making at both planning and operational stages, additional • Secondly, the requirements for the control were defined.
efforts are required on their quantification and monitoring to In this case, whether an inertial effect is required or not,
account for PEIR. Suitable mitigation measures and opera- the type of controller is different.
tional procedures should also be developed accordingly. • Thirdly, when the model and the requirements are stated,
In the manufacturer part of the paper, it has been illus- the design of the controller can be achieved. It was
trated how different requirements are reflected on converter recalled that there is a link between the number of
performance. Some emphasis has been put on the internal DC poles under control and the number of parameters of the
side behaviour of the converter, which is often overlooked. controller. If an inertial effect is targeted, the integral part
The impact on converter and AC system performance of of the PI controller is imposed by external requirements,
large system disturbances in combination with virtual inertia hence the single remaining parameter, the proportional
and virtual reactance has been illustrated through simulations. gain, cannot control the two poles of a second-order
The possible contradictions between inertia requirements and system.
desired response speed of the control have also been stressed.
A general conclusion is that it is important to keep the distur- These three steps to design the active power controller are
bance specifications at a realistic level to avoid oversizing the very classical but to the best knowledge it is the first time that
converter and, hence, unnecessary costs. In the same way, the they have been emphasized in a such a way.
transmission and current capability of the converter should It is not claimed that the control derived from that analysis is
be respected in order to guarantee performance. In a severe the only possible one but the proposed approach has the merit
case, the converter may even trip jeopardizing stability of of integrating a wide range of control principles found in the
the system. The converter dimensioning may then need to literature. For example, highlighting that the same static model
be increased, which in turn will increase the cost. To give can be used for the control design with or without current loop
an example, a large phase jump may be the dimensioning leads to a unification of both control design which had not
scenario for the converter design. As long as general HVDC been showcased before. Moreover, the per unit design of the
performance is of concern, generic models and control system controller enhances the genericity of the proposed controls.
representations make up a valuable basis of discussions re- One of the interests of these models is to derive some
garding control impact on system performance between TSOs general trends of the dynamic behaviour for this kind of sys-
and manufacturers. For detailed design studies, a manufacturer tem. The following features, have been confirmed by detailed
model with a detailed representation of the DC side and simulations on the HVDC link presented in Section IV:
accurate time delays will always be needed. • If an inertial effect is required, the dynamics of the system
As it has been shown in the TSOs section, the classical depends on the inertia. This statement is confirmed by the
GFL converters are not aligned with the system needs: it simulation results of the HVDC link in Fig. 9

PSCC 2024
• A solution to prevent large active power variations in case MMC [143]. It appears that, in normal operation, the behaviour
of phase shift is to add a virtual impedance. This is also of this converter is very similar to an ideal VSC. In case
confirmed by the simulation results in Fig. 8a. of large disturbance, it may be needed to adjust some low
• In case of a simple active power control, it is possible level controls to achieve a good balance of the energy in the
to decouple active power control from primary frequency hundreds of submodules. Some extensions of the methodology
control by using a frequency estimation, typically pro- outlined in this paper to medium-voltage VSCs have been also
vided by a PLL. It must be stressed that this is not a considered [96]. This requires to control the internal energy
“frequency droop” control anymore. If needed, frequency of the LC filter and to take into account the non negligible
support can be added using an external loop but this is not resistive component of the grid.
compulsory and the frequency support can be limited to The last words will be to quote a few directions for future
a maximum power variation (∆Pmax ). Depending on the work:
grid frequency dynamic estimation, it is possible to obtain 1) The integration of the negative sequence in the GFM
a wide range of dynamics between no frequency support control: the two types of control, with and without
(high bandwidth for the PLL) and temporary primary current loop, can be applied to the treatment of the
frequency support (see Fig. 13) negative and positive sequence. This includes the defini-
While active power control is at the heart of grid forming tion of optimal references for the positive and negative
control, it is embedded in a larger set of control functionalities current/voltages.
including voltage control, current limitation, the stabilisation 2) Some works have already been successfully performed
of the system exposed to a large disturbance. In this last case, [144], [145] to extend the concept of using the voltage
it has been shown that the static models can also be used to angle to control the DC bus voltage of an HVDC link.
improve stability. This is a kind of extension of the power synchronisation
Unlike the synchronous machine, the controls of a grid control concept. The first results clearly highlight that
forming converter make it possible to manipulate the voltage this type of control is also bringing stability to the
angle to significantly stabilize its response to short- circuits or system.
phase shifts. Furthermore, using the virtual power method al- 3) Once the model is well defined, it is possible to en-
lows decoupling the current limitation from large-disturbance large the analysis to system-wide (large-scale) studies
stability enhancement. such as optimal placement of the GFM converters or
In order to study interactions between converters, a ded- studies involving converter-driven instability of the slow
icated two-VSC benchmark has been proposed. It allows type [146] and possibly interactions with other stability
studying the dynamics of the converters connected to a very issues, such as short term voltage instability.
weak grid without hitting the static loadability limit of a
Thévenin equivalent. The significant difference in terms of
stabilizing effect of respectively the VC-GFM and the CC-
GFM has been highlighted.
Next, a four-VSC benchmark system has been proposed
to increase the complexity of the system under study. This
benchmark opens the way to a wider range of tests and already
confirms the findings drawn from the previous tests. The
benchmark is fully documented and its model is available in
open source [117]. It is hoped that the models and data of this
benchmark will serve as reference when proposing alternative
control schemes, but also when validating them on commonly
shared test cases and scenarios, which is seldom the case in
the available literature.
The overall objective of the academic part of the paper
was to propose a generic and coherent approach to grid
forming control; it is hoped that the underlying concepts will
lead to a better understanding of this promising technology
and, hopefully, provide material for teaching this topic ! The
proposed models are well in line to study some of the TSO
requirements in terms of bringing stability, strength and inertia
to the grid. To also ensure DC side stability, a more detailed
representation of the converter is needed
Of course, not all aspects of grid forming have been
addressed in this paper. For instance this work has been
successfully extended to more realistic converters such as

PSCC 2024
D ISCLAIMER [21] TERNA, “Benefici di sistema e analisi robustezza rete,” Terna, 2023.
[Online]. Available: https://download.terna.it/terna/Terna Piano Svilu
The views expressed in this article are those of the authors ppo 2023 Benefici Sistema analisi Robustezza Rete 8db2549a9e7
and not necessarily those of AEMO. The information in this 19c7.pdf
[22] Y. Zhu, T. C. Green, X. Zhou, Y. Li, D. Kong, and Y. Gu, “Impedance
article does not constitute technical, legal or business advice. margin ratio: a new metric for small-signal system strength,” 2023.
Before relying on it, you should make your own enquiries. [Online]. Available: https://www.techrxiv.org/users/693819/articles/6
83279-impedance-margin-ratio-a-new-metric-for-small-signal-syste
R EFERENCES m-strength
[23] C. Henderson, A. Egea-Alvarez, T. Kneuppel, G. Yang, and L. Xu,
[1] Migrate H2020, “D3.1 description of system needs and test cases,” “Grid strength impedance metric: An alternative to scr for evaluating
2016. [Online]. Available: https://www.nationalgrideso.com/document system strength in converter dominated systems,” IEEE Transactions
/84261/download on Power Delivery, 2024.
[2] National Grid, “System needs and product strategy,” 2017. [Online]. [24] Q. Piraud, S. Dennetiere, and Y. Vernay, “Harmonic stability study for
Available: https://www.nationalgrideso.com/document/84261/downloa wind farm connection,” B4 International SC Meeting and Colloquium,
d 2023.
[3] J. Bialek, T. Bowen, T. Green, D. Lew, Y. Li, J. MacDowell, J. Mat- [25] A. Jalali, B. Badrzadeh, J.Lu, N.Modi, and M.Gordon, “System
evosyan, N. Miller, M. O’Malley, and D. Ramasubramanian, “System strength challenges and solutions developed for a remote area of aus-
needs and services for systems with high ibr penetratio,” 2021. tralian power system with high penetration of inverter-based resources,”
[4] Energy System Integration Group, “Grid-forming technology in energy CIGRE Science & Engineering, vol. 20, p. 27, 2021.
systems integration,” 2022. [26] NERC, “Odessa disturbance texas events: May 9, 2021 and june 26,”
[5] “Great britain grid forming best practice guide,” 2023. [Online]. NERC and Texas RE Staff Report, 2021. [Online]. Available: https://
Available: https://www.nationalgrideso.com/document/278491/downlo www.nerc.com/pa/rrm/ea/Documents/Odessa Disturbance Report.pdf
ad [27] ——, “2022 odessa disturbance texas event: June 4, 2022,”
[6] ACER, “Draft: Network code on requirements for grid connection of NERC and Texas RE Staff Report, 2022. [Online]. Available:
generators v2,” 2023. [Online]. Available: https://www.acer.europa.eu https://www.nerc.com/comm/RSTC Reliability Guidelines/NERC 2
/sites/default/files/documents/Recommendations annex/ACER Recom 022 Odessa Disturbance Report%20(1).pdf
mendation 03-2023 Annex 1a NC RfG TC to original.pdf [28] RTE, “Documentation technique de référence,” https://www.services-r
[7] AEMO, “Voluntary specification for grid-forming inverters,” 2023. te.com/files/live//sites/services-rte/files/documentsLibrary/31-01-202
[Online]. Available: https://aemo.com.au/-/media/files/initiatives/prima 3 COMPLET 1548 fr, 2023.
ry-frequency-response/2023/gfm-voluntary-spec.pdf [29] P. Panciatici, G. Bareux, and L. Wehenkel, “Operating in the fog:
[8] RTE, “Bilan annuel de la qualité de l’électricité,” 2022. Security management under uncertainty,” IEEE Power and Energy
[9] Bureau of Meteorology, “Severe thunderstorm and tornado outbreak Magazine, vol. 10, no. 5, pp. 40–49, 2012.
south australia 28 september 2016,” 2016. [Online]. Available: [30] RTE, “Reliability report,” 2022.
http://www.bom.gov.au/announcements/sevwx/
[31] AEMO, “Power system security guidelines,” 2023. [Online]. Available:
[10] AEMO, “Black system south australia september 2016 report,” 2017.
https://www.aemo.com.au/-/media/Files/Electricity/NEM/Security and
[Online]. Available: https://www.aemo.com.au/-/media/files/electricity
Reliability/Power System Ops/Procedures/SO OP 3715%20Powe
/nem/market notices and events/power system incident reports/201
r-System-Security-Guidelines.pdf
7/integrated-final-report-sa-black-system-28-september-2016.pdf
[32] ——, “Power system stability guidelines,” 2022. [Online]. Available:
[11] ENTSOE, “Network code on requirements for grid connection of
https://aemo.com.au/-/media/files/electricity/nem/security and reliabi
generators,” 2016. [Online]. Available: http://data.europa.eu/eli/reg/2
lity/congestion-information/power-system-stability-guidelines.pdf?la
016/631/oj
=en
[12] “Ieee standard for interconnection and interoperability of inverter-based
[33] RTE, “R#space,” 2020. [Online]. Available: https://codra.net/wp-conte
resources (ibrs) interconnecting with associated transmission electric
nt/uploads/2020/08/energy-sucess-story RTE.pdf
power systems,” IEEE Std 2800-2022, pp. 1–180, 2022.
[13] “Nf en 50549-2:2019 requirements for generating plants to be [34] A. Bahmanyar, D. Ernst, Y. Vanaubel, Q. Gemine, C. Pache, and
connected in parallel with distribution networks - part 2: Connection P. Panciatici, “Extended equal area criterion revisited: A direct method
to a mv distribution network - generating plants up to and including for fast transient stability analysis,” Energies, vol. 14, no. 21, 2021.
type b,” 2019. [Online]. Available: https://standards.iteh.ai/catalog/stan [Online]. Available: https://www.mdpi.com/1996-1073/14/21/7259
dards/clc/bd975a08-a7fc-47a1-9e46-b9c96cc2dafe/en-50549-2-2019 [35] ENTSOE, “Network code on high voltage direct current connections,”
[14] Z. H. Rather and D. Flynn, “Impact of voltage dip induced delayed 2016. [Online]. Available: https://www.entsoe.eu/network codes/hvdc/
active power recovery on wind integrated power systems,” Control [36] H. Saad, A. Schwob, and Y. Vernay, “Study of resonance issues be-
Engineering Practice, 2017. tween hvdc link and power system components using emt simulations,”
[15] AEMO, “System strength requirements methodology,” 2022. [Online]. 2018.
Available: https://aemo.com.au/-/media/files/electricity/nem/security [37] S. Dennetiere, H. Saad, Y. Vernay, P. Rault, C. Martin, and B. Clerc,
and reliability/system-strength-requirements/system-strength-require “Supporting energy transition in transmission systems: An operator’s
ments-methodology.pdf experience using electromagnetic transient simulation,” IEEE Power
[16] ——, “Power system requirements reference paper,” 2020. [Online]. and Energy Magazine, 2019.
Available: Power-system-requirements.pdf(aemo.com.au) [38] “Wind energy generation systems - Part 27-2: Electrical simulation
[17] C. KARAWITA and U. ANNAKKAGE, TB 909: Guidelines for models - Model validation,” IEC Standard 61400-27-2:2020.
Subsynchronous Oscillation Studies in Power Electronics Dominated [39] E.-M. Karmani, E. Mberou, Y. Vernay, F. Morretton, and M. Thibert,
Power Systems,JWG C4/B4.52. CIGRE, 2023. “New modeling requirements for inverter based resources and storage
[18] K. Elkington, A. Isaacs, M. Davies, C. Smith, R. Gagnon, P. Chris- system connection on the french transmission grid,” in MatPost 2023
tensen, and V. Lo, TB 671: Connection of wind farms To Weak Ac Conference, 11 2023.
Networks, WG B4.62, 2016, no. December. [40] C. Cardozo, A. Guironnet, J.-L. Marin, T. Prevost, and G. Torresan,
[19] L. Huang, H. Xin, Z. Li, P. Ju, H. Yuan, and G. Wang, “Identification of “Compliance criteria to assess rms model accuracy for wind power
generalized short-circuit ratio for on-line stability monitoring of wind plants based on iec 61400-27-2,” in 22nd Wind & Solar Integration
farms,” IEEE Transactions on Power Systems, vol. 35, no. 4, pp. 3282– Workshop, 09 2023.
3285, 2020. [41] AEMO, “Power system model guidelines,” 2018. [Online]. Available:
[20] M. O. Qays, I. Ahmad, D. Habibi, A. Aziz, and T. Mahmoud, https://aemo.com.au/-/media/Files/Electricity/NEM/Security and Rel
“System strength shortfall challenges for renewable energy-based iability/System-Security-Market-Frameworks-Review/2018/Power S
power systems: A review,” Renewable and Sustainable Energy ystems Model Guidelines PUBLISHED.pdf
Reviews, vol. 183, p. 113447, 2023. [Online]. Available: https: [42] ——, “Power system model guidelines,” 2023. [Online]. Available:
//www.sciencedirect.com/science/article/pii/S1364032123003040 https://aemo.com.au/-/media/files/electricity/nem/security and reliabi

PSCC 2024
lity/system-security-market-frameworks-review/2023/power systems [60] D. M. Schmieg, “P2021 gutachten zur nemo viii, los 1 - stabilität
model guidelines 2023 published .pdf (kurzfassung),” 2021. [Online]. Available: https://www.netzausbau.de/
[43] A. Dissanayaka and A. Isaacs, “System strength assessment of SharedDocs/Downloads/DE/Bedarfsermittlung/2035/NEP/NEP2035 N
the panhandle system pscad study,” 2016. [Online]. Available: EMOVIII-1.pdf? blob=publicationFile
https://www.ercot.com/files/docs/2016/03/01/panhandle system stren [61] AEMO, “2023 system strength report,” 2023. [Online]. Available: https:
gth study feb 23 2016 public .pdf //aemo.com.au/-/media/files/electricity/nem/security and reliability/s
[44] A. Dissanayaka, J. Wiebe, and A. Isaacs, “Panhandle and south texas ystem-strength-requirements/2023-system-strength-report.pdf?la=en
stability and system strength assessment,” 2018. [Online]. Available: [62] ——, “System strength impact assessment guidelines,” 2023. [Online].
https://www.ercot.com/files/docs/2018/04/19/Panhandle and South T Available: https://aemo.com.au/-/media/files/stakeholder consultation/
exas Stability and System Strength Assessment March....pdf consultations/nem-consultations/2022/ssrmiag/final-report/system-str
[45] M. Amin and M. Molinas, “Small-signal stability assessment of power ength-impact-assessment-guideline v2.pdf
electronics based power systems: A discussion of impedance- and [63] M. Zubiaga, J. J. Valera, C. Cardozo, T. Prevost, G. Denis, and
eigenvalue-based methods,” IEEE Transactions on Industry Applica- M. Paolone, “D3.3 analysis of the synchronisation capabilities of bess
tions, 2017. power converters,,” OSMOSE, 2022.
[46] L. Orellana, L. Sainz, E. Prieto-Araujo, M. Cheah-Mané, H. Mehrjerdi, [64] E. Martı́nez Carrasco, M. P. Comech Moreno, M. T. Villén Martı́nez,
and O. Gomis-Bellmunt, “Study of black-box models and participation and S. Borroy Vicente, “Improved faulted phase selection algorithm
factors for the positive-mode damping stability criterion,” 2023. for distance protection under high penetration of renewable energies,”
[47] “Sras guideline system restart ancillary services,” 2021. [Online]. 2020.
Available: https://aemo.com.au/-/media/files/electricity/nem/security a [65] V. Pradhan, N. George, O. Naidu, Z. Gajic, and S. Zubic, “Distance
nd reliability/ancillary services/sras/sras-guideline-2021.pdf?la=en protection of inverter based renewables power evacuating lines and
[48] “Fiche i4 : Reconstitution de reseau et contribution au maintien d’un downstream network: Issues and mitigation approach, 2022 ieee pes
reseau separe,” 2023. [Online]. Available: https://www.services-rte.c innovative smart grid technologies - asia (isgt asia),” 2022.
om/files/live//sites/services-rte/files/documentsLibrary/Article%208.24 [66] J. Quispe and E. Orduña, “Transmission line protection challenges
%20-%20Cahier%20des%20charges%20des%20capacit%C3%A9s% influenced by inverter-based resources: a review,” Prot Control Mod
20constructives%20pour%20un%20syst%C3%A8me%20HVDC fr Power Syst, 2022. [Online]. Available: https://doi.org/10.1186/s41601
[49] J. Loncle, P. Hondaa, J. M. A. Enjuanes, G. D. Castejón, A. D. -022-00249-8
Ferrer, L. Benard, J. Santana, V. Lehmann, T. Westerweller, and [67] N. G. ESO, “Frequently asked questions stability pathfinder -
M. Häusler, “Feedback on inelfe france-spain hvdc project,” 2016. phase 1 17 december 2019,” TBA, 2019. [Online]. Available:
[Online]. Available: https://www.e-cigre.org/publications/detail/b4-1 https://www.nationalgrideso.com/document/185281/download
24-2016-feedback-on-inelfe-france-spain-hvdc-project.html [68] “A unique window of opportunity capturing the reliability benefits of
[50] J. Dumas, V. Terrier, F. Bienvenu, S. Finet, and N. Grisey, “Dynamic grid-forming batteries,” 2023. [Online]. Available: https://www.esig.ene
sizing of required balancing capacities: the operational approach in rgy/wp-content/uploads/2023/03/ESIG-GFM-batteries-brief-2023.pdf
france,” in 2023 19th International Conference on the European [69] B. Badrzadeh, C. Cardozo, M. Hishida, S. Shah, I. Huq, N. Modi,
Energy Market (EEM). IEEE, Jun. 2023. [Online]. Available: and A. Morton, “Grid-forming inverters: Project demonstrations and
http://dx.doi.org/10.1109/EEM58374.2023.10161932 pilots,” IEEE Power and Energy Magazine, 2024.
[51] FinGrid, “Inertia of the nordic power system.” [Online]. Available:
[70] N. Modi, M. V. Escudero, K. Aramaki, X. Zhou, and P. Partinen, “High
https://www.fingrid.fi/en/electricity-market-information/InertiaofNord
inverter-based resource integration: The experience of five system
icpowersystem/
operators,” IEEE Power and Energy Magazine, 2024.
[52] MIGRATE, “D2.3: Lessons learned from monitoring & forecasting
[71] S.Sproul, N.Modi, S.Cherevatskiy, A.Jalali, and S.Zabihi, “System
kpis on impact of pe penetration,” 2018. [Online]. Available:
strength support using grid-forming energy storage to enable high
https://ec.europa.eu/research/participants/documents/downloadPublic?
penetrations of inverter-based resources to operate on weak networks,”
documentIds=080166e5bf1893ec&appId=PPGMS
CIGRE Paris 2022, 2022.
[53] Australian Energy Market Committee, “Rule determination national
[72] Y. Zhou, R.Zhang, D.Kathriarachchi, J. Dennis, and S. Goyal, “Grid
electricity amendment (managing the rate of change of power system
forming inverter and its applications to support system strength–a case
frequency) rule 2017,” 2017. [Online]. Available: Availableat:https:
study,” IET Gener. Transm. Distrib., vol. 17, pp. 391–398, 2023.
//www.aemc.gov.au/sites/default/files/content/ed599cb5-91b5-45db-b
157-99c0914ef9c3/ERC0214-Final-Determination-19-September-201 [73] Y. Cheng, “Preliminary assessment of grid forming inverter-based
7-published-version.pdf energy storage resources (gfm-ibr-esr) in the ercot grid,” 2023.
[54] AEMO, “Inertia requirements methodology,” 2018. [Online]. Available: [Online]. Available: https://www.ercot.com/files/docs/2023/08/11/G
https://www.aemo.com.au/-/media/files/electricity/nem/security and r FM ERCOT IBRWG(08112023).pdf
eliability/system-security-market-frameworks-review/2018/inertia req [74] [Online]. Available: https://www.agistin.eu/
uirements methodology published.pdf [75] M. Hishida, T. Redfern, X. Li, D. Hughes, D. Duckwitz, and
[55] ——, ““2023 inertia report”,” 2023. [Online]. Available: https: A. Knobloch, “Addressing transmission system operability challenges
//aemo.com.au/-/media/files/electricity/nem/security and reliability/s using multi-function large scale grid-forming bess solutions,” Energy
ystem-strength-requirements/2023-system-strength-report.pdf?la=en Storage Conference 2023 (ESC 2023), 2023.
[56] ICS Investigation Expert Panel, “Continental europe synchronous area [76] “Guidance notes for grid forming plant,” 2023. [Online]. Available:
separation on 08 january 2021,” 2021. [Online]. Available: https: https://www.nationalgrideso.com/document/289921/download
//eepublicdownloads.azureedge.net/clean-documents/SOC%20docum [77] AEMO, “Voluntary specification for grid-forming inverters: Core
ents/SOC%20Reports/entso-e CESysSep Final Report 210715.pdf requirements test framework,” 2024. [Online]. Available: https:
[57] ——, “Continental europe synchronous area separation on 24 july //aemo.com.au/-/media/files/initiatives/engineering-framework/2023/gr
2021,” 2022. [Online]. Available: https://eepublicdownloads.azureedg id-forming-inverters-jan-2024.pdf?la=en
e.net/clean-documents/Publications/2022/entso-e CESysSep 210724 [78] V. F. Guideline, “Vde fnn guideline : Grid forming behaviour of
02 Final Report 220325.pdf hvdc systems and dc-connected ppms,” 2020. [Online]. Available:
[58] ENTSOE, “Project inertia – phase ii: Updated frequency stability https://shop.vde.com/en/fnn-guideline-hvdc-systems-2
analysis in long term scenarios, relevant solutions and mitigation [79] “Specifications for grid-forming inverter-based resources,” 2022.
measures,” 2022. [Online]. Available: https://eepublicdownloads.blob [Online]. Available: https://www.energy.gov/sites/default/files/2023-0
.core.windows.net/public-cdn-container/clean-documents/sdc-docum 9/Specs%20for%20GFM%20IBRs%20Version%201.pdf
ents/231108 Project Inertia Phase II First Report FOR PUBLICA [80] “White paper: Grid forming functional specifications for bps-
TION clean.pdf connected battery energy storage systems,” 2023. [Online]. Available:
[59] J. Song, M. Cheah-Mane, E. Prieto-Araujo, and O. Gomis-Bellmunt, https://www.nerc.com/comm/RSTC Reliability Guidelines/White Pap
“Short-circuit analysis of grid-connected pv power plants considering er GFM Functional Specification.pdf
inverter limits,” International Journal of Electrical Power & [81] ENTSOE, “Expert group on connection requirements for offshore
Energy Systems, vol. 149, p. 109045, 2023. [Online]. Available: systems – phase ii (proposal for the nc hvdc amendment),” 2023.
https://www.sciencedirect.com/science/article/pii/S0142061523001023 [Online]. Available: https://eepublicdownloads.blob.core.windows.net/

PSCC 2024
public-cdn-container/clean-documents/Network%20codes%20docum [99] Qing-Chang Zhong, Phi-Long Nguyen, Zhenyu Ma, and Wanxing
ents/GC%20ESC/CROS/Final Report - Phase 2 01.pdf Sheng, “Self-Synchronized Synchronverters: Inverters Without a
[82] InterOpera, “D2.2 grid-forming functional requirements for hvdc Dedicated Synchronization Unit,” IEEE Trans. Power Electron.,
converter stations and dc-connected ppms in multi-terminal multi- vol. 29, no. 2, pp. 617–630, Feb. 2014. [Online]. Available:
vendor hvdc systems,” 2024. [Online]. Available: https://interopera.e http://ieeexplore.ieee.org/document/6504534/
u/wp-content/uploads/Docs/deliverables/InterOPERA-D2.2-GFM-fun [100] M. Ashabani, F. D. Freijedo, S. Golestan, and J. M. Guerrero,
ctional-requirements-v1.1-Submitted-PU.pdf “Inducverters: PLL-Less Converters With Auto-Synchronization and
[83] AEMO, “Nem engineering framework priority actions,” 2022. [Online]. Emulated Inertia Capability,” IEEE Trans. Smart Grid, vol. 7,
Available: https://aemo.com.au/-/media/files/initiatives/engineering-fra no. 3, pp. 1660–1674, May 2016. [Online]. Available: http:
mework/2022/nem-engineering-framework-priority-actions.pdf?la=en //ieeexplore.ieee.org/document/7258370/
&hash=F5297316185EDBD4390CDE4AE64F48BB [101] T. Qoria, E. Rokrok, A. Bruyere, B. Francois, and X. Guillaud,
[84] ——, “Application of advanced grid-scale inverters in the nem,” 2021. “A PLL-Free Grid-Forming Control With Decoupled Functionalities
[Online]. Available: https://aemo.com.au/-/media/files/initiatives/engin for High-Power Transmission System Applications,” IEEE Access,
eering-framework/2021/application-of-advanced-grid-scale-inverters-i vol. 8, pp. 197 363–197 378, 2020. [Online]. Available: https:
n-the-nem.pdf //ieeexplore.ieee.org/document/9240975/
[85] A. Cerretti and T. SChaup, “Cenelec tc8x wg03 contribute: Acer [102] C. Arghir and F. Dorfler, “The Electronic Realization of Synchronous
workshop on rate of change of frequency and grid forming capabilities,” Machines: Model Matching, Angle Tracking, and Energy Shaping
2023. Techniques,” IEEE Transactions on Power Electronics, vol. 35,
[86] B. Bahrani, M. H. Ravanji, B. Kroposki, D. Ramasubramanian, no. 4, pp. 4398–4410, Apr. 2020. [Online]. Available: https:
X. Guillaud, T. Prevost, and N.-A. Cutululis, “Grid-forming inverter- //ieeexplore.ieee.org/document/8825565/
based resource research landscape: Understanding the key assets for [103] B. B. Johnson, S. V. Dhople, A. O. Hamadeh, and P. T.
renewable-rich power systems,” IEEE Power and Energy Magazine, Krein, “Synchronization of Parallel Single-Phase Inverters With
2024. Virtual Oscillator Control,” IEEE Transactions on Power Electronics,
[87] T. Engelbrecht, A. Isaacs, S. Kynev, J. Matevosyan, B. Niemann, A. J. vol. 29, no. 11, pp. 6124–6138, Nov. 2014. [Online]. Available:
Owens, B. Singh, and A. Grondona, “Statcom technology evolution for http://ieeexplore.ieee.org/document/6692879/
tomorrow’s grid: E-statcom, statcom with supercapacitor-based active [104] S. D’Arco and J. A. Suul, “Phase Angle Feed-Forward Control for
power capability,” IEEE Power and Energy Magazine, 2023. Improving the Power Reference Tracking of Virtual Synchronous
[88] A. Abdalrahman, Y.-J. Hafner, M. Kumar Sahu, and K. Kumar Nayak, Machines,” IEEE Transactions on Industry Applications, pp. 1–13,
“Grid Forming Control for HVDC Systems: Opportunities and Chal- 2023. [Online]. Available: https://ieeexplore.ieee.org/document/10286
lenges,” in 2022 24th European Conference on Power Electronics and 151/
Applications (EPE’22 ECCE Europe). Hanover, Germany: IEEE, Sep. [105] S. D’Arco, J. A. Suul, and O. B. Fosso, “Automatic Tuning
2022, pp. 1–10. of Cascaded Controllers for Power Converters Using Eigenvalue
[89] C. Barker, “Voltage Source Converter (VSC) HVDC responses to Parametric Sensitivities,” IEEE Trans. on Ind. Applicat., vol. 51,
disturbances and faults in AC systems which have low synchronous no. 2, pp. 1743–1753, Mar. 2015. [Online]. Available: https:
generation,” CIGRE, Vienna, Austria, Tech. Rep., Sep. 2023. //ieeexplore.ieee.org/document/6891273
[90] Y. Jiang Hafner and M. Manchen, “Stability Enhancement and Blackout
[106] J. Wang, L. Chen, Z. Liu, Z. Zhang, and X. Zhang, “Optimized
Prevention by VSC Based HVDC,” Bologna, Italy, 2011, p. 9.
Parameter Design of the Dual-Loop Control for Grid-Forming VSCs
[91] E. Rokrok, T. Qoria, A. Bruyere, B. Francois, and X. Guillaud,
With LC Filters,” IEEE Transactions on Industry Applications,
“Integration of a storage device to the DC bus of a grid-
vol. 58, no. 1, pp. 820–829, Jan. 2022. [Online]. Available:
forming controlled HVDC interconnection,” Electric Power Systems
https://ieeexplore.ieee.org/document/9580573/
Research, vol. 212, p. 108601, Nov. 2022. [Online]. Available:
https://linkinghub.elsevier.com/retrieve/pii/S0378779622006848 [107] J.-M. De Paris, H. Pinheiro, F. D. M. Carnielutti, V. F. Montagner,
and D. M. Lima, “MPC for Grid Forming Converters with Current
[92] X. Xiong, C. Wu, B. Hu, D. Pan, and F. Blaabjerg, “Transient
Limiting,” in IECON 2022 – 48th Annual Conference of the IEEE
Damping Method for Improving the Synchronization Stability of
Industrial Electronics Society. Brussels, Belgium: IEEE, Oct. 2022,
Virtual Synchronous Generators,” IEEE Trans. Power Electron.,
pp. 1–5. [Online]. Available: https://ieeexplore.ieee.org/document/996
vol. 36, no. 7, pp. 7820–7831, Jul. 2021. [Online]. Available:
8998/
https://ieeexplore.ieee.org/document/9303430/
[93] A. Narula, M. Bongiorno, M. Beza, J. R. Svensson, X. Guillaud, and [108] H. Wu and X. Wang, “Passivity-Based Dual-Loop Vector Voltage
L. Harnefors, “Impact of steady-state grid-frequency deviations on and Current Control for Grid-Forming VSCs,” IEEE Transactions
the performance of grid-forming converter control strategies,” in 2020 on Power Electronics, vol. 36, no. 8, pp. 8647–8652, Aug. 2021.
22nd European Conference on Power Electronics and Applications [Online]. Available: https://ieeexplore.ieee.org/document/9311428/
(EPE’20 ECCE Europe). Lyon, France: IEEE, Sep. 2020, pp. P.1–P.10. [109] M. Nurunnabi, S. Li, and H. S. Das, “Control and Operation
[Online]. Available: https://ieeexplore.ieee.org/document/9215947/ Evaluation of Grid-Forming Inverters with L, LC, and LCL Filters,”
[94] Q.-C. Zhong, P.-L. Nguyen, Z. Ma, and W. Sheng, “Self-Synchronized in 2023 IEEE Power & Energy Society General Meeting (PESGM).
Synchronverters: Inverters Without a Dedicated Synchronization Unit,” Orlando, FL, USA: IEEE, Jul. 2023, pp. 1–5. [Online]. Available:
IEEE Trans. Power Electron., vol. 29, no. 2, pp. 617–630, Feb. 2014. https://ieeexplore.ieee.org/document/10253060/
[Online]. Available: http://ieeexplore.ieee.org/document/6504534/ [110] R. Rosso, X. Wang, M. Liserre, X. Lu, and S. Engelken, “Grid-
[95] Y. Mohamed and E. El-Saadany, “Adaptive Decentralized Droop Forming Converters: Control Approaches, Grid-Synchronization, and
Controller to Preserve Power Sharing Stability of Paralleled Inverters Future Trends—A Review,” IEEE Open J. Ind. Applicat., vol. 2, pp.
in Distributed Generation Microgrids,” IEEE Transactions on Power 93–109, 2021. [Online]. Available: https://ieeexplore.ieee.org/docume
Electronics, vol. 23, no. 6, pp. 2806–2816, Nov. 2008. [Online]. nt/9408354/
Available: http://ieeexplore.ieee.org/document/4696040/ [111] G. Song, B. Cao, and L. Chang, “Review of Grid-forming Inverters in
[96] T. Qoria, “Grid-forming control to achieve a 100% power electronics Support of Power System Operation,” Chinese Journal of Electrical
interfaced power transmission systems,” PhD Thesis, ENSAM, Lille, Engineering, vol. 8, no. 1, pp. 1–15, Mar. 2022. [Online]. Available:
Nov. 2020. https://ieeexplore.ieee.org/document/9751441/
[97] S. D’Arco, J. A. Suul, and O. B. Fosso, “Small-signal modelling and [112] D. B. Rathnayake, M. Akrami, C. Phurailatpam, S. P. Me,
parametric sensitivity of a Virtual Synchronous Machine,” in 2014 S. Hadavi, G. Jayasinghe, S. Zabihi, and B. Bahrani, “Grid
Power Systems Computation Conference. Wrocław, Poland: IEEE, Forming Inverter Modeling, Control, and Applications,” IEEE
Aug. 2014, pp. 1–9. [Online]. Available: http://ieeexplore.ieee.org/do Access, vol. 9, pp. 114 781–114 807, 2021. [Online]. Available:
cument/7038410/ https://ieeexplore.ieee.org/document/9513281/
[98] O. Mo, S. D’Arco, and J. A. Suul, “Evaluation of Virtual Synchronous [113] E. Rokrok, T. Qoria, A. Bruyere, B. Francois, and X. Guillaud,
Machines With Dynamic or Quasi-Stationary Machine Models,” IEEE “Effect of Using PLL-Based Grid-Forming Control on Active Power
Trans. Ind. Electron., vol. 64, no. 7, pp. 5952–5962, Jul. 2017. Dynamics Under Various SCR,” in IECON 2019 - 45th Annual
[Online]. Available: http://ieeexplore.ieee.org/document/7781612/ Conference of the IEEE Industrial Electronics Society. Lisbon,

PSCC 2024
Portugal: IEEE, Oct. 2019, pp. 4799–4804. [Online]. Available: to Very Weak Grids,” IEEE Transactions on Power Systems, vol. 32,
https://ieeexplore.ieee.org/document/8927648/ no. 1, pp. 684–693, Jan. 2017.
[114] L. Zhang, L. Harnefors, and H.-P. Nee, “Power-Synchronization [131] J. F. Morris, K. H. Ahmed, and A. Egea-Àlvarez, “Analysis of Con-
Control of Grid-Connected Voltage-Source Converters,” IEEE troller Bandwidth Interactions for Vector-Controlled VSC Connected to
Transactions on Power Systems, vol. 25, no. 2, pp. 809–820, May 2010. Very Weak AC Grids,” IEEE Journal of Emerging and Selected Topics
[Online]. Available: http://ieeexplore.ieee.org/document/5308285/ in Power Electronics, vol. 9, no. 6, pp. 7343–7354, Dec. 2021.
[115] P. Imgart, A. Narula, M. Bongiorno, M. Beza, and J. R. Svensson, “A [132] X. Wang, M. G. Taul, H. Wu, Y. Liao, F. Blaabjerg, and L. Harnefors,
Cascaded Power Controller for Robust Frequency Ride-Through of “Grid-Synchronization Stability of Converter-Based Resources—An
Grid-Forming Converters,” in 2022 IEEE Energy Conversion Congress Overview,” IEEE Open Journal of Industry Applications, vol. 1, pp.
and Exposition (ECCE). Detroit, MI, USA: IEEE, Oct. 2022, pp. 1–8. 115–134, 2020. [Online]. Available: https://ieeexplore.ieee.org/docu
[Online]. Available: https://ieeexplore.ieee.org/document/9947721/ ment/9181463/
[116] P. Imgart, M. Bongiorno, J. R. Svensson, and M. Beza, “Stability [133] Y. Lamrani, L. Huang, F. Colas, X. Guillaud, F. Blaajberg, C. Cardozo,
Limits and Improved Robustness of Grid-Forming Converters With and T. Prevost, “Grid following converters stability study and control
External Inertia-Emulation Loop,” in 2023 25th European Conference enhancements using an improved test setup,” in 19th International
on Power Electronics and Applications (EPE’23 ECCE Europe). Conference on AC and DC Power Transmission (ACDC 2023), vol.
Aalborg, Denmark: IEEE, Sep. 2023, pp. 1–8. [Online]. Available: 2023, Mar. 2023, pp. 64–69.
https://ieeexplore.ieee.org/document/10264526/ [134] H. Yuan, H. Xin, D. Wu, Z. Li, X. Qin, Y. Zhou, and L. Huang,
[117] “Github L2EP EPMLAB, link to simulation files,” Tech. Rep. “Assessing Maximal Capacity of Grid-Following Converters With
[Online]. Available: https://github.com/l2ep-epmlab/VSC Lib/tree/ma Grid Strength Constraints,” IEEE Transactions on Sustainable Energy,
ster/R2021a/Examples/PSCC vol. 13, no. 4, pp. 2119–2132, Oct. 2022. [Online]. Available:
[118] T. Qoria, F. Gruson, F. Colas, X. Kestelyn, and X. Guillaud, “Current https://ieeexplore.ieee.org/document/9800140/
limiting algorithms and transient stability analysis of grid-forming [135] Z. Zou, J. Tang, X. Wang, Z. Wang, W. Chen, G. Buticchi, and
VSCs,” Electric Power Systems Research, vol. 189, p. 106726, Dec. M. Liserre, “Modeling and Control of A Two-bus System With Grid-
2020. [Online]. Available: https://linkinghub.elsevier.com/retrieve/pii forming and Grid-following Converters,” IEEE Journal of Emerging
/S0378779620305290 and Selected Topics in Power Electronics, pp. 1–1, 2022.
[119] B. Fan, T. Liu, F. Zhao, H. Wu, and X. Wang, “A Review of Current- [136] S. Jiang and G. Konstantinou, “Generalized impedance model and
Limiting Control of Grid-Forming Inverters Under Symmetrical interaction analysis for multiple grid-forming and grid-following
Disturbances,” IEEE Open J. Power Electron., vol. 3, pp. 955–969, converters,” Electric Power Systems Research, vol. 214, p. 108912,
2022. [Online]. Available: https://ieeexplore.ieee.org/document/99733 Jan. 2023. [Online]. Available: https://linkinghub.elsevier.com/retrieve
69/ /pii/S0378779622009634
[120] A. D. Paquette and D. M. Divan, “Virtual Impedance Current Limiting [137] H. Xin, Y. Wang, X. Liu, B. Tang, G. Yu, and L. Huang,
for Inverters in Microgrids With Synchronous Generators,” IEEE “How Many Grid-Forming Converters do We Need? A Perspective
Trans. on Ind. Applicat., vol. 51, no. 2, pp. 1630–1638, Mar. 2015. From Power Grid Strength,” Sep. 2022. [Online]. Available:
[Online]. Available: https://ieeexplore.ieee.org/document/6872529 http://arxiv.org/abs/2209.10465
[121] A. Narula, P. Imgart, M. Bongiorno, M. Beza, J. R. Svensson, [138] P. F. Mayer, M. Gordon, W.-C. Huang, and C. Hardt, “Improving
and A. J.-P. Hasler, “Voltage-based Current Limitation Strategy to grid strength in a wide-area transmission system with grid forming
Preserve Grid-forming Properties Under Severe Grid Disturbances,” inverters,” IET Generation, Transmission & Distribution, vol. n/a, no.
IEEE Open J. Power Electron., pp. 1–13, 2023. [Online]. Available: n/a. [Online]. Available: https://onlinelibrary.wiley.com/doi/abs/10.104
https://ieeexplore.ieee.org/document/10049117/ 9/gtd2.12498
[122] T. Qoria, “Grid-forming control to achieve a 100% power electronics [139] Y. Lamrani, F. Colas, T. Van Cutsem, C. Cardozo, T. Prevost,
interfaced power transmission systems,” Ph.D. dissertation, HESAM and X. Guillaud, “On the Stabilizing Contribution of Different
University, Ecole Nationale Supérieure d’Arts et Métiers, Nov. 2020. Grid-Forming Controls to Power Systems,” Jan. 2024. [Online].
[123] K. V. Kkuni and G. Yang, “Effects of current limit for grid forming Available: https://www.techrxiv.org/users/688223/articles/680612-on-t
converters on transient stability: analysis and solution,” Jun. 2021, he-stabilizing-contribution-of-different-grid-forming-controls-to-pow
number: arXiv:2106.13555 arXiv:2106.13555 [cs, eess]. [Online]. er-systems?commit=98ac9b64fe0bab0ac532c8a1f090f8b5e39e3e0a
Available: http://arxiv.org/abs/2106.13555 [140] ABB, “XLPE Submarine Cable Systems Attachment to XLPE Land
[124] K. G. Saffar, S. Driss, and F. B. Ajaei, “Impacts of Current Limiting Cable Systems - User’s Guide Rev 5.”
on the Transient Stability of the Virtual Synchronous Generator,” IEEE [141] RTE, “Article 8.3.3 – Trame de procédure de contrôle de conformité
Trans. Power Electron., vol. 38, no. 2, pp. 1509–1521, Feb. 2023. pour le raccordement d’une installation de production ou de stockage.”
[Online]. Available: https://ieeexplore.ieee.org/document/9899737/ [142] ——, “Cahier des charges des Capacités Constructives Conditions
[125] L. Harnefors, M. Hinkkanen, U. Riaz, F. M. M. Rahman, and L. Zhang, générales - Unité de stockage non synchrone,” Feb. 2021.
“Robust Analytic Design of Power-Synchronization Control,” IEEE [143] E. Rokrok, T. Qoria, A. Bruyere, B. Francois, H. Zhang,
Trans. Ind. Electron., vol. 66, no. 8, pp. 5810–5819, Aug. 2019. M. Belhaouane, and X. Guillaud, “Impact of grid-forming control on
[Online]. Available: https://ieeexplore.ieee.org/document/8490668/ the internal energy of a modular multilevel converter,” in 2020 22nd
[126] H. Wu and X. Wang, “Design-Oriented Transient Stability Analysis European Conference on Power Electronics and Applications (EPE’20
of PLL-Synchronized Voltage-Source Converters,” IEEE Trans. Power ECCE Europe). Lyon, France: IEEE, Sep. 2020, pp. 1–10. [Online].
Electron., vol. 35, no. 4, pp. 3573–3589, Apr. 2020. [Online]. Available: https://ieeexplore.ieee.org/document/9215646/
Available: https://ieeexplore.ieee.org/document/8820019/ [144] L. Zhao, Z. Jin, and X. Wang, “Analysis and Damping of Low-
[127] M. Chen, D. Zhou, and F. Blaabjerg, “Enhanced Transient Angle Frequency Oscillation for DC-Link Voltage-Synchronized VSCs,”
Stability Control of Grid-Forming Converter Based on Virtual IEEE Transactions on Power Electronics, vol. 38, no. 7, pp.
Synchronous Generator,” IEEE Trans. Ind. Electron., vol. 69, 8177–8189, Jul. 2023. [Online]. Available: https://ieeexplore.ieee.org/
no. 9, pp. 9133–9144, Sep. 2022. [Online]. Available: https: document/10089557/
//ieeexplore.ieee.org/document/9552499/ [145] ——, “Small-Signal Synchronization Stability of Grid-Forming
[128] Y. Laba, A. Bruyère, F. Colas, and X. Guillaud, “Virtual Power-Based Converters With Regulated DC-Link Dynamics,” IEEE Transactions
Technique for Enhancing the Large Voltage Disturbance Stability of on Industrial Electronics, vol. 70, no. 12, pp. 12 399–12 409, Dec. 2023.
HV Grid-Forming Converters,” in 2023 25th European Conference [Online]. Available: https://ieeexplore.ieee.org/document/10012586/
on Power Electronics and Applications (EPE’23 ECCE Europe). [146] N. Hatziargyriou, J. Milanovic, C. Rahmann, V. Ajjarapu, C. Canizares,
Aalborg, Denmark: IEEE, Sep. 2023, pp. 1–8. [Online]. Available: I. Erlich, D. Hill, I. Hiskens, I. Kamwa, B. Pal, P. Pourbeik, J. Sanchez-
https://ieeexplore.ieee.org/document/10264486/ Gasca, A. Stankovic, T. Van Cutsem, V. Vittal, and C. Vournas,
[129] ENTSO-E, “Stability Management in Power Electronics Dominated “Definition and classification of power system stability – revisited &
Systems: A Prerequisite to the Success of the Energy Transition,” Jun. extended,” IEEE Transactions on Power Systems, vol. 36, no. 4, pp.
2022. 3271–3281, 2021.
[130] M. F. M. Arani and Y. A.-R. I. Mohamed, “Analysis and Performance
Enhancement of Vector-Controlled VSC in HVDC Links Connected

PSCC 2024

2024 Promises and Challenges of Grid Forming Transmission System Operator Manufacture and Academic View Points

Uploaded by

Copyright:

Available Formats

2024 Promises and Challenges of Grid Forming Transmission System Operator Manufacture and Academic View Points

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

2024 Promises and Challenges of Grid Forming Transmission System Operator Manufacture and Academic View Points

Uploaded by

Copyright:

Available Formats

Promises and Challenges of Grid Forming:

Transmission System Operator, Manufacturer and

Corresponding author: Xavier Guillaud (xavier.guillaud@centralelille.fr)

23rd Power Systems Computation Conference Paris, France — June 4 – 7, 2024

23rd Power Systems Computation Conference Paris, France — June 4 – 7, 2024

23rd Power Systems Computation Conference Paris, France — June 4 – 7, 2024

23rd Power Systems Computation Conference Paris, France — June 4 – 7, 2024

23rd Power Systems Computation Conference Paris, France — June 4 – 7, 2024

23rd Power Systems Computation Conference Paris, France — June 4 – 7, 2024

23rd Power Systems Computation Conference Paris, France — June 4 – 7, 2024

23rd Power Systems Computation Conference Paris, France — June 4 – 7, 2024

23rd Power Systems Computation Conference Paris, France — June 4 – 7, 2024

23rd Power Systems Computation Conference Paris, France — June 4 – 7, 2024

23rd Power Systems Computation Conference Paris, France — June 4 – 7, 2024

23rd Power Systems Computation Conference Paris, France — June 4 – 7, 2024

23rd Power Systems Computation Conference Paris, France — June 4 – 7, 2024

Fig. 9: Step response in active power for different inertia, H

the DC voltage control, see also [91]. Any such requirement

23rd Power Systems Computation Conference Paris, France — June 4 – 7, 2024

a case where the purpose is to determine that a design is not 1

only technically, but also economically feasible a detailed 0.95

23rd Power Systems Computation Conference Paris, France — June 4 – 7, 2024

D. Virtual impedance impact on large power disturbances

23rd Power Systems Computation Conference Paris, France — June 4 – 7, 2024

23rd Power Systems Computation Conference Paris, France — June 4 – 7, 2024

the literature [110]–[112]. However, to the best of the authors’ Q∗ VSC ∗

phase voltages vm . The dynamics of transistor switching can jXc Ig

be approximated by a delay that takes into account the internal Vg

(b) phasor representation

23rd Power Systems Computation Conference Paris, France — June 4 – 7, 2024

vmd = V ∗ − Xv igq vmq = Xv igd (10) Calculation P,Q

23rd Power Systems Computation Conference Paris, France — June 4 – 7, 2024

ρ P Fig. 21 shows the corresponding blocks to be added. The

23rd Power Systems Computation Conference Paris, France — June 4 – 7, 2024

(b) 2nd formulation

23rd Power Systems Computation Conference Paris, France — June 4 – 7, 2024

Active Power (pu)

consists of specifying the damping. For a desired value ζ̃,

23rd Power Systems Computation Conference Paris, France — June 4 – 7, 2024

Inverse current control loop Current control loop

23rd Power Systems Computation Conference Paris, France — June 4 – 7, 2024

vgdq = vedq + jXg igdq (40)

23rd Power Systems Computation Conference Paris, France — June 4 – 7, 2024

i∗gdq Pvirt = vgd i∗gd + vgq i∗gq

4) Application of Virtual Power Method to VC-GFM: Ex-

To assess the effectiveness of the proposed method, the same 3

process following fault clearance.

been tested to evaluate the virtual power method within the

F. Stabilizing effect of the grid forming control 1

While there is a consensus that GFM controls are expected

23rd Power Systems Computation Conference Paris, France — June 4 – 7, 2024

23rd Power Systems Computation Conference Paris, France — June 4 – 7, 2024

Fig. 36: The Four-VSC benchmark system

G. Extension of findings to a 4 VSC meshed benchmark

23rd Power Systems Computation Conference Paris, France — June 4 – 7, 2024

TABLE IV: 4-VSC system: transformer data

to cope with the maximum production of each park. The cables

23rd Power Systems Computation Conference Paris, France — June 4 – 7, 2024

23rd Power Systems Computation Conference Paris, France — June 4 – 7, 2024