Electrical Power and Energy Systems 127 (2021) 106679

Contents lists available at ScienceDirect

International Journal of Electrical Power and Energy Systems

journal homepage: [Link]/locate/ijepes

Over-frequency support in large-scale photovoltaic power plants using

non-conventional control architectures
Queralt Madorell-Batlle , Eduard Bullich-Massagué *, Marc Cheah-Mañé , Oriol Gomis-Bellmunt
Centre d’Innovació Tecnològica en Convertidors Estàtics i Accionaments (CITCEA-UPC), ETS d’Enginyeria Industrial de Barcelona, Universitat Politècnica de Catalunya,
Av. Diagonal, 647, Pl. 2., 08028 Barcelona, Spain

A R T I C L E I N F O A B S T R A C T

Keywords: Large scale photovoltaic power plants must provide a frequency regulation service, which is defined in the grid
PV power plant codes. This service has commonly required a response time between 15 and 30 s. But some countries are now
Grid code introducing more strict regulations and requiring response times below 2 s. The typical centralized control ar­
Control architecture
chitecture of photovoltaic power plants for frequency regulation can present undesired oscillatory responses (or
Frequency
Frequency regulation
even become unstable) when tuning the controller to achieve these small time response requirements. The
present article proposes an alternative solution based on a hierarchical control architecture. In the proposed
solution, inverter controllers apply a local frequency regulation action and the central controller corrects active
power errors at the point of connection, which can be caused by power losses or lack of irradiance in some
inverters. Simulation models are used to study and test the response of this new control approach. The proposed
hierarchical control architecture is compared with a fully centralized and a fully decentralized archirectures.
Results show that the hierarchical control architecture is not only capable to obtain a fast and accurate response,
but also is robust against communication failures. The proposed hierarchical control architecture advantages
could be extrapolated to other services. So, further research is proposed to confirm this hypothesis.

1. Introduction according to the size of the power plant [7]. LS-PVPPs can be included as
type C and D power-generating modules (minimum size of 5 and 10 MW
Large-scale photovoltaic power plants1 (LS-PVPPs) must include a respectively). Typically, the primary frequency regulation service has
control system to fulfill the interconnection requirements specified by required a response time between 15 and 30 s [6–11]. But now, several
the transmission system operators (TSOs). These requirements, defined countries such as UK, Ireland or Australia are starting to introduce a new
in the so-called grid codes, usually specify a set of grid services that LS- service that, despite having different names, refers to the same concept,
PVPPs must provide [3]. As more and more non-synchronous wind and i.e. fast frequency response, in which the regulation must respond be­
solar power plants are being interconnected to the grid, frequency tween 0.5 and 2 s [5,12].
control becomes more crucial and complex [4], specially in small re­ The control of LS-PVPPs has traditionally been designed in a
gions such as islands with weak interconnection [5]. Accordingly, pri­ centralized structure [13,14]. While this centralized architecture has
mary frequency regulation requirements are commonly found in the grid been effective during the last years, it can present several limitations in
codes. These requirements, are tipically based on a droop characteristic terms of time response [15,16]. In [15], a detailed dynamic analysis of a
where the active power have to be increased or reduced based on the LS-PVPP with a central control architecture has been conducted
grid frequency as shown in Fig. 1. A number of countries are already concluding that tunning the controller to obtain a response time of 1-2 s
considering specific frequency regulation requirements for LS-PVPPs at the point of interconnection can present undesired oscillatory modes,
[6–11]. In Europe, ENTSO-E defines a grid code called Requirements even if the inverter power response is very fast, i.e. 100 ms. In [16] a
for Grid Connection of Generators, where frequency support is defined typical centralized frequency control implementation is presented and

* Corresponding author.
E-mail addresses: [Link]@[Link] (Q. Madorell-Batlle), [Link]@[Link] (E. Bullich-Massagué), [Link]@[Link] (M. Cheah-Mañé),
[Link]@[Link] (O. Gomis-Bellmunt).
1
Despite there is not a clear definition about what is considered a LS-PVPPs, it is well accepted that they are power plants from several MW to GW scale [1]. For
example, the National Renewable Energy Laboratory (NREL) sets the threshold at 5 MW [2].

[Link]
Received 3 July 2020; Received in revised form 9 October 2020; Accepted 28 November 2020
Available online 29 December 2020
0142-0615/© 2020 Elsevier Ltd. All rights reserved.
Q. Madorell-Batlle et al. International Journal of Electrical Power and Energy Systems 127 (2021) 106679

challenges in terms of time response are explained, which are typically 2. Proposed control architecture for over-frequency support
between 3 to 10 s. The study done in [16] achieves a total response time
of 2 s using a centralized approach. 2.1. State of the art: current photovoltaic (PV) plant control architecture
Considering (i) this new service that has been recenlty introduced in
some grid codes requires a time response of less than 2 s and (ii) the Fig. 2 shows a general scheme of the centralized control architecture,
limitations found in [15,16] for controlling the active power in a time which is an industry standard for LS-PVPPs. In such power plants tens or
scale of less than 2 s using fully central controllers, alternative solutions hundreds of PV arrays with a rated power from 100 kW to 2 MW are
to the centralized control architecture can be explored. In addition, the interconnected, through PV inverters and 3-winding transformers, to an
centralized control relies on a communication network that must func­ internal Medium Voltage (MV) AC network called collection grid. The
tion continuously. Although the complete or partial loss of communi­ main transmission network is connected to the collection grid through a
cations is a rare event, its huge impact in centralized architectures may main feeder and a High Voltage - Medium Voltage (HV-MV) transformer.
lead to malfunction of the LS-PVPPs [17]. This also motivates the need Grid codes specify services to be provided at the point of connection
to explore alternative control architectures. with the transmission network, which is called point of common
Different control architectures can be implemented in power sys­ coupling (PCC). Therefore, a coordination between PV inverters is
tems, but as explained, in LS-PVPPs mainly the centralized approach has required. This is achieved with local controllers in the inverters that
been considered. In the field of microgrids, three control architectures follow active and reactive power setpoints and a central controller that
have commonly been considered, namely centralized, decentralized and monitors and controls the power exchange at the PCC. [22].
distributed [18,19]. In the centralized architecture, as in LSPVPPs, all The central controller first computes the required setpoints at the
the information is collected in a central unit, where the setpoints are PCC, P*PCC and Q*PCC , according to the TSO requested grid code re­
processed and sent back to each controllable device using direct quirements. Then, these setpoints are compared with the power mea­
communication links. In the decentralized approach, each unit perfoms surements at the PCC, PPCC and QPCC [13], and a PI controller computes
its local control without a direct communication link but using the the aggregated setpoint that is sent to the PV inverters, P*tot . As the rated
power lines to communicate by varying the voltage and frequency. power of the LS-PVPP is different to the rated power of PV inverters, and
Finally, the distributed approach includes communication links between PV inverters can also have different power rating, a dispatcher is in
controllable units but lacks of a central controller. According to [18],
charge of transforming this P*tot to a p.u. system, dividing the aggegated
centralized approaches require high computational efforts and
setpoint by the nominal power of the PV plant (PN ) and sending this
communication needs while a fully decentralized architecture does not
signal (P*INV [p.u]) to all the PV inverters.
offer a proper coordination level (this is important in LS-PVPPs to
Over-frequency regulation is the service of interest in this study,
comply with the grid code requirements at the point of common
which results in a reduction of the total active power generated. Fig. 3a
coupling, PCC). Thus, despite the distributed control architecture could
shows the general schematic of a centralized architecture for this ser­
be a solution, a hierarchical approach combining the centralized and
vice. As can be seen, the frequency control is implemented at the PPC
decentralized architectures is the most common solution [20,21].
while the inverters only receive power reference signals from the central
In this context, the present paper proposes a novel control archi­
controller. The P-f droop block represents the implementation of the P-f
tecture for LS-PVPPs based on a hierarchical control approach to provide
droop characteristic (see Fig. 1) according to grid code requirements.
over-frequency regulation. This alternative LS-PVPP control architec­
ture aims to face the previous described challenges for the over- The power reference P* is equal to a power curtailment reference P*max
frequency support service. Thus, the present study contributes to i) from the TSO if the frequency control is not active, i.e. the frequency is
introduce the concept of hierarchical control architectures in LS-PVPPs within the deadband of the P-f droop characteristic. When the frequency
as a reliable option and to ii) analyze and compare the proposed hier­ control is active, i.e. the frequency exceeds the deadband, P* is calcu­
archical architecture with the traditional and decentralized options for lated based on the P-f characteristic and considering P*max and the
over-frequency regulation. measured power PPCC and frequency.

Fig. 1. General scheme of a frequency droop characteristic.

2
Q. Madorell-Batlle et al. International Journal of Electrical Power and Energy Systems 127 (2021) 106679

Fig. 2. General scheme of a typical LS-PVPP.

Fig. 3. Existing control architectures for over-frequency regulation.

The communication network between the central controller and the frequency droop characteristic is implemented at each PV inverter,
inverters is essential to ensure a proper operation. In case of commu­ based on their local measurements. In particular the inverter power
nication failure, the plant would be forced to operate blindly, which reference P*INV is the output of the P-f droop block. Thus, the output
could result in a breach of the grid code. In addition, a fast response of active power setpoint is controlled through the active power controller
the frequency control mode is required when a destabilizing event in the of the PV inverter and the effect of cascaded PI controllers is avoided.
grid is detected. Thus, pressure to reduce the response time of the PV Communication requirements are minimal for this architecture, which
plant is increasing. However, the communication network has inherent results in a reliable option. Also, faster responses are achieved when a
delays as well as two cascaded PI controllers (the central PI controller change of the setpoint is detected. However, measurements at the PCC
plus the local PI controller) that might interact, which can become a are not included in the control system. This lack of information results in
limiting factor to reduce the response time. a low performance of the plant, as neither power losses nor power de­
In addition, a decentralized control architecture can be considered in viations caused by the loss of an inverter or the reduction of irradiance
LS-PVPPs to reduce the response time. The decentralized control ar­ (e.g. due to a cloud) are corrected. This blind operation leads to a power
chitecture can be found in microgrid applications, but is less common in injection below the maximum allowed value and might result in a
LS-PVPPs, where it is mainly used for controlling fast current injection mismatch with grid code requirements.
during faults that requires a response in the range of hundreds of mil­
liseconds [23]. The decentralized architecture is based on separated
local controls at inverter level. System operation only requires the in­ 2.2. Proposed hierarchical control architecture for over-frequency support
verter’s local measurements and TSO setpoints. Fig. 3b shows the over-
frequency regulation based on a decentralized architecture. The Thus, the present paper proposes a hierarchical control architecture
to take the advantages of both, the centralized and decentralized ap

3
Q. Madorell-Batlle et al. International Journal of Electrical Power and Energy Systems 127 (2021) 106679

proaches. The scheme of this proposed hierarchical control architecture available power [24]. This available power have been obtained by i)
for over-frequency support is shown in Fig. 4. In this scheme, each local downloading irradiance data at 1 Hz resolution, corresponding to Oahu,
inverter controller provides a power reference P*INV,0 [p.u], which is Hawaii, from the National Renewable Energy Laboratory (NREL’s)
calculated considering the P-f droop characteristics and the power database [25], ii) converting the irradiance data to available active
curtailment reference P*max . Also, the central controller corrects the er­ power generation using the model developed in [26], where a high ac­
curacy have been obtained and iii) applying a delay between each
rors at the PCC with a compensation power reference ΔP*INV [p.u]. Then,
converter profile to simulate the effect of clouds passing above the PV
the main control is performed by local controllers at inverter level, but it
plant. The parameters of this case study are shown in Table 1, where τPPC
is supported by a central controller. As the control function is performed
refers to the sample time of the central controller, τcom is the commu­
locally, communication delays as well as the effect of having cascaded PI
nication delay or the sending delay as shown in Fig. 5 (it is defined
controllers are avoided. Thus, a fast response is achieved. The central
constant in this paper), τinv is the time constant of the inverter response
control will have the same dynamics as in a central control architecture,
(first order function) and the rest of the parameters are the P-f droop
but only applied to the errors caused by the decentralized control ar­
characteristics. Also, the losses are modelled as a 5% of the generated
chitecture (by the local controllers). In particular, the central controller
power PINV .
monitors the PCC, receives TSO setpoints and compensates any control
All the simulations show the PV plant operating under curtailment
function mismatch caused by internal PV plant power losses or by
mode when an over-frequency event is detected. This event activates the
external events, such as loss of an inverter or reduction of irradiance (e.
frequency controller following the P-f droop characteristic, which is
g. due to a cloud). This is done thanks to a central PI controller. By
executed as an active power reduction. For the central controller, the
implementing this additional closed loop of control, the PV plant output
new setpoint is set as a percentage reduction respect to the last active
can be maximized to the desired value. In case of central communication
power measured value at the PCC, PPPC . For the local controllers, the
network failure, the local controls of the hierarchical structure ensure
reduction is done respect the last PINV value. The P − f characteristic is
that the PV plant can keep operation. Commercial inverters are not yet
different for each grid code. This study implements a generic active
prepared to receive a ΔP*INV [p.u]. as an action of control. Modifications of power reduction in response to the over-frequency event, defined in
the actual inverters would be required to implement this non- Table 1. Then, four scenarios are exposed in order to evaluate the over-
conventional control architecture. Nevertheless, the potential advan­ frequency regulation performance of each architecture. Since the P-f
tages of this proposed architectures will be shown in this paper through characteristic is different for each grid code, in this case study the
a set of simulations. requirement from South Africa are considered as example [27].

3. Validation of the control architecture

3.2. Response time
3.1. Case study
The typical primary frequency response currently requires a response
The performance of the centralized, decentralized and the proposed time around 15 and 30 s at the PCC, but future services, such as the Fast
hierarchical control architectures are validated employing a common PV Frequency Response (FFR) in the UK, will require a faster response [12].
plant model. This is based on a LS-PVPP with 20 PV arrays of 1 MW Then, improving the current response time is desirable. The response
including their associated converters. Then, the model is simplified as 2 time can be affected by limitations set by the characteristics of the
PV arrays and 2 converters, where each one represents the aggregation control architecture, the PI controller parameters, the inverters, the
of 10 PV arrays with a total of 10 MW. Fig. 5 shows the simplified model measurement devices and communication delays, specific to each PV
with the centralized and hierarchical architectures. The decentralized plant installation. Hence, the interest in this case study is to compare the
architecture is not shown since is based on the same model as the hi­ response time of these control architectures. The response time of the
erarchical architecture without the PPC. In the centralized control the inverters and the sampling time of the PPC are relevant parameters to
power reference of the inverter model is the output of the PPC after the consider. Fig. 6 shows two simulations to understand the impact of these
sending delays, P*INV [p.u], while in the hierarchical control is the sum of parameters on the system response. Table 2 shows the selected values for
the local inverter reference, P*INV,0 [p.u] , and a correction from the central each simulation as well as the control parameters of each architecture.
controller, ΔP*INV [p.u]. In both cases, the output of the inverter model is The PI parameters of the PPC, in both the centralized and the hierar­
chical control structure, are selected using the Control System Tuner
the active power generated by the inverter, PINV . The inverters are
Toolbox of Matlab to obtain the fastest dynamic response with a small
represented as a first order function, with the output saturated by the
overshoot (a maximum of 0.5 %).

Fig. 4. Proposed hierarchical control architecture for over-frequency regulation.

4
Q. Madorell-Batlle et al. International Journal of Electrical Power and Energy Systems 127 (2021) 106679

Fig. 5. Simplified control architecture models.

Table 1
Parameters used for the over-frequency simulations.
Parameter τPPC τcom τinv Deadband Δfmax PPnom,i Droop constant

Value 100–500 ms 20 ms 100–500 ms 50–50.5 Hz 1.5 Hz 10 MW 0.467 MW/Hz

Fig. 6. Droop response time using different configuration parameters.

5
Q. Madorell-Batlle et al. International Journal of Electrical Power and Energy Systems 127 (2021) 106679

Table 2
Comparison of the two simulations.
Simulation Inverter time constant (τinv ) PPC sampling time (τPPC ) Architcture Central controller System time constant

KP KI

1 100 ms 100 ms Centralized 0 2.63 389.6 ms

Hierarchical 0.23 5.47 121.5 ms
Decentralized – – 119.9 ms

2 500 ms 500 ms Centralized 0.15 0.90 1016.9 ms

Hierarchical 0.23 1.09 521.3 ms
Decentralized – – 519.9 ms

Fig. 6a shows the results when small time constants are considered. hierarchical control is capable to provide a general solution to the pre­
At 500 s, the setpoint P* drops as an over-frequency is detected. sented problem. When the central controller of the hierarchical archi­
Compared to the centralized architecture (PC ), the decentralized (PD ) tecture is turned off, the control becomes local. In general, when only
and hierarchical (PH ) architectures respond faster. This is due to the fact the local control is operative, power losses are not compensated. How­
that the centralized architecture is governed by 2 cascaded PI controllers ever, during droop operation the reduction in active power is performed
(the local PI controller plus the central PI controller), while the decen­ with respect to the last local value sent by each inverter at 240 s, so the
tralized and hierarchical architectures are governed only by the local PI central controller does not act to compensate power losses for over-
controller. Note that in the hierarchical architecture, the central frequency response, unless the available power PAV is lower than the
controller only applies a correction in the setpoint. On the other hand, it setpoint P* . According to Fig. 7, the hierarchical PV plant output (PH ) is
can be observed that the decentralized architecture does not reach the able to reach the desired setpoint with no interruption. If the same so­
setpoint due to the power losses within the power plant, while the lution is applied to the centralized architecture (PC ) at 400 s, the
centralized and hierarchical architectures are capable to deal with this centralized PV plant shuts down.
issue thanks to the central controller. This will be explained in the
following sections in detail. As shown, the hierarchical architecture is
3.4. Losses compensation
taking the advantages of both i) the fast response of the decentralized
architecture and ii) the capability to compensate the power losses of the
Power losses are present in power systems as a result of unwanted
centralized architecture. Fig. 6b shows the results when time constants
energy dissipation effects. They can be as high as the 5% of the gener­
are increased. This second simulation validates that the comparison
ation [28], which results in a big loss of power for LS-PVPPs. In order to
between control architectures does not depend on the inverter
maximize profits, the system may inject the maximum power allowed by
characteristics.
the TSO into the grid. Thus, compensation of losses provided by the
control system is desired. Due to power losses, the aggregated PINV
3.3. Loss of communications output of the inverters is always higher than the PCC measurement. For
the centralized and hierarchical architectures, power losses are cor­
A network of communications is required by the central controller rected by the central controller, so the PCC measurement reaches the
and in some situations can fail. In addition, the central controller itself setpoint when enough power is available. If there is no central
could also experiment a failure. The loss of PCC measurement, the loss of controller, the aggregated inverter output equals the setpoint. Thus, in
the central controller or the loss of the communication network can lead the decentralized architecture the PCC measurement is lower than the
to the plant malfunction. This is often prevented by duplicating the setpoint.
equipment (i.e. with redundancy). However, it is important to note that Note that at the moment that an over-frequency is detected, the latest
communication failure can still occur. In the event of a communication PCC measurement is stored by the central controller and the latest
error detection, inverters could have different modes of configuration, e. inverter output PINV is stored by each local controller. Then, the P-f
g. they can either maintain the last measured value, take a specified droop is performed with respect to these values. The implications of this
value or zero as constant value. There is no standard mode of configu­ operative are clear, as the hierarchical and centralized architectures
ration and it might differ between PV plants. Thus, this paper assumes manage to achieve the active power output during curtailment, while
that the action of control is set at zero when an interruption is detected, the decentralized architecture achieves a lower value. Thus, Fig. 8a
i.e. the central controller is turned off. Fig. 7 shows the P* setpoint shows that the droop setpoint of the hierarchical and centralized ar­
dropping at 240 s as a result of an over-frequency detection. At 400 s the chitectures (P*C− H ) is maximized, which is higher than the droop setpoint
loss of communication between the PPC and the inverters is detected of the decentralized architecture (P*D ). The central controller present in
and the central controller is turned off. The decentralized architecture is the hierarchical and centralized architectures is essential to provide the
not shown because does not have a central PI controller. Only the maximum power allowed at all times. Through the PCC measurement,
the system is able to compute the necessary power that must be injected
into the grid at every given moment and maximize it. As a result, Fig. 8b
shows that the hierarchical (PH ) and centralized (PC ) PV plant output is
equal and maximized, while the decentralized (PD ) PV plant output is
lower when enough power is available.

3.5. PV plant under partial irradiance limitation

Losses due to temporary shading are an important issue in PV gen­

eration. They are caused by clouds passing through or sharp elements
present in the surroundings of the PV plant. The consequences of a shade
can be mitigated by increasing the setpoint sent by the central PI
Fig. 7. Loss of communications during droop mode. controller to all of the inverters, which results in an increase of the

6
Q. Madorell-Batlle et al. International Journal of Electrical Power and Energy Systems 127 (2021) 106679

Fig. 8. Power output during curtailment and droop operation.

output power. Thus, the desired production can be achieved or maxi­

Table 3
mized. The decentralized architecture has no central controller and
Comparison of the control architectures based on the over-frequency support
cannot compensate these losses.
service.
To understand this issue for the decentralized architecture, Fig. 9a
shows the setpoint for each group of inverters and their available power Centralized Decentralized Hierarchical

(PAV1 and PAV2 ). Also, the total available power is also shown as PAV . Control level Central Local Central and
In spite of the availability from the aggregated power, Fig. 9b shows local
Need of communications Important Not necessary Mainly for
that from 430 to 480 s the output power of the decentralized PV plant
central control
does not only suffer from power losses but also from the effect that the Correction of power losses, Yes No Yes
lack of irradiance has on the inverters (PAV2 is lower than P*INV of each irradiance limitations or
inverter). Then, the frequency droop activated at 480 s is applied with inverter failure
Response time Slow Fast Fast
respect a much lower value of active power than for the centralized and Operation without Not possible Possible Possible
hierarchical case, leading to a lower output power during over- communications
frequency support. As Fig. 9b shows, the decentralized architecture Modification in actual No No Yes
performance (PD ) is lower than in the other architectures, both during inverters
Complexity of Low Low Medium
curtailment and frequency droop. On the other hand, the centralized
implementation
(PC ) and hierarchical (PH ) architectures are shown to perform similar Compliance with grid code Ensured Might not be Ensured
results during the event. The maximum power allowed by the TSO is ensured
reached during curtailment as a result of the central controller action. State of the art Commonly Less used Not currently
used used

3.6. Comparison of control architectures characteristics

ensures the continuance of service, while the conventional architecture
The control architectures have been presented and analyzed in cannot support the operation under the same fault. Finally, the hierar­
different scenarios. Table 3 compares the characteristics and capabilities chical architecture can compensate power losses, irradiance limitation
of each control architecture. or inverter failure as the centralized option.
The presented advantages show the potential of the hierarchical
4. Conclusion architecture, which might be more relevant for large scale photovoltaic
power plants as future grid requirements become more strict. This paper
This paper has proposed a hierarchical control architecture for over- has only shown the proposed architecture characteristics for over-
frequency support in large scale photovoltaic power plants. The sug­ frequency regulation, but these conclusions could be also applied for
gested architecture has proven to take the advantages of both the other services. Therefore, further research is required to validate the
centralized and the decentralized architectures. In particular, the hier­ implementation and advantages of this proposed control architecture.
archical architecture offers a similar response to the decentralized ar­
chitecture, i.e. a faster response than the conventional centralized CRediT authorship contribution statement
architecture. Thus, it is an interesting option for future severe re­
quirements related to response time during an over-frequency event. In Queralt Madorell-Batlle: Methodology, Software, Validation,
addition, the hierarchical architecture offers a response as robust as the Formal analysis, Investigation, Writing - original draft, Writing - review
decentralized architecture in front of communication failures, since it & editing. Eduard Bullich-Massagué: Conceptualization,

Fig. 9. Effect of irradiance limitation on some inverters.

7
Q. Madorell-Batlle et al. International Journal of Electrical Power and Energy Systems 127 (2021) 106679

Methodology, Formal analysis, Investigation, Writing - review & editing. [8] Energynet Dk. Technical regulation 3.2.2 for PV power plants above 11 kW; 2016.
[9] Gevorgian V, Booth S. Review of PREPA Technical Requirements for
Marc Cheah-Mañé: Conceptualization, Methodology, Formal analysis,
Interconnecting Wind and Solar Generation, ser. NREL/TP; 2013.
Investigation, Writing - review & editing. Oriol Gomis-Bellmunt: [10] Berndt H, Hermann M, Kreye H, Reinisch R, Scherer U, Vanzetta J.
Conceptualization, Writing - review & editing, Supervision, Project TransmissionCode-2007. Network and System Rules of the German Transmission
administration. System Operators, Verband der Netzbetreiber, Berlin, Tech. Rep.; 2007.
[11] Acts of the national regulatory authorit for ene R G Y domain. Official gazzete of
Romania, Parti I, No. 312/30.V.2013; 2013.
Declaration of Competing Interest [12] National Grid Electricity System Operator, The Grid Code; 2019.
[13] Bullich-Massagué E, Ferrer-San-josé R, Aragüés-Peñalba M, Serrano-Salamanca L,
Pacheco-Navas C, Gomis-Bellmunt O. Power plant control in large-scale
The authors declare that they have no known competing financial photovoltaic plants: design, implementation and validation in a 9.4 MW
interests or personal relationships that could have appeared to influence photovoltaic plant. IET Renew Power Gen 2016; 10 (1):50–62.
the work reported in this paper. [14] Morjaria M, Anichkov D, Chadliev V, Soni S. A grid-friendly plant: the role of
utility-scale photovoltaic plants in grid stability and reliability. IEEE Power Energy
Mag 2014;12(3):87–95.
Acknowledgment [15] Montero-Casinello J, Cheah-Mañé M, Prieto-Araujo E, Gomis-Bellmunt O.
Interaction analysis of large-scale PV power plants considering communication
delays and operational points. In: 9th international workshop on integration of
The work was funded by PVTOOL, which is a project supported solar power into power systems, 15–16 Oct. 2019, Dublir, Ireland; 2019.
under the umbrella of [Link] Cofund by the Ministry of [16] Long J, Qu L, Zhang S, Li L. Frequency control strategy and test technology of
Economy and Competitiveness, the CDTI and the Swedish Energy photovoltaic power plant. In: 2019 IEEE 3rd conference on energy internet and
energy system integration (EI2); 2019. p. 1695–8.
Agency. [Link] is supported by the European Commission
[17] Bosman LB, Leon-Salas WD, Hutzel W, Soto EA. PV system predictive maintenance:
within the EU Framework Programme for Research and Innovation challenges, current approaches, and opportunities. Energies 2020;13(6):1398.
HORIZON 2020 (Cofund ERA-NET Action, no 691664). [18] Olivares DE, Mehrizi-Sani A, Etemadi AH, Cañizares CA, Iravani R, Kazerani M,
Eduard Bullich-Massagué and Marc Cheah-Mañé are lecturers of the et al. Trends in microgrid control. IEEE Trans Smart Grid 2014; 5(4):1905–19.
[19] Dragićević T, Lu X, Vasquez JC, Guerrero JM. DC microgrids – part i: A review of
Serra Húnter programme. The work of Oriol Gomis-Bellmunt is also control strategies and stabilization techniques. IEEE Trans Power Electron 2016;31
supported by the ICREA Academia program. (7):4876–91.
[20] Guerrero JM, Vasquez JC, Matas J, de Vicuna LG, Castilla M. Hierarchical control
of droop-controlled ac and dc microgrids–a general approach toward
References standardization. IEEE Trans Indust Electron 2011;58(1):158–72.
[21] Bidram A, Davoudi A. Hierarchical structure of microgrids control system. IEEE
[1] Komoto K, Kurokawa K, Nishimura T, Kato K, Otani K, Ito M, et al. IEA PVPS task 8: Trans Smart Grid 2012;3(4):1963–76.
Project proposals on very large scale photovoltaic power generation (VLS-PV) [22] Cabrera-Tobar A, Bullich-Massagué E, Aragüés-Peñalba M, Gomis-Bellmunt O.
systems in deserts. In: 2006 IEEE 4th world conference on photovoltaic energy Topologies for large scale photovoltaic power plants. 2016. p. 309–19.
conference, vol. 2; 2006, .p 2359–62. [23] Afshari E, Moradi GR, Rahimi R, Farhangi B, Yang Y, Blaabjerg F, et al. Control
[2] Mendelsohn M, Lowder T, Canavan B, N.R.E.L. (U.S.). Utility-scale concentrating strategy for three-phase grid-connected pv inverters enabling current limitation
solar power and photovoltaics projects: a technology and market overview, ser. under unbalanced faults. IEEE Trans Indust Electron 2017;64(11):8908–18.
NREL/TP. National Renewable Energy Laboratory; 2012. [24] Bullich-Massagué E, Aragüés-Peñalba M, Sumper A, Boix-Aragones O. Active
[3] Cabrera-Tobar A, Bullich-Massagué E, Aragüés-Peñalba M, Gomis-Bellmunt O. power control in a hybrid PV-storage power plant for frequency support. Sol
Review of advanced grid requirements for the integration of large scale Energy 2017; 144:49–62.
photovoltaic power plants in the transmission system. Renew. Sustain. Energy Rev [25] Sengupta M, Andreas A. oahu solar measurement grid (1-year archive), nrel report
2016; 62: 971–87. no. da-5500-56506; 2010. Available: [Link]
[4] Eirgrid and SONI, DS3: Frequency Control Workstream 2015; 2015. [26] Marcos J, Marroyo L, Lorenzo E, Alvira D, Izco E. From irradiance to output power
[5] Meng L, Zafar J, Khadem SK, Collinson A, Murchie KC, Coffele F, et al. Fast fluctuations: the pv plant as a low pass filter. Prog Photovolt: Res Appl 2011;19(1):
frequency response from energy storage systems–a review of grid standards, 505–10.
projects and technical issues. IEEE Trans Smart Grid 2020;11(2):1566–81. [27] Africa NERS. Gird Connection Code for Renewable Power Plants (RPPs) Connected
[6] Bullich-Massagué E, Cifuentes-García F-J, Glenny-Crende I, Cheah-Mañé M, to the Electricity Transmission System (TS) or the Distribution System (DS) in
Aragüés-Peñalba M, et al. A review of energy storage technologies for large scale South Africa; 2019.
photovoltaic power plants. Appl. Energy 2020; 274:115213. Available: http:// [28] Malvoni M, Leggieri A, Maggiotto G, Congedo PM, De Giorgi MG. Long term
[Link]/science/article/pii/S030626192030725X. performance, losses and efficiency analysis of a 960 kWP photovoltaic system in
[7] Commission regulation (EU) 2016/631 of 14 April 2016 establishing a network the Mediterranean climate. Energy Convers Manage 2017;145(August):169–81.
code on requirements for grid connection of generators. Official Journal of the
European Union; 2016.