See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/317385936

Providing Primary Frequency Response from Photovoltaic Power Plants

Conference Paper · May 2017

CITATIONS READS
13 4,103

4 authors, including:

Pouyan Pourbeik Sachin Soni

Power and Energy, Analysis, Consulting and Education, PLLC First Solar
118 PUBLICATIONS 4,844 CITATIONS 10 PUBLICATIONS 188 CITATIONS

SEE PROFILE SEE PROFILE

Anish Gaikwad
Electric Power Research Institute
53 PUBLICATIONS 671 CITATIONS

SEE PROFILE

All content following this page was uploaded by Sachin Soni on 08 June 2017.

The user has requested enhancement of the downloaded file.

See discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/317385936

Providing Primary Frequency Response from

Photovoltaic Power Plants

Conference Paper · May 2017

CITATIONS READS

0 7

4 authors, including:

Pouyan Pourbeik
Power and Energy, Analysis, Consulting and Education, PLLC
95 PUBLICATIONS 1,675 CITATIONS

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

IEC 61400-27 View project

All content following this page was uploaded by Pouyan Pourbeik on 07 June 2017.

The user has requested enhancement of the downloaded file. All in-text references underlined in blue are added to the original document
and are linked to publications on ResearchGate, letting you access and read them immediately.
 DUBLIN 2017
Paper Reference Number: 022
http : //www.cigre.org
CIGRE Symposium 2017
Experiencing the Future Power System ….. Today

Providing Primary Frequency Response from Photovoltaic Power Plants

P. POURBEIK1 S. SONI A. GAIKWAD V.CHADLIEV

PEACE First Solar EPRI First Solar
USA USA USA USA

SUMMARY
Photovoltaic (PV) power generation, which is typically connected to the electric grid through
power electronic inverters, is rapidly growing worldwide as a significant source of energy in
many regions. Within North America, several recent installations of PV power plants have
exceeded 500 MW of nameplate capacity in California2.
With the ever-increasing penetration of renewable resources, both wind and PV, there is
increasing interest by utilities and transmission system operators with respect to the ability of
such resources to provide primary frequency response. Frequency response and stability is
one of the key aspects of bulk power system reliability and technical performance.
In this paper, we will discuss how large PV power plants can provide primary frequency
control at the point of interconnection, and illustrate this through presenting results from
actual field tests of PV plants showing their response to frequency deviations. Furthermore,
simulations, using the recently developed 2nd generation generic renewable energy system
models [1], will be provided showing that the performance of the PV plant can be adequately
modeled, for transmission planning studies, for such primary frequency response control.
Also, a few examples are shown, with validated simulations, of the voltage and reactive
response of the PV plants at the point of interconnection.
KEYWORDS
Solar PV, Model Validation, Frequency Response

1
Corresponding author: P. Pourbeik, Power and Energy, Analysis, Consulting and Education, PLLC
[email protected] or [email protected]
2 http://www.firstsolar.com/en/about-us/projects/topaz-solar-farm

1
1.0 INTRODUCTION
With the increasing penetration of renewable energy resources into the bulk electric power
system, there has been a growing concern and interest in being able to rely on these resources
to provide primary frequency response. When a large generator trips due to an unpredictable
forced outage, the sudden imbalance between electrical power generation and demand (load)
on the bulk electric system (BES) will result in a frequency decline, since all of the large
interconnected power systems around the world are still predominantly served by
synchronous generation. This decline in frequency is characterized by three stages: (i) the
initial inertial response of the synchronous generation that helps to restrain the rate of decline
of system frequency, (ii) the primary frequency response of the generators that have
headroom to maneuver, and a functioning turbine-governor, that helps to restrain the decline
in frequency and establish a new equilibrium point, and (iii) automatic generation control,
which through control action gradually restores the system frequency back to its nominal
value [2], [3]. This is shown diagrammatically in Figure 1.

Figure 1: Typical power system frequency response to a sudden loss of generation (©IEEE
2013, from reference [3]).
Therefore, as the penetration of renewable generation increases, and displaces the
conventional synchronous generators on the BES there has been significant effort to identify
ways in which the inverter based wind and PV generation technologies can provide the same
system technical performance ancillary services, such as frequency response [4]. There is a
significant amount of literature already published in the past several years that has shown the
ability of wind turbine generators to provide frequency response (some examples are [5], [6],
[7] and [8]). However, there are fewer examples illustrating the capabilities of PV generation
to also provide the same type of response. One recent and comprehensive study on this
subject is [9].
The key aspects of this paper a twofold:

2
 1. Showing results from actual field tests of two large utility scale solar PV plants that
clearly show the capability of such plants to provide primary frequency response, and
a concise description of how this is achieved.
2. Illustrating the ability of the newly developed 2nd generation generic renewable energy
system (RES) models [1] to adequately capture the performance of the equipment for
the purposes of large scale power system studies. This is the key new result presented
in this paper, since hitherto these generic models had not been tested in simulating the
primary frequency response of PV power plants.
2.0 PROVIDING PRIMARY FREQUENCY RESPONSE WITH PV GENERATION
As briefly explained in the introduction, and illustrated in Figure 1, conventional synchronous
generation provided frequency response in three stages [2]: (i) inertial response, (ii) primary
frequency response (or sometimes called governor response), and (iii) automatic-generation
control. Inertial response is an inherent physical response of synchronous generation to a
sudden imbalance in generation and load. For example, for the sudden loss of a large
generating unit on the system, in the immediate instant that follows the loss of generation the
load (or demand) does not instantaneous change, and thus due to the laws of physics (not
controls) all the remaining synchronous generators will respond electrically, in proportion to
their electrical rating, to provide a portion of the lost power to continue to supply the load.
This is an inherent electrical response. Following this essentially instantaneous response, the
extra electrical power delivered by each synchronous generator will need to be supplied from
a mechanical source. This additional mechanical energy is drawn out of the stored rotational
kinetic energy in the inertia of the generator shafts – the turbine-generator rotating mechanical
shaft. This is the inherent inertial response of synchronous generators, and is an inherent
physical response due to the laws of physics and not a controlled response. Also, due to this
inertial response as mechanical energy is drawn out of the shafts of all the synchronous
generators, all the synchronous generators on the system slowdown in rotational speed, and
thus the system frequency declines. This is the initial rate of decline in system frequency
shown in Figure 1. The combined effective total “inertia” of the system thus determines this
initial rate of decline in frequency. If there were no controls on any of the generators, then the
system frequency would continue to decline until under-frequency load shedding starts to
shed load and eventually other protection may also start to trip generation, and in extreme
cases lead to a blackout. This is the very consequence which needs to be avoided. Thus, on
any power system an adequate number of generating facilities must carry some reserve – that
is, some of the generators will not be making the maximum amount of power they are capable
of generating, and will keep some head-room for being able to generate additional power, if
requested to do so by their controls, in response to a frequency decline. Then when the initial
decline in frequency is detected, the turbine-governor controls on these units act to increase
the megawatt (MW) output of the unit in proportion to the drop in frequency – this is called
turbine-governor droop-response or frequency response [2], [3]3. Thus, a new equilibrium is
achieved as some generating facilities increase their output and thus the system wide balance
between generation and load is restored – this is the second phase of the response shown in
Figure 1, where the frequency turns around from its nadir and settles down at a new
equilibrium. Finally, automatic-generation control (AGC) acts in the third stage to bring the
frequency back up to its nominal value [2], [9]. AGC is a centralized controller that acts on
selected generation with head-room.

3 It should be noted that load damping (the reduction of load with frequency decline due to the inherent
frequency dependency of some loads) also contributes to frequency response [2].

3
In the case of power-electronic interfaced renewable generation technologies, such as type 3
and 4 wind turbine generators and PV generation, the above inherent physical responses are
none existent. This is because the response of the generator is completely governed by the
controls of the power-electronic interface. Recent technological advances by many vendors
now provide the capability for providing frequency response from wind and PV generation.
A modern utility scale solar PV plant equipped with advanced grid-friendly control functions
is able to contribute to some of the ancillary services and help improve the system reliability
[15]. Examples of, and detailed discussions on, the ability of wind generation to provide
frequency response can be found in many recent publications [7], [10], [11], [12]. Also, as
shown in reference [1], that the 2nd generation generic RES models can adequately simulate
the observed response of wind turbine generators to provide frequency response. In the case
of wind turbine generators, “emulated” or “synthetic” inertial response can also be provided
[13].
Photovoltaic (PV) generation can also provide primary frequency response, much the same
way as wind generation. The mechanism, however, is slightly different in the energy source
and how it is utilized. The active power output of a PV power plant is determined at any
point in time by the available incident solar irradiance. The solar irradiance, falling upon the
solar PV array, creates direct-current in the photo-cell arrays due to the photovoltaic effect.
This direct-current is then converted to alternating-current, in synchronism with the grid
frequency, by a voltage-source converter (VSC) that interfaces the photo-cell array to the grid.
All such power inverters will have an associated so-called maximum power-point tracking
(MPPT) algorithm. The aim of the MPPT algorithm is to vary the apparent impedance seen
by the PV array in order to drive the PV array to operate at its maximum power point (shown
diagrammatically in Figure 2). The reason why a MPPT algorithm is needed, is that the
maximum power point (MPP) of the PV array is constantly changing, because the voltage-
current (VI) characteristic (shown in Figure 2) constantly changes based on the incident solar
irradiance and the ambient temperature. As irradiance increases the VI curve moves up, and
as the ambient conditions change and the array heats up, the VI curve moves down. Thus, the
optimal point, at which maximum power is transferred from the PV array (i.e. the optimal V-I
operating point, shown in Figure 2) changes and the MPPT algorithm follows this point.

Maximum Power Point

Power / Current

Voltage

Figure 2: Maximum power point of a PV array.

4
Now consider, if the MPPT algorithm is used to determine the MPP, but the controls are set
such that the inverter would deliberately operate at a point that was 10% below the MPP, then
that would mean that at any given time the PV array would be producing 10% less power
(MW) than its full potential. Thus, in this way 10% of the current power being produced
could always be kept in reserve. Clearly, since the resource is variable (solar irradiance) the
absolute amount of MW reserve is also variable, however, the relative amount (10%) is
always constant. Now by introducing a droop control function into the PV array controls we
can provide primary frequency response both up and down – this is shown diagrammatically
in Figure 3. This is the approach that is taken to allow PV plants to provide primary
frequency response. It should be noted that providing primary frequency response for an
over-frequency event can always be done, even when the PV is operating at the MPP. This is
because for an over-frequency event the PV plant needs to reduce its output. However, to be
able to respond to an under-frequency event, to increase the PV plant output, the PV arrays
must be operated at a sub-optimal power point to keep some of the potential power generation
in reserve. Furthermore, in order to provide any primary frequency response, the PV plant
must employ the appropriate control functions to provide droop response. This is not
necessarily a standard feature and must be requested from the equipment manufacturer and
has associated costs. Also, there are obviously opportunity costs for the plant owner
associated with operating a PV plant at sub-optimal power points to keep MWs in reserve.
Such considerations are outside the scope of this paper, but must be considered.
Power

Deadband

Droop
Frequency

Droop

Figure 3: Droop function

3.0 FIELD TESTS FOR PV PRIMARY FREQUENCY RESPONSE
Field tests were performed on two different PV plants. One of these plants is a 23.5 MW
plant located in the Electric Reliability Council of Texas (ERCOT) region in the United States
(US), and the other is a 250 MW plant located in Western Interconnection region in the US.
Both are connected to the bulk electric power system, at extra high voltage (EHV)
transmission level. The test performed in the field where to inject the following signals into
the plant level controller at the PV power plant:
 A step change in the reactive power (MVAr) reference, with the plant in constant-Q
control mode – this is to illustrate the response rate of the reactive power control at the
point of interconnection.

5
  A step change in the voltage reference of the plant level controller, which is
controlling the voltage at the point of interconnection. This was done with the plant in
voltage control mode.
 Play-in a frequency signal at power plant controller by deliberately over-riding the
frequency reference. By doing this, the inverters are exposed to a signal as if a real
frequency event has happened at the grid and provide the desired frequency response.
The played-in signal can be a real-time frequency signal measured from a system wide
frequency event or a synthetic frequency signal mimicking a large frequency dip or
rise. This was done with the plant in droop-control and holding 10% of its current
megawatts in reserve.
The third item above deserves a little further explanation. The primary frequency response
capability of utility scale solar PV plants has been less explored, especially for under-
frequency regulation services. Utility scale solar PV plants can demonstrate frequency
response for over-frequency events easily through its governor-like control systems by
curtailing the active power on the directive of the system operator. The purpose of this test
was to demonstrate the potential of utility scale solar PV plants to provide both up-and down
regulation during under-and over-frequency event through its governor-like control systems.
Figure 4 shows the actual system frequency of the ERCOT grid in the US during an under-
frequency event that happened as a result of a loss in generation. To perform the third test
above, this measured signal (from an actual event, or others similar to this) were digitally
played back into the controls of the PV plant, in lieu of the actual grid frequency, thus making
it appear to the controls that an actual under-frequency event has occurred, and to thus see if
the plant provides frequency response.

Figure 4: Under-frequency event on the ERCOT grid of the US.

4.0 MODEL VALIDATION AND SIMULATION OF THE PV PLANT PRIMARY
FREQUENCY RESPONSE
23.5 MW Plant
The tests described above were performed on a 23.5 MW PV plant, located in ERCOT.
Namely, a Q-reference step-test, a voltage-reference step test and a frequency response test.
This plant was then modeled as a single aggregated equivalent generator, with an equivalent
collector system model using the 2nd generation generic RES models [13]. Then these same

6
tests were simulated and the measured and simulated responses were compared. The
simulations were done in the EPRI PPPD tool [14]. The resultant models were also then
implemented in a commercial software platform and the simulations also performed there,
giving the same results. For the sake of brevity, here only the simulations with PPPD are
shown. The results for these tests are shown in Figures 5, 6 and 7.
Note the clear response of the plant to an emulated frequency event in Figure 5, and the fact
that the simulation model can quite adequately simulate the plants behavior. For these tests
the droop constant for the frequency control was deliberately set to a large value, i.e. much
large droop gain, as compared to typical fossil fuel plants (e.g. for a typical fossil fuel plant
the droop settings are 4 – 5%, that is a gain of 20 to 25). Thus, if kept in reserve, note that the
PV plant is readily able to provide a response of nearly 20% of its nameplate rating in less
than 3 to 4 seconds. This is a very fast and large response, typically not feasible on large
thermal power plants. The reason simply is that the PV plant has no mechanically moving
parts, the response is driven solely by the electrical and thermal ratings of the equipment, the
response time of the power-electronic controls and the amount of MW kept in reserve. In the
case of a large thermal plant, such as a steam-turbine, there are many other factors related to
the mechanical side that limit the rate at which the plant can respond.
250 MW Plant
Similar tests were also conducted on a 250 MW PV plant in the Western Interconnection of
the US. In this case, tests were done for both over- and under-frequency control, by playing
into the controls a simulated over-frequency, as well as an under-frequency event. Actual
under-frequency events do occur on the system form time-to-time. It can take several months
of monitoring to collect such data. Over-frequency events are much rarer. Even if such
events do occur, it is important that plant have some means of collecting the event data
through installed devices like Phasor Measurement Units (PMUs). This 250 MW plant under
study is equipped with a PMU, but actual event data was not available during testing.
Therefore, a synthetic frequency event signal demonstrating a large change in frequency was
used. The comparisons of the measured and simulated response are shown in Figures 8 and 9.

Figure 5: Frequency response test, showing simulation versus measured response in MW and
MVAr of the plant. Plant in frequency responsive control mode, with head-room. The 23.5
MW PV Plant.

7
Figure 6: Q-reference step test response, showing simulation versus measured response in
MW and MVAr of the plant. Plant in constant Q-cotnrol mode. The 23.5 MW PV Plant.

Figure 7: V-reference step test (step down) response, showing simulation versus measured
response in MW and MVAr of the plant. Plant in constant voltage control mode. The 23.5
MW PV Plant.

8
Figure 8: Under-frequency response for the 250 MW PV plant, showing simulation versus
measured response in MW and MVAr of the plant. Plant in frequency responsive control
mode, with head-room.

Figure 9: Over-frequency response for the 250 MW PV plant, showing simulation versus
measured response in MW and MVAr of the plant. Plant in frequency responsive control
mode, with head-room.

5.0 CONCLUSION
In this paper, the basis on which frequency response capability can be provided by a PV
power plant, connected to the bulk electric power system, has been described. Field tests on
two large PV plants have shown this to be successful and effective. Furthermore, it has been
illustrated here that the newly develop generic 2nd generation renewable energy system
models, developed in the Western Electricity Coordinating Council [1], can adequately
simulate the frequency response capability of PV power plants, as well as some of the other
advanced functionalities such as reactive power and voltage control at the point of
interconnection.
Finally, it should be noted, as shown here, that the frequency response capability of PV plants
is significant faster than conventional fossil fuel plants, since the response is driven by the
control of power-electronics. There are no mechanical moving parts, as there are in steam-
turbine, gas-turbines and other similar technologies. Thus, the actual response rate can be
quite fast and very predictable. There are of course economic consequences to operating a
renewable power plant, such as a PV plant, at sub-optimal conditions in order to keep

9
 megawatts of incident solar energy in reserve for under-frequency response. Such
considerations must not be neglected, but are outside of the scope of this paper.
REFERENCES
[1] P. Pourbeik, J. Sanchez-Gasca, J. Senthil, J. Weber, P. Zadehkhost, Y. Kazachkov, S. Tacke and J. Wen,
“Generic Dynamic Models for Modeling Wind Power Plants and other Renewable Technologies in Large
Scale Power System Studies”, To be published in IEEE Trans. on Energy Conversion, 2017; DOI:
10.1109/TEC.2016.2639050
[2] IEEE Task Force on Large Interconnected Power Systems Response to Generation Governing,
Interconnected Power System Response to Generation Governing: Present Practice and Outstanding
Concerns, IEEE Special Publication 07TP180, May 2007.
[3] IEEE Task Force on Turbine-Governor Modeling, Dynamic Models for Turbine-Governors in Power
System Studies, IEEE Technical Report PES-TR1, January 2013.
[4] Advanced Grid-Friendly Controls Demonstration Project for Utility-Scale PV Power Plants, January
2016. http://www.nrel.gov/docs/fy16osti/65368.pdf
[5] Western Wind and Solar Integration Study (NREL)
http://www.nrel.gov/docs/fy10osti/47434.pdf
[6] Western Wind and Solar Integration Study – Phase 3 (NREL)
http://www.nrel.gov/docs/fy15osti/62906.pdf
[7] NREL Report TP-5D00-60574, Active Power Controls from Wind Power: Bridging the Gaps, January
2014 http://www.nrel.gov/docs/fy14osti/60574.pdf
[8] Eastern Frequency Response Study by GE http://www.nrel.gov/docs/fy13osti/58077.pdf
[9] CalISO Report, Using Renewables to Operate a Low-Carbon Grid: Demonstration of Advanced
Reliability Services from a Utility-Scale Solar PV Plant. January 11th , 2017 Available at:
http://www.caiso.com/Documents/UsingRenewablesToOperateLow-CarbonGrid.pdf
[10] V. Singvi, P. Pourbeik, N. Bhatt, D. Brooks, Y. Zhang, V. Gevorgian, E. Ela, and K. Clark, “Impact of
Wind Active Power Control Strategies on Frequency Response of an Interconnection”, Proceedings of the
IEEE PES General Meeting, July 2013, Vancouver, Canada.
[11] N. Miller, K. Clark, R. Walling, “WindINERTIA: Controlled Inertial Response from GE Wind Turbine
Generators", 45th Annual Minnesota Power Systems Conference, Minneapolis, MN, November 2009.
[12] N. W. Miller, M. Shao and S. Venkataraman, “Impact of Frequency Responsive Wind Plant Controls on
Grid Performance”, CIGRE Session 2012, paper C4-113.
[13] N. W. Miller, K. Clark and M. Shao, “Frequency responsive wind plant controls: Impacts on grid
performance”, Proceedings of the IEEE PES GM 2011.
[14] Power Plant Parameter Derivation (PPPD) Tool Version 9.0, October 2017, EPRI.
http://www.epri.com/abstracts/Pages/ProductAbstract.aspx?ProductId=000000003002008339
[15] M. Morjaria, D. Anichkov, V. Chadliev and S. Soni, "A Grid Friendly plant: The role of utility-scale
photovoltaic plants in grid stability and reliability", IEEE Power Energy Mag., vol 12, no.3, pp.87-95,
May/June 2014.

10

View publication stats