A Grid-Friendly

Plant
By Mahesh Morjaria, Dmitriy Anichkov,
Vladimir Chadliev, and Sachin Soni

The Role of
Utility-Scale
Photovoltaic Plants
in Grid Stability
and Reliability

A
A Recent rapid Photovoltaic (PV)
price reduction has significantly increased the
competitiveness of PV-generated electricity, which
has now reached grid parity in many markets,
plant—© fotosearch, background—image licensed by ingram publishing

i.e., electricity is produced at the same or a

lower price than conventional electricity sources.
Along with other factors stimulating the global
growth of renewable energy, this cost decline has
greatly increased both the number and size of
utility-scale solar plants, especially PVs, that are
deployed on the power grid. The impact on power
systems of integrating PV generation, especially
as it relates to grid reliability and stability, needs
to be addressed.
A typical PV solar generation plant is composed
of multiple individual “generators” connected to the
electrical network via power electronics (inverters),
rather than synchronous machines. The PV plant’s
response to grid system disturbances is not similar
to the inherent electromechanical dynamics of syn-
chronous machines. Through sophisticated control
functions, however, the PV plant is able to contribute
Digital Object Identifier 10.1109/MPE.2014.2302221
actively to grid stability and reliability and operate
Date of publication: 17 April 2014 effectively in the grid.

may/june 2014 1540-7977/14/$31.00©2014IEEE ieee power & energy magazine 87

 We have described a utility-scale, grid-friendly
PV power plant that incorporates advanced capabilities
essential to supporting grid stability and reliability.

A task force under the aegis of the North American Electric In this article, we describe our design of a grid-friendly
Reliability Corporation (NERC) has made several recommen- PV plant, including the development of a plant controller
dations on specific requirements that such variable generation that complies with these requirements, resulting in a PV
plants must meet to provide their share of grid support. These plant that actively contributes to the reliability and stabil-
recommendations address grid requirements such as voltage ity of electrical transmission and distribution system. Field
control and regulation, voltage and frequency fault ride-through, data from First Solar-developed utility-scale PV plants are
reactive and real power control, and frequency response criteria used to illustrate the concepts. The grid-friendly PV plant
in the context of the technical characteristics and physical capa- also includes the ability to ride through specific low and high
bilities of variable-­generation equipment. voltages or low- and high-frequency ranges.

Substation POI Measurements

POI Voltage,
Current, and Frequency
SCADA HMI

Plant Controller
Substation RTU
Set Points Set Points

RTU Data HMI Data

Set Points (Power,

Voltage, or PF)

Plant Network
Inverter Commands

PCS 1 PCS n

DAS/PLC DAS/PLC

Inverter 1 Inverter 2 Inverter 1 Inverter 2

Legend DAS Data Acquisition System HMI Human–Machine Interface

PCS Power Conversion Station PLC Programmable Logic Controller
POI Point of Interconnection RTU Remote Terminal Unit
SCADA Supervisory Control and Data Acquisition System

figure 1. A plant control system and interfaces to other components.

88 ieee power & energy magazine may/june 2014

 A typical PV solar generation plant is composed
of multiple individual “generators” connected to the
electrical network via power electronics.

Power Plant Controller Architecture the real and reactive power necessary to meet the desired
A key component of a grid-friendly PV power plant is a settings at the POI.
plant-level controller. It is designed to regulate real and The plant operator can set an active power curtailment
reactive power output from the PV plant, such that it command to the controller. In this case, the controller calcu-
behaves as a single large generator. While the plant is com- lates and distributes active power curtailment to individual
posed of individual small generators (or, more specifically, inverters. In general, the inverters can be throttled back only
inverters), with each generator performing its own energy to a certain specified level of active power and not any lower
production based on local solar array conditions, the func- without causing the dc voltage to rise beyond its operating
tion of the plant controller is to coordinate the power output range. Therefore, the plant controller dynamically stops and
to provide typical large power-plant features such as active starts inverters as needed to manage the specified active power
power control and voltage regulation (through reactive
power regulation).
The plant controller provides the following plant-level Automatic Voltage Regulation (AVR)
control functions:
✔✔ dynamic voltage and/or power factor regulation of the AVR Control Mode
solar plant at the point of interconnection (POI)
On
✔✔ real power output curtailment of the solar plant
when required so that it does not exceed an operator-­ Off
specified limit
✔✔ ramp-rate controls to ensure that the plant output does Mode Current Setpoint New Setpoint
not ramp up or down faster than a specified ramp-rate Voltage Control 528.0 kV 528.0 kV
limit, to the extent possible
Power Factor Mode 0.98 0.98
✔✔ frequency control to lower plant output in case of over-
frequency situation or increase plant output (if pos- Var Mode 0.2 MVar 0.2 MVar
sible) in case of under-frequency
✔✔ start-up and shut-down control. Confirm Cancel
The plant controller implements plant-level logic and Changes Changes
closed-loop control schemes with real-time commands to
the inverters to achieve fast and reliable regulation. It relies
figure 2. An HMI screen on AVR modes.
on the ability of the inverters to provide a rapid response to
commands from the plant controller. Typically there is one
controller per plant controlling the output at a single high-
voltage bus (referred to as POI). The commands to the plant 10 1.000
controller can be provided through the supervisory control 0
0.995
Power Factor

and data acquisition system (SCADA) human-machine –10

MVars

interface (HMI) or even through other interface equipment, –20 0.990

such as a substation remote terminal unit. –30
Figure 1 illustrates a block-diagram overview of the con- 0.985
–40
trol system and its interfaces to other devices in the plant.
–50 0.980
The power plant controller monitors system-level measure- –4 –2 0 2 4 6 8 10
ments and determines the desired operating conditions of Time (s)
various plant devices to meet the specified targets. It man- Measured Power Factor
ages capacitor banks and/or reactor banks, if present. It has Commanded Power Factor
the critical responsibility of managing all the inverters in the Total Vars
plant, continuously monitoring the conditions of the invert-
ers and commanding them to ensure that they are producing figure 3. Dynamic power-factor regulation.

may/june 2014 ieee power & energy magazine 89

 Active Power Curtailment
Dynamic Voltage Regulation Modes
Through the SCADA HMI, the plant control system can be
set to operate in one of the three modes of automatic voltage
Current SP New SP regulation (AVR): voltage regulation, power-factor regula-
Active Power Limit 290.00 290.00 MW tion, or reactive-power control. The three AVR modes are
illustrated in the SCADA HMI screen shown in Figure 2.
Ramp Up Rate 15.0 15.0 MW/min
Note that the plant can operate in only one of the three oper-
Ramp Down Rate 15.0 15.0 MW/min ating modes at any time.
In the voltage-regulation mode, the controller maintains
the specified voltage set point at the POI by regulating the
Confirm Cancel reactive power produced by the inverters as well as other
Changes Changes
devices such as capacitor banks. In the power-factor regula-
tion mode, the controller maintains the specified power fac-
figure 4. An active power curtailment HMI. tor. The operation of the controller is illustrated in Figure 3,
which shows field data from a PV plant producing about
212 MW of active power at that time.
output limit. It also uses the active power management func- Figure 3 illustrates the response of the plant when the
tion to ensure that the plant output does not exceed the desired power factor set point is changed from 0.98 to 1.0. The
ramp rates, to the extent possible. It cannot, however, always controller commands the inverters to change their reactive
accommodate rapid reduction in irradiance due to cloud cover. power output to meet the new power factor set point, using
a closed-loop control mecha-
nism. The figure illustrates that
the inverters respond very rap-
Power Curtailment at Different Levels idly. Within a few seconds (<4 s)
100 100
the new set point is achieved in a
95 95 closed-loop control mode. More
specifically, the rise time to reach
Set Point (MW)

90 Inverters Are Turned Down/Off 90

Power (MW)

to Curtail Output 90% of steady-state value shown

85 85 in the figure is about 3.2 s.

80
Ramp Rates Between
80 Active Power
Set Point Changes Management
75 Are Controlled 75
Through the SCADA HMI, the
70 70 controller can also be set to con-
0 2 4 6 8 10 12 14 16 18 20 trol the active power output of the
Time (min) plant. When the control system
detects that the active power at POI
figure 5. Power curtailment at different levels. exceeds the specified set point, it

Output of Each Block Plant Output Power

Increase in Additional Energy Captured

Potential
Power Limit Power
Power

Power

Plant
Output Limit
Curtailed

Block Time

figure 6. The impact of cloud passage under curtailment.

90 ieee power & energy magazine may/june 2014

calculates and sends the commands
for each inverter individually to Stopping and Starting One Block (30 MW)
lower its output to achieve the 90 90
desired set point, using a closed-
85 Low Plant Stop Command 85
loop control mechanism. Note that,
Irradiance
in some cases the plant controller 80 80

Commanded (MW)
will turn off certain inverters to

Power (MW)
achieve this desired set point since 75 75
Inverters Are
the output of each inverter cannot 70 Started in 70
be lowered below a certain thresh- Inverters Are Sequence
Turned Off in
old without causing a high dc volt- 65
Sequence;
65
age operating condition. Ramp Rates
60 60
Figure 4 illustrates field data Are Controlled
Plant Start Command
from a PV plant operating at 55 55
around 90-MW power. The cur- 0 1 2 3 4 5 6 7 8 9 10
tailment limit is initially changed Time (min)
from 100 to 82.5 MW. The plant Source: Agua Caliete PV Plant in Yuma, Arizona, USA
controller turns down the invert- 13 March 2012 ~90 MW Online
ers (and turns off some of them if
required) to achieve the new set figure 7. Plant shut-down and start-up controls.
point. Note that the turndown of
power is gradual to meet the spec-
ified ramp-rate limit.
Frequency Droop Parameters
The curtailment limit is reduced
again to around 75 MW, and the MW
controller responds as expected.
When the limit is raised, the con-
troller adjusts the output of the Droop
inverters to increase the total plant Low
Deadband
High Limits Decrease Active
output. Finally, when the limit is Limits
Power in Case of
raised to 100 MW, the plant is no Hz
High Frequency
longer curtailed since the plant is Deadband
producing less than the limit. Droop
In all the control actions, the con-
troller’s command to each inverter is
unique, given the specific conditions
each inverter is experiencing. For
example, when the plant is under Frequency Deadband 0.036 Hz
curtailment, the plant controller Frequency Droop 5.0 %
can release the power limit of indi- High Limit 61.500 Hz
vidual inverters if the total output of
Low Limit 58.500 Hz
the plant starts falling below the set
point. So in case of a cloud passage, Active Power High Limit 290.0 MW
which results in a reduction of the Active Power Low Limit 10.0 MW
output of a part of the plant, the con-
troller can make the adjustment to
increase the output of other inverters
that are not impacted.
The plant-level control strategy figure 8. The frequency droop function.
results in a capture of energy from
inverters that would have been otherwise unnecessarily cur- that are not impacted by the cloud cover to dynamically increase
tailed. This concept is illustrated in Figure 5. The left side of their previously curtailed limit. Since the total potential power
the figure represents the reduction in power output of some of of the plant is greater than the specified plant output limit (as
the inverters (grouped in blocks for illustration purposes) due illustrated on the right-hand side of Figure 5), the plant is able to
to partial cloud cover. The controller commands other inverters output the total power all the way to the limit.

may/june 2014 ieee power & energy magazine 91

 The grid-friendly PV plant also includes
the ability to ride through specific low and high voltages
or low- and high-frequency ranges.

An illustration of an additional function related to active required ramp rate. As mentioned earlier, some inverters
power management is shown in Figure 6, which shows the are turned down while others are shut down. Note that the
reaction of the plant when it is shut down and started up. control is quite effective even with moderately varying irra-
The field data in this case illustrates when one block (30 diance conditions.
MW), which is under the control of the plant controller, Figure 7 also illustrates the plant start command resulting in
is commanded to shut down. The active power manage- the controller gradually increasing the plant output by adjusting
ment function reduces plant output while maintaining the the inverters’ output and turning on the inverters in sequence.

REPC_A Vt REEC_B Vt REGC_A

Vreg ′
Iqcmd Iqcmd Iq
Vref Plant-Level Qext Q
Qref V/Q Control Control
Current
Qbranch Generator Network
Limit
Pref Model Solution
′
Ipcmd Logic Ipcmd Ip
Pbranch Plant-Level Pref P
Fref P Control Control
Freg
Pqflag

figure 9. An overall model structure of a solar PV plant.

Ibranch 1 VcompFlag +
Vreg Vreg – (Rc + jXc) × Ibranch 1 – ∑ Qmax
0 1 + sTfltr
dbd emax
+ 1 Ki 1 + sTft
+ RefFlag Kp + Qext
Qbranch Kc ∑ S 1 + sTiv
0
1 – emin
∑ Freeze State If
1 + sTfltr Qmin Vreg < Vfrz
+
Qref

Plant_Pref Pmax
Freg_flag
+ femax 0
1 – Kig 1 Pref
Pbranch ∑ Kpg +
1 + sTp 0 S 1 + sTlag 1
+ femin
fdbd1, fdbd2 Ddn + Pmin
Freg ∑ ∑
–
+ Dup +
Freg_ref
0

figure 10. An REPC_A model block diagram.

92 ieee power & energy magazine may/june 2014

Frequency Droop
Control Interconnection Station Collector Equivalent
The control system also pro- Transmission Transformer System Pad-Mounted
vides frequency droop control to Line Equivalent Transformer
handle unusual grid situations.
PV
For example, in case of above- kV, R, X, B
normal frequency, the control- kV, R, X, B kV, MVA, R, X kV, MVA, R, X PV System
ler will reduce the active power Equivalent
POI or Connection to Plant-Level
of the plant, as illustrated in Figure 8. Transmission System Reactive
If the plant is under curtailment, Compensation
the power can also be increased
if the below-normal frequency is figure 11. A single-machine equivalent load flow representation.
detected. Note that all the param-
eters illustrated in the figure are
configurable and are shown here for illustrative purposes only. Power Plant Controller Model
The plant controller model REPC_A is used when plant-level
Fault Ride-Through Capability control of active and/or reactive power is desired. A func-
A significant benefit of utility-scale PV systems that incorpo- tional block diagram for REPC_A model is shown in ­Figure
rate fault ride-through capability is that they do not trip off 10. The model incorporates the following functionalities
during system disturbances but continue to provide power Renewable Energy Modeling Task Force (WECC-REMTF):
when the grid needs it. The ability to ride through specific low ✔✔ Closed-loop voltage regulation at a user-designated
and high voltages or low- and high-frequency ranges is being bus with the provisions for line drop compensation,
designed effectively into all modern variable generators. Most voltage droop response, and a user-settable dead-band
utility-scale inverters have this capability. With proper design on the voltage error signal.
practices, the PV plant is engineered to ensure that all com- ✔✔ Closed-loop reactive power regulation on a user-des-
ponents in addition to inverters also have the ability to ride ignated branch with a user-settable dead-band on the
through short-term grid events. reactive power error signal.
✔✔ A plant-level governor response signal derived from
Modeling and Validation frequency deviation from the nominal user-designated
A general structure of a PV power plant recommended by branch. The frequency droop response can be applied
the Western Electricity Coordinating Council (WECC) to active power flow on a user-designated branch. Fre-
is shown in Figure 9. For the simplification of a solar PV quency droop control is capable of being activated in
plant, the dynamics related to the dc side of the inverter both over and under frequency conditions. The fre-
(PV array dynamics, inverter dc link, and voltage regulator) quency deviation applied to the droop gain can be
are ignored. The overall model structure shown in Figure 9 subject to a user-settable dead-band.
represents the complete PV plant and consists of an equiva-
lent generator model “REGC_A” to provide current injec- Test Model Description
tions into the network solution, an electrical control model This section demonstrates the outcome from a set of field tests
“REEC_B” for local active and reactive power control, and performed at a PV plant when operating at 90-MW active
a centralized plant controller model “REPC_A” to allow for power output. The test plant equivalent model is config-
plant-level active and reactive power control. ured as shown in Figure 11, where multiple medium-voltage

3
1
1 2 4 5
12-MVar Cap Bank
90.0 89.75 89.58 89.54 89.54 89.54 89.54
1.000

1.000

1 90.00 89.75 89.58

9.6 1
9.63 8.35 8.35 13.20 13.20 9.58 9.57 9.55 9.54
PV Plant 0.328 37.74 525.0 525.0 Grid
1.011 1.007 1.050 1.050

34.62
1.003

figure 12. A PSLF equivalent model for a plant under test.

may/june 2014 ieee power & energy magazine 93

 In the test system, a 12-MVar
Field Test Versus Simulation for 12-MVar Capacitor Bank Switching capacitor bank, located at the 34.5-
15.0 kV collector bus, is engaged as an
Reactive Power Command (MVar)

12.0 external stimulus. F ­ igure 13 illus-

9.0 Cap Bank Switching trates that, when the capacitor bank
6.0 Fast Initial Response Qcmd_actual is engaged at “t = 0” s relative time,
Qcmd_simulation the power plant controller sends
3.0
0.0
control signals to each individual
–3.0
inverter to immediately address
this switching event, and eventually
–6.0
Slower Plant Level Response
inverters reactive power contribu-
–9.0 Capacitor Bank Switched Online
tion to the plant drops. First Solar’s
–12.0
–1.10 0.40 1.90 3.40 4.90 6.40 7.90 power plant controller reactive
Relative Time (s) power command (Qcmd_actual)
distributed to the individual invert-
figure 13. The Qcmd response field test versus simulation. ers is shown in green in Figure 13.
Both field measured reactive power
command (Qcmd_actual) and sim-
ulated model command (Qcmd_
Field Test Versus Model Simulation for 12-MVar Capacitor Bank Switching simulation) are shown in Figure
15.0 13. The figure illustrates the initial
12.0 quick response by the PV plant,
which took almost 100 ms, followed
Reactive Power (MVar)

9.0 Inverter Vars (Measured)

6.0 Inverter Vars (Simulated) by the dominance of the power plant
Cap Bank Switching controller in the order of seconds, to
3.0
0.0
maintain post-disturbance stability.
–3.0
Figure 14 shows the detailed
plant-reactive power response to
–6.0
Capacitor Bank Switched Online capacitor switching. This figure
–9.0
illustrates that the simulation model
–12.0
–5.90 –4.40 –2.90 –1.40 0.10 1.60 3.10 4.60 6.10 7.60 performance adequately mimics
Relative Time (s) actual plant behavior. The response
closely matches actual behavior,
figure 14. Inverter VARs during a 12-MVar cap bank switching test. with a difference immediately fol-
lowing the switching operation that
could be attributed to the lower sam-
feeders are equivalenced at one collector, a 34.5-kV bus at pling rate in the field measurement than in PSLF simulation.
the low side of the plant substation transformer. Load-flow
model equivalencing is performed using the method sug- Daily Grid Operation
gested in the WECC modeling guide for solar PV plants Another grid integration concern, especially for a grid
Modeling and Validation Work Group/Technical Studies operator, is daily load balancing. Solar generation is a type
Subcommittee. of variable power generation that is not fully dispatch-
This plant uses SMA 800-kVA inverters controlled by able since the energy source is influenced by the presence
First Solar’s power plant controller. The power plant control- of solar radiation and by atmospheric conditions. Reliable
ler allows for coordination of all online inverters for plant- power-system operation requires the continuous balance of
level voltage regulation at the POI, located at the 500-kV supply and demand. To successfully manage a variable gen-
substation bus. The purpose of testing at this plant is for the eration source like solar, grid operators treat PV generation
model validation of First Solar’s controller against WECC’s as “negative” load, and they utilize short-term forecasts to
proposed power plant controller. Data captured from the field schedule and dispatch compensatory controllable resources.
tests were filtered and then compared to simulation results The operators are already familiar with a certain amount of
obtained from a plant model built in GE’s Positive Sequence variability and uncertainty, particularly with system load (or
Load Flow (PSLF) simulation software. A single-line dia- demand). They have successfully utilized a variety of tools
gram of the plant model in PSLF is shown in Figure 12. such as generator and transmission flexibility, ancillary ser-
A 12-MVar capacitor bank switching test is performed to vices, and demand-side resources to achieve reliable sys-
examine the plant controller performance. tem operation. The growing sophistication and accuracy of

94 ieee power & energy magazine may/june 2014

 In all the control actions, the controller’s command
to each inverter is unique, given the specific conditions
each inverter is experiencing.

short-term solar generation forecasts is facilitating continued can reduce the impact of cloud passage and increase over-
efficient and reliable system operations. By providing PV all yield. Plant-model validation against measured field data
plants that can support the forecasting needs of grid opera- demonstrates that the WECC-proposed model for a solar PV
tors, this variable resource can address the load-balancing plant is adequately capturing actual plant behavior.
needs of grid operators. An advantage of solar in many mar-
kets is that its peak generation coincides with higher load For Further Reading
demand, making it more a valuable generation resource. R. Binz. (2012. Apr.). Practicing risk aware electricity regu-
lation: What every state regulators need to know. A Ceres
Grid Flexibility and Power Report. [Online]. Available: http://www.ceres.org/resources/
Systems Resource Planning reports/practicing-risk-aware-electricity-regulation
As the proportion of variable generation increases in the M. G. Lauby, M. Ahlstorm, D. L. Brooks, S. Beuning, J.
overall generation portfolio, another integration concern is Caspray, W. Grant, B. Kirby, M. Milligan, M. O’Malley, M.
that greater grid flexibility is required to provide the neces- Patel, R. Piwko, P. Pourbeik, D. Shirmohammadi, and J. C.
sary power backup when the variable generation resource is Smith., “Balancing act,” IEEE Power Energy Mag., vol. 9,
not adequate to meet the demand. This dictates an increased no. 6, pp. 75–85, Nov./Dec. 2011.
use of conventional resources that are able to respond and Modeling and Validation Work Group/Technical Stud-
ramp up more quickly and a reduced use of inflexible gen- ies Subcommittee. (2011, Jan.). [Online]. Available: https://
eration resources. A recent analysis points out “that planning www.wecc.biz/committees/StandingCommittees/PCC/TSS/
the lowest-cost, lowest-risk investment route aligns with a MVWG/Shared%20Documents/MVWG%20Approved%20
low-carbon future. From a risk management standpoint, Do cu ment s/ W ECC%20PV%20Pla nt%20Power%20
diversifying utility portfolios today by expanding invest- Flow%20Modeling%20Guide.pdf
ment in clean energy and energy efficiency makes sense R. Piwko, A. Ellis, R Walling, B. Zavadil, D. Jacobson,
regardless of how and when carbon controls come into play. M. Patel, D. Brooks, B. Nelson, J. MacDowell, C. Barker,
Placing too many bets on the conventional basket of genera- E. V. Engeln, E Seymour, L. Casey, E. John, W. Peter, W. Lasher, C.
tion technologies is the highest risk route.” Loutan, L. Dangelmaier, E. Seymour, and M. Behnke. (2012,
Mar.). Special reliability assessment: Interconnection require-
Summary ments for variable generation. NERC. [Online]. Available:
We have described a utility-scale, grid-friendly PV power http://www.nerc.com/comm/PC/Integration%20of%20Vari-
plant that incorporates advanced capabilities essential to able%20Generation%20Task%20Force%20IVGT/Sub%20
supporting grid stability and reliability. It includes features Teams/Interconnection/NERC_IVGTF_1-3_DRAFT.pdf
such as voltage regulation, active power controls, ramp-rate WECC Renewable Energy Modeling Task Force. (2012,
controls, fault ride through, and frequency control. These Sept.). [Online]. Available: https://www.wecc.biz. https://
capabilities provide the intrinsic benefits of reliable plant www.wecc.biz/committees/StandingCommittees/PCC/TSS/
operation in the grid, which in turn results in additional MVWG/Shared%20Documents/MVWG%20Approved%20
plant yield and potential additional revenue. Such capa- Documents/WECC%20Solar%20PV%20Dynamic%20
bilities are essential for successful the deployment of large- Model%20Specification%20-%20September%202012.pdf
scale PV plants.
A key component of such a grid-friendly plant is a plant- Biographies
level controller specifically engineered to regulate real and Mahesh Morjaria is with First Solar, Tempe, Arizona.
reactive power output of the solar facility such that it behaves Dmitriy Anichkov is with First Solar, Tempe, Arizona.
as a single large conventional generator, although within Vladimir Chadliev is with First Solar, Tempe, Arizona.
the limits dictated by the intermittent nature of the solar Sachin Soni is with Arizona State University, Tempe,
resource. In cases where the plant output is constrained but Arizona.
p&e
the plant has additional generation capability, this controller 

may/june 2014 ieee power & energy magazine 95