Photovoltaic Plant and BESS Integration

Photovoltaic Plant and Battery Energy
Storage System Integration at NREL’s

Flatirons Campus
Vahan Gevorgian, Przemyslaw Koralewicz,
Shahil Shah, Emanuel Mendiola, Robb Wallen,
and Hugo Villegas Pico
National Renewable Energy Laboratory
With collaboration from First Solar
NREL is a national laboratory of the U.S. Department of Energy Technical Report

Office of Energy Efficiency & Renewable Energy NREL/TP-5D00-81104
Operated by the Alliance for Sustainable Energy, LLC February 2022
This report is available at no cost from the National Renewable Energy
Laboratory (NREL) at www.nrel.gov/publications.
Contract No. DE-AC36-08GO28308

Photovoltaic Plant and Battery Energy
Storage System Integration at NREL’s
Flatirons Campus
Vahan Gevorgian, Przemyslaw Koralewicz,
Shahil Shah, Emanuel Mendiola, Robb Wallen,
and Hugo Villegas Pico
National Renewable Energy Laboratory
With collaboration from First Solar
Suggested Citation
Gevorgian, Vahan, Przemyslaw Koralewicz, Shahil Shah, Emanuel Mendiola, Robb
Wallen, and Hugo Villegas Pico. 2022. Photovoltaic Plant and Battery Energy Storage
System Integration at NREL’s Flatirons Campus. Golden, CO: National Renewable
Energy Laboratory. NREL/TP-5D00-81104. https://www.nrel.gov/docs/fy21osti/81104.pdf.
NREL is a national laboratory of the U.S. Department of Energy Technical Report

Office of Energy Efficiency & Renewable Energy NREL/TP-5D00-81104
Operated by the Alliance for Sustainable Energy, LLC February 2022
This report is available at no cost from the National Renewable Energy National Renewable Energy Laboratory
Laboratory (NREL) at www.nrel.gov/publications. 15013 Denver West Parkway
Golden, CO 80401
Contract No. DE-AC36-08GO28308 303-275-3000 • www.nrel.gov
NOTICE
This work was authored in part by the National Renewable Energy Laboratory, operated by Alliance for Sustainable
Energy, LLC, for the U.S. Department of Energy (DOE) under Contract No. DE-AC36-08GO28308. Funding
provided by U.S. Department of Energy Office of Energy Efficiency and Renewable Energy Solar Energy
Technologies Office Agreement Number 31797. The views expressed herein do not necessarily represent the views
of the DOE or the U.S. Government.
This report is available at no cost from the National Renewable

Energy Laboratory (NREL) at www.nrel.gov/publications.
U.S. Department of Energy (DOE) reports produced after 1991

and a growing number of pre-1991 documents are available
free via www.OSTI.gov.
Cover Photos by Dennis Schroeder: (clockwise, left to right) NREL 51934, NREL 45897, NREL 42160, NREL 45891, NREL 48097,
NREL 46526.
NREL prints on paper that contains recycled content.

Acknowledgments
The authors thank Dr. Anastasios Golnas, Dr. Guohui Yuan, and Dr. Charlie Gay (retired) of the
U.S. Department of Energy (DOE) Solar Energy Technologies Office (SETO) for their
continuous support of this project.
We express our gratitude to the whole First Solar organization for providing substantial
contributions to this project in the form of a fully operational 430-kW photovoltaic (PV) power
plant and control system, valuable guidance, and countless hours of engineering and logistics
support. Special thanks to Mahesh Morjaria, of Terabase Energy (formerly of First Solar, vice
president of PV systems development), for his vision and guidance. We are also grateful to the
late Kevin Collins, formerly of First Solar, for his contributions to this work.
We also thank and acknowledge the substantial contributions to this work provided by the staff
of the National Renewable Energy Laboratory’s (NREL’s) Flatirons Campus, formerly known as
the National Wind Technology Center, and NREL’s Power Systems Engineering Center during
the various stages of this project.
NREL was supported by DOE SETO under Agreement Number 31797 in the production of this
report.
iii
This report is available at no cost from the National Renewable Energy Laboratory at www.nrel.gov/publications.
List of Acronyms
AGC automatic generation control
ARIES Advanced Research on Integrated Energy Systems
BESS battery energy storage system
BMS battery management system
CAISO California Independent System Operator
CGI controllable grid interface
DAS data acquisition system
DOE U.S. Department of Energy
EMS energy management system
ERCOT Electric Reliability Council of Texas
FFR fast frequency response
GHI global horizontal irradiance
GPS Global Positioning System
IBR inverter-based resource
IEEE Institute of Electrical and Electronics Engineers
IESS Integrated Energy Systems at Scale
IR infrared
Li-ion lithium-ion
MPP maximum power point
MPPE maximum power point estimation
MPPT maximum power point tracking
NERC North American Electric Reliability Corporation
NREL National Renewable Energy Laboratory
NWTC National Wind Technology Center
PCC point of common coupling
PCC point of common coupling
PFR primary frequency response
PHIL power-hardware-in-the-loop
PLL phase-locked loop
PMU phasor measurement unit
POD power oscillation damping
POI point of interconnection
PPC power plant controller
PREPA Puerto Rica Electric Power Authority
PSS power system stabilizer
PV photovoltaic
ROCOF rate of change of frequency
RPS renewables portfolio standards
RTAC Real-Time Automation Controller
RTDS real-time digital simulator
SCADA supervisory control and data acquisition
SDS safety data sheet
SETO Solar Energy Technologies Office
SF synchronous frame
SOC state of charge
iv
SSI sinusoidal signal integrator
THD total harmonic distortion
v
Executive Summary
Although utility-scale solar photovoltaic (PV) power plants are becoming a cost-effective energy
resource, there is belief within the energy industry that the increasing penetrations of PV
technologies could potentially impact grid reliability. This is due to the variability across
timescales, the forecast uncertainty of the solar energy resource, and the impacts on both
distribution and transmission systems. This can cause utilities to severely limit PV installations
or to assign PV integration costs when considering a least-cost portfolio of resources. As
California and other regions in the United States are marching toward fulfilling—and probably
exceeding—their renewables portfolio standards goals, grid operators are beginning to face
operational challenges that could have implications for existing renewable and nonrenewable
generators and that will shape opportunities for future projects. In this work, we focused on
developing controls and conducting demonstrations for AC-coupled PV-battery energy storage
systems (BESS) in which PV and BESS are colocated and share a point of common coupling
(PCC). The PV and BESS systems do not share any physical components (such as inverters,
transformers, protection, or energy metering equipment), but they have a common controller that
can operate both PV and BESS either as a single plant or as two independent systems. The
general configuration of a utility-scale PV-BESS plant used to develop and demonstrate many
control concepts during this project is shown in Figure ES-1. . As colocated resources, PV
generation and BESS naturally share the same infrastructure (substation, PCC, tie-lines) and
plant-level controller that operates both technologies as one utility-scale asset interacting with a
single market interface.
Figure ES-1. Colocated AC-coupled PV-storage power plant

The cost of BESS has dramatically declined in recent years, presenting an opportunity for energy
storage not only to perform functions currently met by conventional generators that serve peak
electricity demand but also to provide new opportunities for economic hybridization with
variable generation, including PV. The use of storage can change and customize the “shape” of
PV production to better match load and peak demand in many power systems, make PV
generation more flexible, and facilitate very high levels of PV generation without curtailment.
vi
This cooperative research-and-development agreement project conducted by the National
Renewable Energy Laboratory (NREL) and First Solar leveraged and benefited from the
knowledge and field experience accumulated from NREL’s previous work with First Solar and
other partners—e.g., the California Independent System Operator, the Electric Reliability
Council of Texas, AES, Puerto Rica Electric Power Authority (PREPA)—and advanced prior
work by:
• Combining PV plant and BESS controls to provide the existing and future projected
ancillary service products by various balancing authorities and system operators in U.S.
interconnections and island systems
• Developing controls with higher levels of complexity due to the multi-technology nature
of PV-BESS operation
• Using an advanced PV power plant controller based on the Schweitzer Engineering
Laboratories Real-Time Automation Controller (RTAC) platform supplemented with new
control algorithms to include BESS and provide many types of essential reliability
services
• Using advanced, smart string PV inverters that provide high performance, robust air-
cooled design, easy serviceability, and a quicker return on investment
• Demonstrating controls by the utility-scale PV plant combined with BESS that can
qualify for participating in the existing ancillary service spinning and nonspinning
reserve markets
• Demonstrating new PV-BESS controls that can be required from variable generation by
evolving ancillary service markets, such as primary frequency response (PFR), fast
frequency response (FFR), down-ramp control, short-term variability smoothing, and
voltage control
• Developing new advanced ancillary service controls by PV-BESS plants for interarea
oscillations damping, subsynchronous resonance mitigation, mitigation of control
interactions, etc.
• Demonstration of grid-forming controls by BESS to provide resilience services (black
start, islanded operation, system restoration) to hybrid PV-plus-storage plants.
The project resulted in the development of a new research asset—a grid-scale PV-storage hybrid
system test bed that can be used by the industry and research community for the validation and
demonstration of new control concepts, stakeholder engagement, workforce education, and as a
validation platform for the future standardization of hybrid technologies. It will lead to the
development of both technology and vendor-neutral, “plug-and-play” hybridization guidelines
and control architecture openly available at all segments of the stakeholder community. This
project was the first research effort that demonstrated the multi-technology aspect of grid
integration research possible at NREL’s Flatirons Campus and facilitated the shift toward the
development of new research platforms, such as Integrated Energy Systems at Scale (IESS) and
the Advanced Research on Integrated Energy Systems (ARIES). Capabilities developed under
this project were also foundational for continued hybrid systems research conducted at NREL
and were used in other U.S. Department of Energy-funded research work, such as the Grid
Modernization Laboratory Consortium FlexPower Hybrid Plants project.
The emergence of new integrated energy systems provides many opportunities, and it also
presents urgent challenges that must be addressed to ensure that these new integrated energy
vii
systems meet evolving consumer expectations and continue to be economic, reliable, resilient,
and secure. From this perspective, integrated PV-BESS systems can play a crucial role for power
systems in the transition to high shares of inverter-based resources.
Controls developed and demonstrated under this project for PV-storage plants under both real
grid-connected and controlled grid conditions (emulated by NREL’s multi-MW grid simulator)
include:
• Controls for inertia-like response when the PV-storage plant was responding to the rate of
change of frequency
• Controls for FFR
• Controls for PFR
• Ability of PV storage plants to participate in automatic generation control
• Dispatchability controls
• Variability smoothing and ramp-limiting tests
• Advanced power system oscillations damping controls
• Real-time potential available power estimation in curtailed PV plants
• Resilience service controls in the form of black start by grid-forming battery energy
storage and the islanded operation of a PV-storage system.
The research and testing activities under this cooperative research-and-development project help
address many goals in the areas of achieving a dramatic increase of PV-generated electricity that
can be dispatched any time and in maintaining the adequate levels of reliability. The results of
this work can be used in designing future ancillary service markets in the United States and
abroad to incentivize renewable generators and energy storage to provide such services for
enhanced grid stability and reliability.
viii
Table of Contents
1 Introduction ........................................................................................................................................... 1
1.1 Solar PV Integration Challenges ................................................................................................... 1
1.2 Configurations of PV-Storage Systems ......................................................................................... 4
1.3 Structure of This Report ................................................................................................................ 5
2 Project Objectives ................................................................................................................................ 6
3 NREL-First Solar Collaboration........................................................................................................... 8
4 PV Plant Construction at NREL........................................................................................................... 9
5 Plant Commissioning ......................................................................................................................... 15
6 Test Platform and Control Architecture ........................................................................................... 17
7 Measurement of Response Times of the PV-BESS System ........................................................... 24
8 Active Power Control Tests ............................................................................................................... 26
8.1 Inertial Response Tests................................................................................................................ 27
8.2 Fast Frequency Response Tests................................................................................................... 30
8.2.1 FFR-PHIL Simulations .................................................................................................. 30
8.2.2 PMU-Based FFR Concept.............................................................................................. 32
8.3 Frequency Droop Tests ............................................................................................................... 36
8.3.1 Frequency Droop Test Results by PV-BESS System .................................................... 36
8.3.2 Comparison of Frequency Response Methods ............................................................... 39
8.4 AGC Tests ................................................................................................................................... 40
8.5 Dispatchability of PV-BESS Plant .............................................................................................. 42
8.6 Ramp-Limiting Tests................................................................................................................... 44
8.7 Oscillation Damping Controls ..................................................................................................... 47
8.8 Modeling of POD by PV_BESS plants ....................................................................................... 50
9 Reactive Power Control Tests ........................................................................................................... 54
10 Impedance Characterization of the PV-BESS Plant ........................................................................ 57
10.1 Impedance Measurement System ................................................................................................ 57
10.1.1 Grid Simulator ................................................................................................................ 57
10.1.2 GPS-Synchronized Medium-Voltage Measurements .................................................... 58
10.1.3 Impedance Measurement System ................................................................................... 58
10.2 Impedance Measurements ........................................................................................................... 59
10.3 Impedance-Based Prediction of Oscillatory Instability During Weak Grid Operation ............... 61
11 Black Start of PV Plant with BESS .................................................................................................... 63
11.1 Black-Start Experiment with CGI ............................................................................................... 63
11.2 Black-Start Experiment with BESS ............................................................................................ 65
11.2.1 Soft Black Start .............................................................................................................. 65
11.2.2 Hard Black Start ............................................................................................................. 67
11.3 Other Black-Start Configurations................................................................................................ 68
12 Solar Forecast ..................................................................................................................................... 73
13 Accurate Method for Real-Time Power Reserve Estimation for Utility-Scale PV Power Plant ... 75
13.1 Importance of Accurate Estimation of Available PV Power....................................................... 75
13.2 Proposed Method......................................................................................................................... 78
13.3 Implementation of Reserve Estimation Method on the 430-kW PV Power Plant ...................... 80
13.4 Evaluation of Peak Power Estimation Method for a Large 300-MW PV Power Plant ............... 84
14 Test Platform for PV-BESS Systems in Distribution Grids ............................................................ 87
15 Conclusions ........................................................................................................................................ 90
16 Future Opportunities .......................................................................................................................... 92
References ................................................................................................................................................. 93
ix
List of Figures
Figure ES-1. Colocated AC-coupled PV-storage power plant..................................................................... vi
Figure 1. CAISO generation breakdown for April 21, 2019......................................................................... 2
Figure 2. Curtailed portion of PV generation on April 21, 2019 .................................................................. 3
Figure 3. Available and curtailed PV power on April 21, 2019 ................................................................... 3
Figure 4. PV-BESS plant configuration used in this work ........................................................................... 5
Figure 5. ARIES virtual emulation platform at NREL’s Flatirons Campus ................................................. 7
Figure 6. Physical layout of First Solar’s PV plant at NREL’s Flatirons Campus ....................................... 9
Figure 7. Array-mounting method .............................................................................................................. 10
Figure 8. Aerial photo of First Solar’s PV plant at NREL’s Flatirons Campus. Photo by Red Mountain
Scientific ................................................................................................................................. 10
Figure 9. Layout of First Solar’s PV plant site at NREL’s Flatirons Campus ............................................ 10
Figure 10. Interconnection diagram for six string inverters ........................................................................ 11
Figure 11. A 50-kVA string inverter (SUN2000-45KTL-US-HV) ............................................................ 12
Figure 12. Circuit diagram of a 45-kW string inverter ............................................................................... 12
Figure 13. A 125-kVA string inverter (Sungrow SG125HV) ..................................................................... 13
Figure 14. View of First Solar’s PV plant at NREL’s Flatirons Campus. Photo by NREL ........................ 13
Figure 15. Aerial photo of BESS. Photo by NREL ..................................................................................... 14
Figure 16. Virtual plant zones for IR imaging. Image by Red Mountain Scientific .................................... 15
Figure 17. Thermal image for the whole plant. Image by Red Mountain Scientific ................................... 16
Figure 18. Typical daily production profile for First Solar’s PV plant ....................................................... 16
Figure 19. PV plant and BESS integrated into the hybrid testing platform at NREL's Flatirons Campus . 17
Figure 20. Main characteristics of the 7-MVA CGI ................................................................................... 18
Figure 21. PV-BESS plant communications ............................................................................................... 20
Figure 22. Flatirons Campus main SCADA display ................................................................................... 21
Figure 23. PV plant SCADA display .......................................................................................................... 21
Figure 24. BESS SCADA display............................................................................................................... 22
Figure 25. Controller for the BESS to provide plant services implemented in the RTAC ......................... 22
Figure 26. Controller diagram for the active and reactive power control services by BESS ...................... 23
Figure 27. Measured PV plant response times ............................................................................................ 24
Figure 28. Measured BESS response time in grid-following mode............................................................ 25
Figure 29. Measured BESS response time in grid-forming mode .............................................................. 25
Figure 30. BESS response to active power set points ................................................................................. 25
Figure 31. PHIL setup using 9-bus power system model in RTDS ............................................................ 27
Figure 32. Results of the PHIL tests for PV-BESS plants providing inertial response at different
penetration levels: (a) 0%, (b) 15%, (c) 20%, (d) 40%, and (e) 60% .................................... 28
Figure 33. Comparison of different PLL algorithms under balanced grid conditions ................................ 29
Figure 34. Comparison of different PLL algorithms under 5% imbalance in grid conditions.................... 29
Figure 35. Calculated range of programmable H for NREL’s 1-MW/1-MWh BESS ................................ 30
Figure 36. Result of FFR PHIL tests at different IBR penetration levels: (a) 0%, (b) 15%, (c) 20%, (d)
40%, and (e) 60% ................................................................................................................... 31
Figure 37. PMU networks embedded in the power system ........................................................................ 32
Figure 38. Concept of PMU-based wide-area reliability service by PV-BESS plants................................ 33
Figure 39. PMU-controls test bed developed at NREL's Flatirons Campus ............................................... 33
Figure 40. PMU-based system characterization test setup.......................................................................... 34
Figure 41. Transfer function of PMU data flows to 5 % voltage magnitude injection by the CGI ............ 34
Figure 42. Transfer functions of PMU data paths in response to the 0.5-Hz frequency injection by the CGI
................................................................................................................................................ 35
Figure 43. Transfer functions of PMU data paths in response to the 0.1-rad phase angle injection by the
CGI ......................................................................................................................................... 35
x
Figure 44. Modified IEEE 39-bus system for optimally placed PMUs ...................................................... 36
Figure 45. Frequency and voltage response of 39-bus system: (left) no FFR and (right) with PMU-based
FFR ........................................................................................................................................ 36
Figure 46. Frequency droop by a PV-BESS system ................................................................................... 37
Figure 47. PV plant providing 5% droop response ..................................................................................... 38
Figure 48. PV plant providing 3% droop response ..................................................................................... 38
Figure 49. BESS providing frequency droop response ............................................................................... 39
Figure 50. Comparison of impacts by inertial, FFR, and PFR controls ...................................................... 40
Figure 51. PV plant following the AGC set point under high-variability conditions ................................. 41
Figure 52. Correlation between the measured PV plant power and the AGC set points under high-
variability conditions.............................................................................................................. 41
Figure 53. PV plant following the AGC set point under low variability conditions................................... 41
Figure 54. Correlation between the measured PV plant power and the AGC set points under low
variability conditions.............................................................................................................. 42
Figure 55. BESS following the AGC set points—grid-following mode .................................................... 42
Figure 56. Example of production profile-shaping PV-BESS system ........................................................ 43
Figure 57. Examples of dispatchable operation by the PV-BESS system .................................................. 44
Figure 58. PV-BESS system in ramp-limiting mode (50-kW/min limit) providing PFR ........................... 45
Figure 59. Ramp-limiting control of the PV-BESS plant (100 k/min) ....................................................... 46
Figure 60. Example of a variability-smoothing test by the BESS in grid-forming mode ........................... 47
Figure 61. Main control screen for the P-Q set points and oscillation control ........................................... 48
Figure 62. PV power plant following the sinusoidal set point .................................................................... 48
Figure 63. BESS in grid-following mode following the sinusoidal set point ............................................. 48
Figure 64. Time-domain captures for fast Fourier transform analysis for First Solar’s 430-kW PV plant........ 49
Figure 65. Time-domain captures for the fast Fourier transform analysis for the BESS ............................ 49
Figure 66. Gain and phase response bode plot for different resources ....................................................... 49
Figure 67. Modified two-area system model .............................................................................................. 50
Figure 68. Undamped oscillations triggered by a small change in load ..................................................... 51
Figure 69. Oscillations damped by POD control applied to the PV plant (Td=200 ms) ............................. 52
Figure 70. PV plant providing POD without continuous curtailment ......................................................... 53
Figure 71. Measured reactive power characteristics of the PV plant (all inverters) ................................... 55
Figure 72. Measured reactive power capability of the BESS system ......................................................... 56
Figure 73. Reactive power capability measured on the low-voltage (SMA) and medium-voltage (CGI)
sides of the BESS transformer ............................................................................................... 56
Figure 74. Combined reactive power capability of the PV-BESS plant ..................................................... 56
Figure 75. Hardware in each medium-voltage DAS sensing node: (left) voltage and current sensors;
(right) data conditioning built around the National Instruments’ cRIO-9030 platform. Photos
by NREL ................................................................................................................................. 58
Figure 76. Impedance measurement system diagram ................................................................................. 59
Figure 77. Inverters for impedance characterization: the (a) 2.2-MVA inverter interfacing the 1-MW/1-
MWh BESS and the (b) 430-kW PV plant with six string inverters. Photos by NREL ......... 60
Figure 78. Positive-sequence impedance response of the 1-MW/1-MWh BESS plant measured for
different magnitudes of the voltage perturbation injected by the CGI for the impedance
measurement .......................................................................................................................... 60
Figure 79. Positive-sequence impedance response of two sets of string inverters (4 kW x 40 kW and 2 kW
x125 kW) interfacing First Solar’s 430-kW PV plant............................................................ 61
Figure 80. Comparison of positive-sequence impedance response of the 1-MW/1-MWh BESS plant when
it is operating in grid-following and grid-forming control modes ......................................... 62
Figure 81. Impedance-based prediction of instability of First Solar’s 430-kW PV plant during operation
under a weak grid condition: the (a) impedance analysis comparing the impedance responses
xi
of the PV inverters with the grid impedance and the (b) PV plant output current during
instability................................................................................................................................ 62
Figure 82. Black start of PV system with the CGI...................................................................................... 64
Figure 83. Series of black-start experiments using the CGI ....................................................................... 64
Figure 84. Test setup for the black start of the PV plant with BESS .......................................................... 65
Figure 85. Soft black start of the PV system using the BESS..................................................................... 66
Figure 86. (Top) Soft black start of the PV system using the BESS and (bottom) magnified view of the
voltage ramping by the inverter. The voltage ramp was performed within 200 ms............... 66
Figure 87. Hard black start of the PV system by closing the PV breaker on the live voltage bus .............. 67
Figure 88. Hard black start of the PV system by closing the PV breaker on the live voltage bus—
magnified view of the voltage and current transient .............................................................. 68
Figure 89. New black-start paradigm.......................................................................................................... 69
Figure 90. Configurations of integrated PV-plus-storage plants as black-start resource for conventional
plants ...................................................................................................................................... 70
Figure 91. PV-plus-storage plant to start auxiliary loads in a thermal power plant .................................... 71
Figure 92. Consecutive starts of auxiliary motors ...................................................................................... 71
Figure 93. Constant volt/Hertz startup ........................................................................................................ 72
Figure 94. Correlation between the measured hourly and the forecasted GHI ........................................... 73
Figure 95. Measured hourly average site temperature during a two-week period ...................................... 74
Figure 96. Example of the forecasted and the measured PV plant hourly power production during 2 weeks
of operation ............................................................................................................................ 74
Figure 97. Absolute error in the estimated power using an hour-ahead forecast ........................................ 74
Figure 98. Example comparison of an hour-ahead and a week-ahead forecast .......................................... 74
Figure 99. Example of inaccurate maximum peak power estimation ......................................................... 77
Figure 100. Example of maximum peak power estimation using a single reference inverter .................... 77
Figure 101. Estimation of available power for a PV power plant ............................................................... 78
Figure 102. Large PV power plant divided into control zones during cloud conditions ............................ 79
Figure 103. Examples of 430-kW plant daily production profiles.............................................................. 81
Figure 104. Measured 1-s total power ramps .............................................................................................. 81
Figure 105. Comparison of estimated and measured power using simple method ..................................... 82
Figure 106. Peak power estimation error .................................................................................................... 82
Figure 107. Correlation between the estimation error and the 1-s variability in the power output of the
reference inverter and linear regression ................................................................................. 83
Figure 108. Distribution of peak power estimation errors .......................................................................... 84
Figure 109. Example of the daily operation profile of a 300-MW PV plant .............................................. 84
Figure 110. Dynamic reference inverters (light green circles represent reference inverters) ..................... 85
Figure 111. Peak power estimation errors .................................................................................................. 86
Figure 112. Test bed for PV-BESS systems in distribution microgrids and islanded grids ....................... 88
Figure 113. Measured response of grid-forming BESS after Circuit Breaker 1 trips in the PHIL model .. 88
Figure 114. Minimum voltages and maximum currents recorded in all buses during a line-to-line low-
impedance fault in Bus 691 .................................................................................................... 89
Figure 115. Main characteristics of the 19.9-MVA CGI ............................................................................ 92
List of Tables
Table 1. Parameters of First Solar’s PV Plant ............................................................................................ 11
Table 2. PHIL Scenarios ............................................................................................................................. 26
xii
1 Introduction
1.1 Solar PV Integration Challenges
Although utility-scale solar photovoltaic (PV) power plants are becoming a cost-effective energy
resource, there is belief within the industry that the increasing penetrations of PV technologies
could potentially impact grid reliability. This is due to the variability across timescales and
forecast uncertainty of the solar energy resource and the impacts on both distribution and
transmission systems. This can cause utilities to severely limit PV installations or assign PV
integration costs when considering a least-cost portfolio of resources. With PV’s increased
proportion of energy to the generation mix, advanced PV controls and grid integration features
can minimize grid impacts from variability and, in many cases, improve reliability, stability, and
power quality. The deployment of utility-scale, grid-friendly PV power plants that incorporate
advanced capabilities to support grid stability and reliability is essential for the large-scale
integration of PV generation into the electric power grid.
The rapidly declining cost of PV technology in combination with evolving renewables portfolio
standards (RPS) is driving increased PV deployment in many states. A key limiting factor for PV
deployment is curtailment due to overgeneration and other constraints. Energy storage can play a
crucial role in keeping PV curtailment to acceptable levels, thus the enabling cost-effective
integration of larger amounts of PV generation. For example, in fall 2015, the California state
legislature passed Senate Bill 350, which requires all utilities in the state to produce 50% of their
electricity sales from renewable sources by 2030. In 2018, Senate Bill 100 was signed into law,
which increased the RPS to 60% by 2030 and required all the state’s electricity to come from
carbon-free resources by 2045 [1]. To reach these RPS goals, California operators will need to
find additional ways to balance generation and load to manage the variability of increased
renewable generation. In combination with the 1,325-GW California energy storage procurement
mandate, ancillary services provided by renewables can enhance system flexibility and reliability
and reduce needs in spinning reserves by conventional power plants. The exact amount of energy
storage needed to meet the RPS goals depends on the complex interdependence of a range of grid
flexibility options. One National Renewable Energy Laboratory (NREL) study [2] estimated that
under certain scenarios of flexibility and PV levelized cost of energy, nearly 19 GW of energy
storage will be required to meet California’s 50% RPS goals. Similar studies have been
conducted by others looking into the economic and operational impacts of energy storage on PV
integration for creating regionally specific low-cost, balanced, and low-carbon generation
portfolios; however, the role of energy storage in combination with high levels of variable PV
generation when providing a wide range of reliability services in the most cost-effective and co-
optimized way is largely unknown to the industry. Many questions are still unknown and depend
on future research efforts, including control objectives and architectures, optimization methods,
scalability issues, and technology-specific response characteristics. In addition, the lack of
publicly available testing and demonstration results for both energy storage and PV technologies
operating with the common system-level control objectives at large scales is a significant barrier
to bridging the knowledge gaps in the research community and among industry stakeholders.
As California and other regions in the United States are marching toward fulfilling—and
probably exceeding—these RPS goals, grid operators are beginning to face operational
challenges that could have implications for existing renewable and nonrenewable generators and
1
that will shape opportunities for future projects. One example of such a challenge took place on
April 21, 2019, as shown in Figure 1, when up to 4.5 GW of PV generation in California had to
be curtailed because of overgeneration. The actual curtailed portion of PV generation on the
same day is shown in Figure 2. This curtailed energy was essentially lost, and despite being
available as a massive spinning reserve, it was not used to provide any services to California’s
electric grid. Advanced inverter functions and the design and operation of PV projects can help
address grid stability problems during such periods. A typical modern utility-scale PV power
plant is a complex system of large PV arrays and multiple power electronic inverters, and it can
contribute to mitigating the impacts on grid stability and reliability through sophisticated,
automatic grid-supportive controls. Integration with energy storage can assist PV integration in
any number of ways by increasing power system flexibility.
Figure 1. CAISO generation breakdown for April 21, 2019.
2
Figure 2. Curtailed portion of PV generation on April 21, 2019
It is remarkable that a small portion of this curtailed energy would have been sufficient to cover
all California Independent System Operator (CAISO) frequency regulation needs during almost
the entire period of curtailment, as shown in Figure 3. The dotted line represents the 600-MW
frequency regulation reserve margin that PV generation could have provided if it had been
included in CAISO’s automatic generation control (AGC). This example shows how curtailed
energy could have been used for certain grid services or could have been stored for later use.
Figure 3. Available and curtailed PV power on April 21, 2019

In November 2015, the North American Electric Reliability Corporation (NERC) task force set a
framework of reliability services recommended for variable generation (including solar) to
provide their share of grid services, including sufficient voltage control, frequency support, and
ramping capability—essential components of a reliable bulk power system. The capabilities of
utility-scale PV generation technology to provide such services has been demonstrated by NREL
and partners in recent U.S. Department of Energy (DOE)-funded groundbreaking demonstration
3
projects conducted in West Texas and Puerto Rico using modern PV plants with rated capacities
of +20 MW [3]. In addition, a more recent DOE Solar Energy Technologies Office (SETO)-
funded project by NREL in collaboration with CAISO using First Solar’s 300-MW PV power
plant in California has advanced the industry knowledge even further by demonstrating the
ability of PV technology to provide services of unprecedented scale at the system level [4].
Results of these projects are in the public domain and have already triggered significant interest
from various stakeholder groups.
The cost of battery energy storage systems (BESS) has dramatically declined in recent years,
presenting an opportunity for energy storage not only to perform functions currently met by
conventional generators that serve peak electricity demand but also to provide new opportunities
for economic hybridization with variable generation, including PV. The use of storage can
change and customize the “shape” of PV production to better match load and peak demand in
many power systems, make PV generation more flexible, and facilitate very high levels of PV
generation without curtailment.
1.2 Configurations of PV-Storage Systems

The degree to which the PV and storage are coupled (both physically and operationally) can be
divided into four distinct categories:
• Independent: PV and battery are not colocated and do not have a common point of
connection, and energy stored in the battery could come from either PV or the grid.
• AC-coupled: PV and battery are colocated and have a common point of connection at the
plant substation, and energy stored in the battery could come from either PV or the grid.
• DC-coupled: Battery is connected to the DC side of the PV inverters, and energy stored
in the battery could come from either PV or the grid.
• Tightly DC-coupled: Battery is connected to the DC side of PV inverters, and energy
stored in the battery could come only from PV.
Each PV-plus-storage configuration has advantages and disadvantages [5]–[9]. Many analyses of
PV-plus-storage systems have focused on costs and cost minimization. A sufficient amount of
analyses have been done on evaluating the levelized-cost-of-energy metrics for PV-plus-storage
systems [10] and on evaluating the economic performance of PV-BESS systems in terms of their
cost/benefit ratios [5]. In this work, we focused on developing controls and conducting
demonstration testing for AC-coupled PV-BESS systems in which PV and BESS are colocated
and share a point of common coupling (PCC). PV and BESS systems do not share any physical
components (such as inverters, transformers, protection, or energy metering equipment), but they
have a common controller that can operate both PV and BESS either as a single plant or as two
independent systems. The general configuration of a utility-scale PV-BESS plant used to develop
and demonstrate many control concepts during this project is shown in Figure 4. As colocated
resources, PV generation and BESS naturally share the same infrastructure (substation, point of
interconnection [POI], tie-lines) and plant-level controller that operates both technologies as one
utility-scale asset interacting with a single market interface.
4
Figure 4. PV-BESS plant configuration used in this work
1.3 Structure of This Report

To create a fully functional validation testbed for PV-BESS systems, help the industry with
developing and demonstrating many types of controls for PV-BESS plants, and establishing a
common framework for further discussion, analysis, and advances in PV-plus-storage systems,
we developed an “at-scale” system and explored many types of operational aspects of such a
system that are hard to evaluate in the field using real commercial installations. The report is
intended to demonstrate results of 3-yr activities conducted by NREL and First Solar to address
two primary goals:
1. Develop a world-class research facility consisting of an at-scale PV-BESS system at

NREL’s Flatirons Campus that is capable of providing a wide spectrum of reliability and
resilience services when operating under real and simulated grid conditions or as part of
stand-alone islanded grids.
2. Develop and demonstrate controls for the provision of such services using this test bed.
The report is structured as follows. Sections 2–6 describe the process of developing and
commissioning a PV-BESS validation platform. Sections 7–9 describe results of various types of
tests involving different forms of active and reactive power control. Section 10 describes the
innovative impedance characterization of PV and BESS inverters developed by NREL. Section
11 describes black-start controls. Section 12 describes the solar resource forecast system
implemented at NREL’s Flatirons Campus. And Section 13 describes the NREL-developed real-
time method for estimating the available power reserve in curtailed PV power plants.
5
2 Project Objectives
Emerging hybrid renewable energy systems offer (1) new opportunities for the global renewable
energy industry with disruptive market potential and (2) a scalable, economic, and reliable
solution applicable to a power system of any size (e.g., large, interconnected power systems,
islands, microgrids). The BESS, because of its tremendous range of uses and configurations, can
assist PV integration in any number of ways by increasing power system flexibility. These uses
include (1) matching generation to loads through time shifting, reducing PV curtailment,
managing transmission congestion, and controlling elements of power flow; (2) promoting
higher levels of PV penetrations by balancing the grid through ancillary services, load following,
ramp limiting, and load levelling; (3) managing forecast uncertainty and short-term variability in
PV generation through reserves; (4) smoothing output from individual solar power plants; and
(5) enabling islanded and microgrid applications. In addition, in distribution applications, BESS
can also play crucial roles in increasing hosting capacities of distribution feeders for large levels
of PV penetration by retail energy time shifting, providing voltage support, and improving power
quality. This collaborative project aimed toward producing publicly available valid test data and
a comprehensive performance evaluation to understand the impacts of combining PV and BESS
for all these uses. The project demonstrated many types of services by PV and energy storage
systems based on different forms of active and reactive power controls by PV and BESS in both
grid-connected mode and under controlled grid conditions emulated by the 7-MVA controllable
grid interface (CGI) at NREL’s Flatirons Campus.
This cooperative research-and-development agreement project conducted by NREL and First

Solar leveraged and benefited from the knowledge and field experience accumulated from
NREL’s previous work with First Solar and other partners—e.g., CAISO, the Electric Reliability
Council of Texas (ERCOT), AES, Puerto Rico Electric Power Authority (PREPA)— and
advanced prior work by:
• Combining PV plant and BESS controls to provide the existing and future projected
ancillary service products by various balancing authorities and system operators in U.S.
interconnections and island systems
• Developing controls with higher levels of complexity due to the multi-technology nature
of PV-BESS operation
• Using an advanced PV power plant controller (PPC) based on the SEL Real-Time
Automation Controller (RTAC) platform supplemented with new control algorithms to
include BESS and provide many types of essential reliability services
• Using advanced, smart string PV inverters that provide high performance, robust air-
cooled design, easy serviceability, and a quicker return on investment
• Demonstrating controls by the utility-scale PV plant combined with BESS that can
qualify for participating in the existing ancillary service spinning and nonspinning
reserve markets
• Demonstrating new PV-BESS controls that can be required from variable generation by
evolving ancillary service markets, such as primary frequency response (PFR), fast
frequency response (FFR), down-ramp control, short-term variability smoothing, and
voltage control
6
• Developing new advanced ancillary service controls by PV-BESS plants for interarea
oscillations damping, subsynchronous resonance mitigation, mitigation of control
interactions, etc.
• Demonstration of grid-forming controls by BESS to provide resilience services (black
start, islanded operation, system restoration) to hybrid PV-plus-storage plants. 1
The project resulted in the development of a new valuable testing and validation asset—a grid-
scale PV-plus-storage hybrid system test bed that can be used by industry and research
community for validation and demonstration of new control concepts, stakeholder engagement,
workforce education, and as a validation platform for the future standardization of hybrid
technologies. In fact, this test bed has already been included as a crucial testing capability in
many ongoing research projects and in research proposals led by NREL, industry, and academia
in response to various funding opportunity announcements by several DOE programs and
requests for proposals from industry. This research is intended to lead development toward both
technology- and vendor-neutral, “plug-and-play” hybridization guidelines and control
architectures openly available at all segments of the stakeholder community.
This project was the first research effort that demonstrated the multi-technology aspect of grid
integration research possible at NREL’s Flatirons Campus, and it facilitated the shift toward the
development of new research platforms, such as the Integrated Energy Systems at Scale (IESS)
and Advanced Research on Integrated Energy Systems (ARIES), as shown in Figure 5.
Capabilities developed under this project were also foundational for continued hybrid systems
research conducted at NREL and were used in other DOE-funded research work, such as the
Grid Modernization Laboratory Consortium FlexPower hybrid plants project.
Figure 5. ARIES virtual emulation platform at NREL’s Flatirons Campus
1
Grid-forming controls by utility-scale PV inverters are being developed and will be tested during 2020–2022 by
GE and NREL under another SETO-funded project.
7
3 NREL-First Solar Collaboration
This project was built on NREL and First Solar’s prior efforts that demonstrated that large
utility-scale PV plants are capable of providing a range of essential grid reliability services [3],
[4]. The objective of this research project is to further advance the accumulated controls
knowledge from the PV-only area to the multi-technology domain by developing and testing the
coordinated controls for PV generation and energy storage cooperation for energy management
and ancillary services. This project is unique from other DOE-funded PV-storage integration
projects in that the evaluation conditions will be fully controllable because of the distinctive
medium-voltage grid emulation capabilities at NREL’s Flatirons Campus.
NREL’s strategic partnership with First Solar has resulted in the continuation of research
collaboration in the area of developing and testing controls for the operation of utility-scale PV
generation with energy storage systems for supporting various use cases related to curtailment
reduction and reliability services. For this purpose, First Solar installed their 430-KW AC fixed-
tilt PV plant at NREL’s Flatirons Campus for long-term joint research as a shared resource
(estimated cost $350K). In addition, First Solar contributed 3 years of labor to this work ($650K
in total). In case of future needs, the First Solar plant is easily expandable to larger capacities
because the plant’s electrical topology will be based on a scalable advanced string inverter-based
configuration (smart air-cooled inverters with advanced grid-friendly controls will be used by
First Solar in this project). First Solar is also open to the long-term use of the plant for continued
research beyond the scope of this project.
To establish the collaboration framework, NREL and First Solar modified the preexisting
cooperative research-and-development agreement to include a new 3-yr research project
involving the installation and operation of a new 430-kW array that uses advanced thin-film PV
modules with string inverters and NREL’s 1-MW/1-MWh BESS. The BESS consists of LG’s
lithium-ion (Li-ion) batteries with an SMA Storage Central inverter/charger. The technical
specifications of NREL’s BESS procured for this project are given in the appendix.
8
4 PV Plant Construction at NREL
The construction stage of the project was completed during the second quarter of Fiscal Year
(FY) 2018 instead of the originally planned completion time of FY 2017 due to unexpectedly
high installation costs quoted by potential contractors. The array was installed at NREL’s
Flatirons Campus next to the existing and future planned energy storage testing facility for
promoting the research of colocated multi-technology hybrid plants. The layout schematic of the
plant is shown in Figure 6. The plant uses fixed-tilt PV modules with a 25-degree angle facing
south and attached to mounting tables assembled on a number of vertical support posts (Figure 7)
combined in 14 parallel rows, as shown in the aerial image of the complete plant (Figure 8). The
array was divided into six subarrays, each connected to individual string inverters (Figure 9).
Figure 6. Physical layout of First Solar’s PV plant at NREL’s Flatirons Campus
9
Figure 7. Array-mounting method
Figure 8. Aerial photo of First Solar’s PV plant at NREL’s Flatirons Campus.

Photo by Red Mountain Scientific
Figure 9. Layout of First Solar’s PV plant site at NREL’s Flatirons Campus
10
The electric design characteristics of the plant are shown in Table 1.
Table 1. Parameters of First Solar’s PV Plant
Power rating 430 kW AC/480 kW DC

DC/AC ratio 1.14865
Number of rows 14
PV modules First Solar Series 4 105-W modules
Number of modules in a string 14
Number of structural tables 72
DC power per table 5.88 kW DC
Inverter Type A 4 x 45 kW AC
Inverter Type B 2 x 125 kW AC
The electric interconnection of all six string inverters with the plant transformer is shown in
Figure 10. Inverters, plant communication equipment, and control interfaces are all mounted on a
panel located in the northwest corner of the plant.
Figure 10. Interconnection diagram for six string inverters

The plant electrical design consists of two types of air-cooled string inverters. The first type
(Type A) is the SUN2000 50-kVA (45-kW) string inverter (Figure 11), and four of these
inverters are installed in total. The electrical diagram of a single SUN2000 inverter with four
maximum power point tracking (MPPT) circuits and 1,500-V DC max input voltage is shown in
Figure 12.
11
Figure 11. A 50-kVA string inverter (SUN2000-45KTL-US-HV)
Figure 12. Circuit diagram of a 45-kW string inverter

The second type (Type B) of inverter used in the plant is the Sungrow 125-kW string inverter
(Figure 13). Two Sungrow inverters are deployed in the plant. This inverter is also designed for
1,500-V DC max voltage (860-V DC nominal) and supports one MPPT circuit.
12
Figure 13. A 125-kVA string inverter (Sungrow SG125HV)
The PV plant was first energized in April 2018, and commissioning activities were conducted
during May–June 2018. A ground-level photo of the installation is shown in Figure 14.
Figure 14. View of First Solar’s PV plant at NREL’s Flatirons Campus. Photo by NREL
13
The battery system was installed at NREL’s Flatirons Campus test site in April 2017. Figure 15
shows the aerial photo of the BESS pad, including the battery enclosure, the SMA inverter, the
step-up transformer, and the medium-voltage switchgear.
Figure 15. Aerial photo of BESS. Photo by NREL
14
5 Plant Commissioning
The NREL and First Solar team conducted several commissioning tests and verified that the
plant components were operating properly and that the plant could produce expected levels of
electric power under various solar resource conditions. In general, the plant operated as expected
during the first energizing attempt; however, after several weeks of observation and
measurement, a magnetic core saturation effect was observed in the plant transformer causing
distortions in current waveforms. After performing PSCAD modeling and experimenting with
different transformer tap positions, it was concluded that the transformer nameplate
specifications were incorrect, so the decision was made to replace the plant transformer. This
caused several months of delays to start the testing activities until the new transformer was
delivered, installed, and commissioned. After the transformer replacement, the power quality
issue was resolved, and the PV plant operated as expected.
In addition, NREL, in collaboration with Red Mountain Scientific, conducted a survey of the
plant under full load using the infrared (IR) camera installed on an unmanned flying vehicle
(drone). The purpose of the test was to identify any abnormalities in the temperatures of
individual PV modules under load. This way, any defective modules or inconsistent module
wiring could be easily identified. The IR-enabled drone flight was conducted on a “clear-sky”
day to make sure that the imaging was performed under uniform irradiance conditions across the
whole plant.
The aerial inspection consisted of two overlapping IR site images, 69 low-pass IR images, and
several visible spectrum images. This inspection required 40 min of flight time and 5 separate
flights. The plant was virtually separated into 69 different zones (as shown in Figure 16).
Figure 16. Virtual plant zones for IR imaging. Image by Red Mountain Scientific
15
An example of a thermal image taken for the whole plant is shown in Figure 17. More high-
resolution detailed analysis was conducted for smaller areas of the plant. Despite the clear-sky
conditions, there were some differences in panel temperatures due to the wind conditions present
at the site. The wind was strong (gusts up to 12 m/s–13 m/s), causing some temperature
differences across the array; however, the detailed thermal analysis did not reveal any
abnormalities or hot spots in the array temperatures, proving that the installation was done in
accordance with all technical requirements, and there were no factory deficiencies or damaged
modules. Note that the ground seems to be warmer than the panels in Figure 17 because the
picture shows blackbody radiation temperature, not absolute thermodynamic temperature; thus, it
is conceivable that the ground is radiating more than the panels.
Figure 17. Thermal image for the whole plant. Image by Red Mountain Scientific
A typical production profile for the individual inverters of the plant under both clear-sky and
cloud variability conditions is shown in Figure 18.
Combined output of four 45 kW inverters
125 kW inverter #1
125 kW inverter #2
Figure 18. Typical daily production profile for First Solar’s PV plant
16
6 Test Platform and Control Architecture
A test platform was developed combining the PV array and BESS, with both interconnected with
the 13.2-kV research bus at NREL’s Flatirons Campus. This system has the flexibility to be
configured in such a manner, so both the PV plant and the BESS can operate in regular grid-
connected mode or can be switched to operate under controlled grid conditions interconnected
with the 7-MVA CGI, as shown in Figure 19. The CGI is combined with a real-time digital
simulator (RTDS) as a power-hardware-in-the-loop (PHIL) platform, so closed-loop experiments
can be conducted for the PV-plus-storage system connected to emulated models of power
systems of different sizes. A detailed description of the CGI and characteristics of the PHIL
platform are described in detail in [11].
13.2 kV Simulated Grid Bus
13.2 kV
NWTC 1 MW/1 MWh Li-ion BESS 7 MVA Controllable Grid Interface (CGI)
480V/13.2kV
40 MVA for 2 sec DC

First Solar 430 kWac PV array 1.2 MVA
AC
DC
AC
DC
DC AC AC
AC
DC
DC
DC
480V/13.2kV
AC
AC
500 kVA
DC
AC
NR RTAC CGI RTDS
First Solar PPC

PMU Data
PMU Data
Labview PXI
PMU Data
PMU protocol
SEL Phasor Data Modbus
Concentrator Real-time model of a power system
NREL DAS
Medium Voltage High-speed DAS nodes
PMU
Solar forecast and price signals
Figure 19. PV plant and BESS integrated into the hybrid testing platform at NREL's Flatirons
Campus
17
Figure 20. Main characteristics of the 7-MVA CGI
The main characteristics of NREL’s 7-MVA CGI are shown in Figure 20. The continuous power
rating of the CGI is 7 MVA. It includes a 9-MVA active line-side rectifier unit with filter that
allows power to flow from the device under test to the steady grid with a controllable power
factor; however, the test-side converter faces many challenges because it needs to provide grid
simulator functionality and maintain full controllability under transient conditions that might
exist at the PCC. Transient overcurrent capability is one important feature of the grid simulator
because certain types of generators under test can inject high short-circuit currents that exceed
their nominal rating by many multiples of their nameplate rating. Substantial short-term
overcurrent capacity is needed by the grid simulator to maintain stable operation during such
transient events. For this purpose, the CGI topology is based on four 3.3-kV medium-voltage,
neutral-point-clamped inverter units that are normally used to drive industrial-grade motors and a
custom step-up transformer to produce 13.2 kV on the test article terminals. Under continuous 7-
MVA operation, the amplitude of the nominal continuous current at the inverter side of the
transformer is 500 A. To allow for a significant overcurrent capability, the selected medium-
voltage, neutral-point-clamped inverters are based on integrated gate-commutated thyristor
devices. Their maximum current is 2.7 kA, which allows for a 540% overcurrent margin
assuming a 7-MVA device under test.
CGI’s custom transformer is designed to match devices under test with various nominal voltages
by using multiple transformer taps. The transformer is rated for 7-MVA continuous operation
and 560% short-term overcurrent operation to handle short-circuit currents that might be
produced by the test articles. The special configuration of the transformer allows synthesizing
17-level low-distortion voltage waveforms by interleaving 3-level phase voltages.
18
For a given semiconductor device, the power capacity can be increased by decreasing the
switching frequency, which, in turn, can lead to voltage waveshape quality degradation, which is
normally measured as total harmonic distortion (THD). Normally, the desired THD level in
power converters can be maintained by a harmonic filter; however, filters also decrease the
dynamic range of operation. To maintain dynamic performance as fast as possible, the CGI uses
advanced modulation control methods rather than a hardware filter. Thus, a balance is found
among three conflicting requirements of power conversion: multimegawatt power ratings; sub-
1% THD; and extremely high response times, typically smaller than 1 ms. The CGI supports two
multilevel modulation schemes, each optimized for different objectives. A pulse-width
modulated scheme is used for highly dynamic operation, and an optimized pulse pattern
modulator is used to achieve minimum line-to-line voltage THD levels. It is possible to transition
from one mode to another during operation; therefore, advanced test scenarios using high power
quality in normal operation and high dynamic response during faults can be implemented.
Both the PV plant and the BESS supervisory controllers are based on an industrial RTAC (SEL-
3555), making it easier to integrate such controls in field systems. In addition, the NREL site
supervisory control and data acquisition (SCADA) is implemented in the RTAC controller as
well. NREL has a National Instruments LabVIEW PXI RT controller integrated into the system
as well for preliminary testing of various control algorithms prior to deploying them in the
RTAC. Both the PV plant and the BESS have phasor measurement unit (PMU-)enabled relays
installed in 13.2-kV switchgear, so the PMU data streams are sent to all RTACs and a phasor
data concentrator (installed in the main control room at the Flatirons Campus). NREL’s custom
50-kHz Global Positioning System (GPS)-synchronized data acquisition system (DAS) is
deployed in each 13.2-kV node as well, allowing for the capture of high-resolution snapshots of
transient events. The control and communication architecture for the whole platform—involving
RTACs, network interfaces, and individual inverters—is shown in Figure 21.
19
Figure 21. PV-BESS plant communications
A snapshot of the main SCADA display developed for the whole Flatirons Campus is shown in
Figure 22. The PV plant and the BESS are connected to the 13.2-kV grid via circuit breakers
ES2 and ES1, respectively. Then, with the desired position combinations in various medium-
voltage switches, both the BESS and the PV plant can be connected to either the regular grid
(green bus) or the controllable CGI grid (red bus), shown in Figure 22.
The SCADA display of the PV plant control is shown in Figure 23. This SCADA allows for
switching the plant control between remote and manual modes, controlling individual inverters,
and configuring the sources for external set point signals.
20
PV-BESS
system
Figure 22. Flatirons Campus main SCADA display
Figure 23. PV plant SCADA display

The NREL team developed and implemented a BESS controller based on the standard SEL
RTAC industrial controller platform. The BESS RTAC controller is communicating with the
First Solar PPC controller, which was developed on the same hardware platform. This makes the
integration of the NREL-developed controller with the existing First Solar PPC an easy process.
The user interface for the BESS SCADA is shown in Figure 24. It allows for the remote start and
stop of the BESS system and the activation of desired control modes, including the ability to
switch between grid-following and grid-forming modes.
21
Figure 24. BESS SCADA display
The BESS controller is designed to provide various forms of active and reactive power control
with the interface shown in Figure 25. It has the capability to provide inertial response, FFR, and
PFR by the PV-BESS plant. In addition, the BESS can provide reactive power control in the
form of kVAR-voltage droop during dynamic and fault conditions. The implemented control
diagram for all these control functions was developed by NREL and is shown in Figure 26.
Figure 25. Controller for the BESS to provide plant services implemented in the RTAC
22
BESS Control Diagram
Selectable single BESS Selectable number of

inertia constant 1MW BESS units
2HBESS NBESS SOC control
df/dt deadband
PI + Target SOC
+1 MW Inertial response
df/dt df/dt ΔPi ΔPi ΔPi.total
Selectable comm delay d/dt -1 x x x
ROCOF
0.5-2 s -1 MW Conversion
Measured SOC
Low-pass filter Limiter ± 1MW limit
0 or 1 to P.U. Total BESS Active
+
ΔPloss -1 x
ΔPFFR.total
x
FFR response
+
+
P.U.
Power Response Pset
ΔPloss +
ΔP deadband Rate limiter
NBESS 0 or 1
60 Hz
Frequency droop with
Pset with
Q priority
f deadband
ΔPdroop.total Droop response Plimit P or Q priority
Δf
-+
ΔPdroop
Grid frequency x x
BESS
NBESS 0 or 1 P-Q Setpoint
Deadband (18 mHz
default)
Srated =1 p.u. OR
Frequency Droop (5% default)
Qlimit Qset with

P priority
Vref
ΔV deadband ± 1MVAR limit

Total BESS Reactive Power
VRMS -+
ΔV ΔV ΔV
-1 ÷
ΔVp.u.
x x P.U.
Response
Voltage at POI Qset

Low-pass filter Conversion
NBESS to P.U.
Vref
P-Q Control – P priority
1
Voltage Droop
÷ P-Q Control – Q priority
(5% default)
Figure 26. Controller diagram for the active and reactive power control services by BESS
All BESS active power controls components can be combined in a single equation, so at any
instance in time, the total BESS power (injecting or absorbing) is:
𝑃𝑃𝑏𝑏𝑏𝑏𝑏𝑏𝑏𝑏 (𝑡𝑡) = 𝑃𝑃𝑜𝑜 (𝑡𝑡) + ∆𝑃𝑃𝑠𝑠𝑠𝑠𝑠𝑠 (𝑡𝑡) + ∆𝑃𝑃𝑖𝑖 (𝑡𝑡) + ∆𝑃𝑃𝐹𝐹𝐹𝐹𝐹𝐹 (𝑡𝑡) + ∆𝑃𝑃𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑 (𝑡𝑡) + ∆𝑃𝑃𝐴𝐴𝐴𝐴𝐴𝐴 (𝑡𝑡) (6.1)
where 𝑃𝑃𝑜𝑜 (𝑡𝑡) is the BESS dispatch set point; ∆𝑃𝑃𝑠𝑠𝑠𝑠𝑠𝑠 (𝑡𝑡) is the portion of the commanded power set
point for the BESS state-of-charge (SOC) control; ∆𝑃𝑃𝑖𝑖 (𝑡𝑡) is the BESS inertial response (or
response proportional to the rate of change of frequency [ROCOF]); ∆𝑃𝑃𝐹𝐹𝐹𝐹𝐹𝐹 (𝑡𝑡) is BESS FFR
response; ∆𝑃𝑃𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑 (𝑡𝑡) is the BESS droop response; and ∆𝑃𝑃𝐴𝐴𝐴𝐴𝐴𝐴 (𝑡𝑡) is BESS AGC response.
Depending on the types of service that the BESS is providing, the individual components in
equation (1) can be activated at the proper times. For example, ∆𝑃𝑃𝑖𝑖 (𝑡𝑡) will first start at the
beginning of the event as soon as a large ROCOF is detected, then either ∆𝑃𝑃𝐹𝐹𝐹𝐹𝐹𝐹 (𝑡𝑡) or ∆𝑃𝑃𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑 (𝑡𝑡)
will kick in (the BESS can provide either FFR or droop response, but it cannot do both at the
same time). The ∆𝑃𝑃𝐴𝐴𝐴𝐴𝐴𝐴 (𝑡𝑡) component will follow the AGC set points from the system operator
for secondary frequency control and frequency regulation.
Equation (6.1) can be expanded to show the components of interest in more detail:
𝑑𝑑𝑑𝑑(𝑡𝑡) 𝑓𝑓𝑜𝑜 −𝑓𝑓(𝑡𝑡)
𝑃𝑃𝑏𝑏𝑏𝑏𝑏𝑏𝑏𝑏 (𝑡𝑡) = 𝑃𝑃𝑜𝑜 (𝑡𝑡) + ∆𝑃𝑃𝑠𝑠𝑠𝑠𝑠𝑠 (𝑡𝑡) − 2𝐻𝐻 + ∆𝑃𝑃𝐹𝐹𝐹𝐹𝐹𝐹 (𝑡𝑡) − + ∆𝑃𝑃𝐴𝐴𝐴𝐴𝐴𝐴 (𝑡𝑡) (6.2)
𝑑𝑑𝑑𝑑 𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑
where 𝑓𝑓𝑜𝑜 is the scheduled grid frequency, and 𝑓𝑓(𝑡𝑡) is the grid frequency at any point in time.
23
7 Measurement of Response Times of the PV-BESS
System
The response time of the 430-kW PV plant for the active and reactive power set points was
measured in grid-connected mode during the peak PV production hour. The plant controller was
set to receive rectangular active and reactive power set points simultaneously, as shown in Figure
27. Before starting this test, the plant was curtailed for testing the active power response in both
the up and down directions. The plant was receiving rectangular active and reactive power set
points at different frequencies (0.1 Hz and 0.05 Hz, respectively). Several important observations
from Figure 27 are:
• The PV plant consisting of string inverters from different vendors is capable of following
both the active and reactive power set points with a high level of precision.
• The plant can follow the reactive power set points in both the kVAR injection and
absorption modes.
• An approximate delay of 300 ms–350 ms was observed until both the active and reactive
power reached the set point due to a combination of communication delays and internal
ramp limits in inverters.
• Active and reactive power can be controlled independently of each other and can follow
the set points with a high level of precision if the current limits of the inverters are not
exceeded.
In this test, both the active and reactive power of the PV plant were measured on the 13.2-kV
side of the PV plant transformer.
Figure 27. Measured PV plant response times

A number of tests for the BESS were conducted as well to measure the ability of the battery
system to follow the active and reactive power set points. The BESS response was tested in both
grid-following and grid-forming modes of operations. Results of the measured BESS responses
to the step changes in the active power set points are shown in Figure 28 and Figure 29 for grid-
following and grid-forming modes, respectively. In both tests, the BESS was commanded to
24
inject full rated power (1 MW) starting from zero. In grid-following mode, the measured power
reaches the set point within approximately 20 ms after the BESS controller receives the set point
from NREL’s SCADA. Similarly, in grid-forming mode, the BESS reaches the set point within
approximately 50 ms. The BESS power was measured on the 13.2-kV side of the transformer for
both tests.
Figure 28. Measured BESS response time in grid-following mode
Figure 29. Measured BESS response time in grid-forming mode

Additional response time tests were conducted for the BESS during this project. One example of the
BESS following rectangular active power set points in grid-following mode is shown in Figure 30.
Figure 30. BESS response to active power set points
25
8 Active Power Control Tests
To demonstrate several types of active power controls by the PV-BESS system, the NREL team
developed a modified real-time model of the Institute of Electrical and Electronics Engineers
(IEEE) 9-bus system and used it in the PHIL setup with the 7-MVA CGI (Figure 19). A
simplified single-line diagram of the 9-bus system implemented in the RTDS is shown in Figure
31. The real-time model has been modified, so different penetration levels of renewable
generation and PV-BESS can be modeled to show how the controls for reliability services by
PV-BESS systems can help improve grid stability and frequency response. Also, the model was
tuned to replicate frequency responses like that of the Western Interconnection using results of
the Western Electricity Coordinating Council Transmission Expansion Planning Policy
Committee 2022 positive-sequence load flow model at different penetration levels. This PHIL
setup interfaced with the NREL CGI and allows for exposing the PV-BESS system to grid
conditions measured at any bus in this RTDS model that represents an approximate 100-GW
power system. The model allows for scaling the capacity of the PV-BESS plant to any desired
level. Fast active power controls—such as inertial response, FFR, and PFR—by the PV-BESS
plants can be tested with this setup by exposing the system to frequency transients caused by the
drop of a large conventional power plant (approximate 3% of dispatched capacity). A test system
similar to this developed by NREL was used in a project that NREL conducted for the Pacific
Gas & Electric Company under the California Energy Commission’s Electric Program
Investment Charge during 2018 but with focus on BESS only [14]. Individual generation units in
the RTDS model can be dispatched to emulate different levels of instantaneous penetration by
inverter-coupled resources.
To evaluate the frequency stability impacts of the large-scale deployment of PV-BESS systems,
we assumed different PV-BESS plant capacities at different levels of inverter-based resource
(IBR) penetration in the test system, as shown in Table 2. The PV-BESS capacity was scaled up
in the model from the capacity of the real system at NREL (430-KW PV and 1-MW/1-MWh
BESS) to levels shown in Table 2. For each case, we evaluated the frequency response of the
system with PV-BESS plants providing various forms of active power controls.
Table 2. PHIL Scenarios
PV-BESS Level of IBRs (% of Load)

Capacity
15% 20% 40% 60%
Deployed in the
Model
0 GW Inertia, FFR, PFR Inertia, FFR, PFR Inertia, FFR, PFR Inertia, FFR, PFR
0.6 GW Inertia, FFR, PFR Inertia, FFR, PFR Inertia, FFR, PFR Inertia, FFR, PFR
26
Figure 31. PHIL setup using 9-bus power system model in RTDS
8.1 Inertial Response Tests

For a PV-BESS plant to provide inertial response like a rotating generator, its output power
needs to be modulated to inject/absorb active power proportional to the ROCOF. Inertia-like
response can be provided by a BESS only, by a PV plant only if it is curtailed during the
frequency event, or by both the PV and BESS; however, note that the maximum level of inertial
response by a PV-BESS plant at any time depends on the operating power level of the plant
components just before the frequency event occurs. We anticipate that in PV-BESS plants the
burden of inertia-like response will be in places mainly on battery storage and not curtailed PV.
So, in this experiment, we assumed that the BESS is the only provider of inertial response.
Detailed explanations of how the BESS was programmed to provide inertia-like response are
given in [11] and [14].
The ROCOF deadband implemented for this test was set at the 20-mHz/s level to avoid
unnecessarily triggering the inertial response by the BESS controller. The tests were conducted
for four aggregate PV-BESS capacities at each IBR penetration level. The frequency and BESS
response time series for each renewable penetration level are consolidated in Figure 32 a, b, c, d,
and e for 0%, 15%,20%, 40%, and 60% IBR penetration cases, respectively. The results show
continued improvements in the frequency nadir with increased aggregated capacity of PV-BESS
27
plants. Inertial response by the BESS improves the frequency nadir but does not impact the level
of the settling frequency because the PFR in the test system remains the same for all test cases.
For all cases, except for the 60% IBR penetration case, the inertial response by the PV-BESS
plant shifts the frequency nadir above the 59.5-Hz underfrequency load-shedding threshold,
In these tests, the initial BESS power was set to zero at the beginning of the fault, so the battery
system had enough headroom to inject the inertia power with a magnitude equal to its full rating
(1-MW battery at NREL scaled to the levels shown in Table 2). For this purpose, the inertial
constant, H, of the battery was scaled from H=125 s to H=50 s when going from lower to higher
penetration levels of IBRs.
Figure 32. Results of the PHIL tests for PV-BESS plants providing inertial response at different
penetration levels: (a) 0%, (b) 15%, (c) 20%, (d) 40%, and (e) 60%
28
The accuracy of the frequency measurements is very important for inertial control of IBRs. IBRs
should ensure that the frequency measurement and protection settings can ride through without
tripping because of phase jumps or other grid disturbances, such as fault events. Phase-locked
loops (PLLs) and synchronization techniques used in inverters are important for the reliable
operation of IBRs under dynamic and transient conditions in the grid. Different PLL algorithms
estimate grid frequency in different ways. Under sinusoidal and balanced grid conditions, they
all can be tuned to produce results that are similar to those shown in Figure 33. We modeled and
analyzed several PLL methods for comparison, including synchronous-frame PLL (SF-PLL), a
PLL method based on instantaneous active and reactive power theory (PQ-PLL), and a
sinusoidal signal integrator PLL (SSI-PLL). All these methods are described in detail in [15].
Under unbalanced grid conditions, however, or when voltage harmonics are present, additional
filtering delays are needed for PLL algorithms to accurately estimate both grid frequency and
ROCOF. An example of frequency estimation by all these PLL algorithms under 5% voltage
imbalance is shown in Figure 34, which emphasizes the need for additional filtering of the
frequency signal, and which will impact the effective magnitude of the inertial constant, H,
emulated by the BESS.
Figure 33. Comparison of different PLL algorithms under balanced grid conditions
Figure 34. Comparison of different PLL algorithms under 5% imbalance in grid conditions
The theoretical range of programmable H for a BESS can be very high, assuming that the
frequency and ROCOF can be estimated very fast, and the magnitude of the ROCOF is low. For
example, the calculated theoretical inertial constant that can be emulated by the 1-MW BESS at
zero power level for the U.S. Western Interconnection is shown in Figure 35 within the range
from H= 200 s–800 s; however, the practical range reduces to H= 50 s–125 s if the frequency
29
filtering delays are applied (a 100-ms delay was applied in the PHIL system). This range for
inertia constants is still much higher than one for conventional plants; however, we would like to
emphasize that this level of inertial response by the BESS highly depends on the initial
conditions of the battery, and care needs to be taken to have adequate power headroom for the
desired inertial response if the BESS is expected to provide such service at any time.
Figure 35. Calculated range of programmable H for NREL’s 1-MW/1-MWh BESS
8.2 Fast Frequency Response Tests

8.2.1 FFR-PHIL Simulations
FFR is another method to use PV-BESS systems to compensate for a sudden generation or load
loss. This can become a very efficient frequency response tool for system operators, but it
requires precise knowledge of the loss of magnitude so participating plants can be commanded to
change their power outputs accordingly. This method depends on the ability of the control
system to rapidly determine the magnitude of the loss and communicate the set point to the
controllers of PV-BESS plants. The required speed of the response (or how fast the PV, BESS,
or both deploy available reserves) will depend on the possible impacts on power system stability:
BESS and PV plants can deploy their available reserves very fast, which can cause unwanted
oscillations in the system. In some cases, FFR activation by PV-BESS plants can be based on
frequency thresholds, similar to underfrequency load-shedding schemes, or based on the
ROCOF. This will require determining the precise FFR magnitudes based on the measured
system frequency or the ROCOF by conducting detailed modeling studies. The main advantage
of FFR by PV-BESS plants is that it has the potential to reducing requirements for both primary
and secondary reserves due to the fast-acting nature of both resources, allowing for the fast
recovery of frequency during the initial stages of the frequency event. Further studies are needed
30
to understand this benefit in terms of a megawatt response comparison that is needed to restore
the frequency using a combination of primary and secondary controls compared to FFR.
For this use case, the BESS of the PV-plus-storage plant was programmed to deploy its available
power to compensate for the loss of generation. For the first series of FFR tests, a very fast 100-
ms response time for the BESS to provide FFR was implemented. Tests were conducted for four
BESS capacities at each renewable penetration level. Frequency and plant active response time
series for each renewable penetration level are shown consolidated in Figure 36 a, b, c, d, and e
for 0%, 15%, 20%, 40%, and 60% IBR penetration cases, respectively.
No PV-BESS No PV-BESS
a. b.
No PV-BESS No PV-BESS
c. d.
No PV-BESS
e.
Figure 36. Result of FFR PHIL tests at different IBR penetration levels: (a) 0%, (b) 15%, (c) 20%, (d)
40%, and (e) 60%
31
With such FFR, continuous improvements for frequency response can be observed at each
penetration level, shifting the frequency nadir well above the 59.5 underfrequency load-shedding
threshold for all sizes of PV-BESS plants. Only for extreme 60% penetration cases, a larger
amount of storage is needed to keep the nadir securely above the underfrequency load-shedding
level. Note that even without the BESS, the system can provide satisfactory frequency response
by conventional generation at lower penetration levels; however, it is conceivable that some
extreme conditions that were not envisioned in the study could result in unsatisfactory
performance. In this regard, the advanced FFR by the BESS can help provide improved
frequency response and reliability of the power system.
8.2.2 PMU-Based FFR Concept

Reliable FFR by inverter-coupled resources can be realized using the embedded network of PMUs
and associated equipment (data concentrators, SCADA, data communication equipment, and data
networks), as shown in Figure 37. Such a network can be used in a wide-area FFR control system
involving data from PMU units distributed across the network to help detect sudden changes in the
system, determine the magnitude of change, and reporting to all participating IBRs (in this case, to
all participating PV-BESS systems) to provide precision response for the rapid restoration of any
load-generation imbalances that might occur due to disturbances (Figure 38).
Figure 37. PMU networks embedded in the power system
32
Time stamp
Power System
PMU1
Time stamp
Time stamp PV-sotrage
Power and Voltage plant FFR, Oscillation Damping
PMU2 Contribution
Modulation Setpoint Setpoint
Delay
D PDC WADC Compensation
D
D Plant Controller
D Wide-area Stability
Control Center
PMUN
Distributed Phasor
Measurement Units
Figure 38. Concept of PMU-based wide-area reliability service by PV-BESS plants

Motivated by these considerations, the NREL team developed a PMU-based test bed for
characterizing and validating wide-area stability controls by IBRs, including FFR. The diagram of
this new fully operational test bed is shown in Figure 39. It consists of a number of GPS-
synchronized, SEL PMU-enabled relays installed on the medium-voltage grid across the Flatirons
Campus. The diagram in Figure 39 shows a portion of the test bed that has been specifically
designed to test wide-area stability services by PV-BESS plants. The test bed includes:
• PMUs installed on the medium-voltage sides of both BESS and PV plant controllers
• PMUs coupled with RTDS, so they can be virtually placed at any bus in a real-time
power system model
• RTAC systems
• SEL phasor data concentrator
• PMU fiber-optic communication network
• Network equipment (switches, routers, firewall devices)
• Fully integrated with the Flatirons Campus site controller and synchronized with custom-
made, advanced, medium-voltage DAS with 50-kHz sampling rate deployed at the site.
Figure 39. PMU-controls test bed developed at NREL's Flatirons Campus
33
This setup was used to characterize and quantify the responses and delays associated with PMU-
based controls using real equipment coupled with controlled, repeatable grid conditions emulated
using the 7-MVA CGI. During this project, the NREL team conducted characterization testing of
PMU-based controls under CGI-emulated voltage, phase angle, and frequency perturbation
injections through a RTDS model. Perturbations were captured by both medium-voltage DAS
and PMUs in a time-aligned manner to measure the response times and communication delays
for all components of the synchrophasor data network (Figure 39):
• PMUs to RTAC
• PMUs to phasor data concentrator
• Phasor data concentrator to RTAC.
The response of the system to various perturbations introduced by the CGI was measured and
characterized; the results are shown in Figure 40–Figure 43.
Figure 40. PMU-based system characterization test setup
Hz
Hz
Figure 41. Transfer function of PMU data flows to 5 % voltage magnitude injection by the CGI
34
Hz
Hz
Figure 42. Transfer functions of PMU data paths in response to the 0.5-Hz frequency injection by
the CGI
Hz
Hz
Figure 43. Transfer functions of PMU data paths in response to the 0.1-rad phase angle injection
by the CGI
It appears that the magnitude response of the whole PMU data flow path is very stable for update
frequencies up to 12 Hz (~80-ms update rate). Phase response has some delays at that frequency,
but it can be compensated. These types of response times make it possible to consider the use of
the proposed method to estimate load-generation imbalances at any bus of a power system and to
develop a control strategy to deploy ultrafast frequency response converter-based assets. The
objective is to prevent relatively large off-nominal frequency transients by counteracting the
impact of sudden imbalances on an electric grid. The implementation of this based on
experimental results and using a modified model of the IEEE 39-bus test system is described in
[20]. The diagram of the model is shown in Figure 44 and can be used to study the impacts of
FFR controls by PV-BESS systems on the frequency stability of the grid.
35
PV-BESS
Figure 44. Modified IEEE 39-bus system for optimally placed PMUs
The method also addresses the PMU placement problem for maximum system observability and
accurate detection of system imbalances. For this purpose, a solution was developed to ensure
that all positive-sequence bus voltages, and the currents leaving a bus, become available at a
phasor data concentrator facility by direct measurement and/or estimation [20]. One example of
simulation results for the 39-bus system under the same contingency event (the loss of a large
conventional unit) without and with the proposed FFR method is shown in Figure 45. The
proposed FFR method results in a better frequency and voltage response of the system compared
to a case with no PMU-based FFR controls. This approach will be investigated further by the
NREL team during future research activities.
Figure 45. Frequency and voltage response of 39-bus system: (left) no FFR and (right) with PMU-
based FFR
8.3 Frequency Droop Tests

8.3.1 Frequency Droop Test Results by PV-BESS System
Adequate droop response in combination with inertia is an important reliability factor for any
power system and needs to be maintained at levels prescribed by NERC BAL-003-1 for
interconnections and individual balancing areas [24]. The most common droop setting used in
36
many power systems is 5%, but in some cases a more aggressive 3% droop is used as well. For
example, the Western Electricity Coordinating Council governor droop criterion allows
individual generator droop settings within a 3%–5% range [21]. Frequency droop control
assumes a linear relationship between power and frequency (with some small deadband). For
example, a 5% droop means that a 5% change in frequency would result in a 100% change in
power, a 3% droop means that a 3% change in frequency would result in a 100% change in
power, etc. Such a linear relationship is theoretical, though, and with real governors, it has many
nonlinearities due to various types of control delays, unintentional deadbands, and physical
characteristics of prime movers. The ability of large utility-scale PV plants to provide droop
response was demonstrated in [4]. For BESS, as was shown by previous testing at NREL [11],
the relationship between power and frequency is essentially linear due to the fast response time
(less than 50 ms) of the battery inverters; therefore, both PV plants and BESS can provide PFR
with much higher levels of precision and speed for a wide range of droop settings (e.g., 1%–5%).
However, the ability of BESS to provide adequate droop response is also subject to its initial
conditions. For the same per unit of power and depending on the initial conditions, the BESS can
provide more benefits to the system’s PFR than conventional generators. A theoretical
comparison of the frequency droop response of a BESS and a synchronous generator-based unit
is shown in Figure 46. In this case, the droop response by the BESS component of the PV-BESS
plant is drawn for a case when the BESS happened to be at zero power, so it has equal headroom
to inject or absorb power depending on the nature of the transient event (underfrequency or
overfrequency). For a PV plant to provide a droop response, it needs to be curtailed to have
enough headroom to increase its production to the level determined by the droop settings during
underfrequency events. For overfrequency events, a PV plant can provide an aggressive droop
response. It is anticipated that the prevailing role of the PFR provision of PV-BESS plants will
be based on the BESS components providing such services instead of curtailing the PV
components of PV-BESS plants.
Figure 46. Frequency droop by a PV-BESS system
37
Several experiments were conducted on the 430-kW PV plant to demonstrate its ability to
provide droop response when exposed to real frequency events emulated by the CGI. The PV
PPC did not have built-in control for droop response. For this purpose, the external NREL
controller was used to command the set points to the PV PPC based on the measured frequency
and droop settings, thus introducing an additional delay from 50 ms–100 ms in response. But
even then, the PV plant demonstrated good PFR for both the 5% and 3% droop settings, as
shown in Figure 47 and Figure 48, respectively. The measured ability of the BESS to provide
droop response is shown in Figure 49.
Figure 47. PV plant providing 5% droop response
Figure 48. PV plant providing 3% droop response
38
Figure 49. BESS providing frequency droop response
8.3.2 Comparison of Frequency Response Methods

In this section, we analyze the combined results of several PHIL tests to better understand and
compare system performance for each method. For this purpose, we selected the most extreme
case—corresponding to 60% penetration by IBRs in the power system emulated in PHIL (Figure
31). Figure 50 shows the results for the following test cases:
• No PV-BESS present in the PHIL test case

• Equivalent of 1,800-MW PV-BESS providing inertial response only
• Equivalent of 1,800-MW PV-BESS providing FFR with 0-ms delay
• Equivalent of 1,800-MW PV-BESS providing FFR with 1-s delay
• Equivalent of 1,800-MW PV-BESS providing inertia plus FFR with 1-s delay.
Apparently, the best performance is achieved for the use case with zero-delay FFR (Td=0-s),
indicated by the yellow trace; however, as discussed earlier, this is not a realistic situation
because actual delays associated with PMU-based FFR controls will exist. After introducing a
conservative 1-s delay, the best-performance among all remaining use cases is demonstrated by
the use case of “Inertia plus FFR with 1-s delay” (green trace in Figure 50). Even at such a high
level of renewable penetration, the combination of inertial response and FFR with a 1-s delay
produces the best performance, keeping the frequency nadir securely above underfrequency load-
shedding level.
39
No PV-BESS
Figure 50. Comparison of impacts by inertial, FFR, and PFR controls
8.4 AGC Tests

The ability of a large 300-MW PV plant to participate in AGC was successfully demonstrated in
[4], with results indicating that PV generation can participate in frequency regulation with much
faster speed and higher precision of response than conventional resources when following an
AGC signal provided by CAISO. In this section, we show some results of a 430-kW PV plant
operation when following the set points generated by the NREL controller. These set points were
calculated from the historic 4-s Public Service Company of Colorado area control error time
series. The historic area control error data were scaled down by the NREL controller to a level
corresponding to the rating of the 430-kW PV plant. The results of such testing are shown in
Figure 51 and Figure 52, each for 1 hour of operation. These figures show that the curtailed plant
(10% below available peak power) is fully capable of following the AGC signal with a high
degree of precision when there is enough headroom between the plant’s operational point and the
peak available power. During fast cloud movements, however, the plant output can decline
rapidly (initial portion of time series in Figure 51). To avoid these types of situations, either a
control error correction is needed by using energy storage or a fast real-time peak power
estimation method is needed to constantly scale the AGC set points based on the available plant
power. Such a method was developed during this project, and it is described in detail in Section
13 of this report.
During less variable conditions, the AGC performance of the plant is much better, as shown in
Figure 53 and Figure 54. The ability of the BESS to follow the same AGC set point is shown in
Figure 55. The BESS can provide this type of response with a very high degree of precision
because it is not subject to resource variability like the PV plant.
40
Cloud impact on AGC performance
Figure 51. PV plant following the AGC set point under high-variability conditions
Cloud impact on AGC performance
Figure 52. Correlation between the measured PV plant power and the AGC set points under high-
variability conditions
Figure 53. PV plant following the AGC set point under low variability conditions
41
Figure 54. Correlation between the measured PV plant power and the AGC set points under low
variability conditions
Figure 55. BESS following the AGC set points—grid-following mode
8.5 Dispatchability of PV-BESS Plant

The combined PV-BESS output can be shaped based on various revenue optimization
algorithms, changing electricity prices, dispatch signals from the system operator, and types of
reliability services the PV-BESS pant is providing. The NREL PV-BESS system combines an
AC-coupled PV array (rated at 430 kW) and a BESS (rated at 1 MW/1 MWh), which creates an
opportunity to test algorithms for multihour PV energy shifting and scheduling in addition to
output profile shaping with a large degree of flexibility. The plant controller developed at NREL
can provide dispatchable operation without PV curtailment while operating within the energy
capacity and power rating constraints of the PV-BESS plant. It can be coupled with any external
economic dispatch and optimization controller linked to the solar resource forecast system, thus
creating the unique platform for industry partners to test, improve, and adapt PV-BESS controls
for various market rules, reliability, and resilience metrics. One example of output-shaping
operation is shown in Figure 56 for a PV-BESS system with measured PV power output under
highly variable solar resource conditions present at NREL’s Flatirons Campus. In this example,
42
the BESS is controlled to provide 3-h morning and 5-h evening ramps as well as steady
baseload-like operation during the middle of the day. The initial SOC of the BESS happened to
be at 70%, and during the day, the BESS is mainly charging (from PV only) in the morning and
discharging in the evening while balancing the PV variability throughout the whole day, with the
end SOC near 60%.
Figure 56. Example of production profile-shaping PV-BESS system

Other examples of such dispatchable operation are shown in Figure 57 (a), (b), and (c), with the
BESS shifting PV energy into the evening hours following the commanded trapezoidal profile.
Figure 57 shows the case when the PV-BESS system is commanded to provide a firm block of
energy similar to baseload units.
43
Figure 57. Examples of dispatchable operation by the PV-BESS system
In all these cases, the BESS—because of its larger power rating compared to the PV plant—has
sufficient headroom to modulate its power above the dispatched level for other services, such as
primary response and up-regulation. Any desired aggregate profile of the PV-BESS plant can be
achieved similar to the cases shown in Figure 56 and Figure 57 as long as all operational and
economic constraints are respected.
8.6 Ramp-Limiting Tests

In the bulk power system, short-term variability and instantaneous ramp rates are largely
mitigated by the geographic diversity of the solar resource [22]. Despite the attractiveness of a
44
smoothed and shifted output from each PV plant using energy storage, this approach results in a
significant decrease in the efficiency of the entire system and eliminates the benefits of resource
aggregation. In fact, operational integration between energy storage and any individual or groups of
PV plants would be nonoptimal and would likely result in simultaneous charging and discharging
with zero net benefit to the system [22]. From this perspective, we did not consider the variability-
smoothing controls an important topic in the context of utility-scale PV operation (the same is
true for utility-scale wind as well). In small, stand-alone microgrids or small island systems that
do not have the advantage of large spatial resource diversity, however, the variability smoothing
of individual plants can still be a useful service. For this purpose, we developed and
demonstrated some of the controls that can be used for variability-smoothing and short-term
ramp-limiting service by PV-BESS plants. One example of a ramp-limiting test for a PV-BESS
plant is shown in Figure 58. In this case, the BESS was controlled to provide ramp limiting for
two PV plants at NREL: the 430-kW First Solar PV plant and a 1-MW MaxGen commercial PV
plant. 2
Figure 58. PV-BESS system in ramp-limiting mode (50-kW/min limit) providing PFR
2
The MaxGen (formerly SunEdison) PV plant is a commercial 1-MW plant that operates under a PPA with Xcel
Energy. NREL does not have control over the inverters of this plant. During most of 2019 and 2020, this plant was
malfunctioning and was not used for testing BESS controls related to variability smoothing except for short periods
of time when the plant operated properly. (It has problems with the tracking system and inverters.)
45
At the beginning, the total combined production from the two plants was near 1,000 kW and
stable because of clear-sky conditions. The clouds started moving over the plant within
approximately 10 min of the beginning of the experiment. The BESS was programmed to
provide a 50-kW/min aggregate ramp rate for both plants combined (orange trace), ensuring
significant short-term variability smoothing. At t≈1,900 s, a frequency event was simulated by
the CGI, so in addition to ramp rate control, the inertial response (H=10 s) and primary
frequency control of the BESS became active. From this point, the PV-BESS plant was
providing two services at the same time (stacked services): a 50-kW/min ramp rate limit and
frequency response. The BESS SOC was declining steadily during the whole test because the
BESS was mostly discharging to provide the desired ramp limit.
Another example of a ramp control test is shown in Figure 59. The BESS was set to provide a
100-kW/min ramp limit. This made the total PV-BESS plant output during the periods of cloud
variability (time periods near t=60 s and t=150 s). At t=160 min, a large cloud moved in, rapidly
collapsing the PV production in only 30 s (yellow trace). The BESS ramp-limiting controls
started injecting power to delay the total plant output reduction. This can be a valuable service in
some microgrids, allowing sufficient time for backup generation to start and synchronize.
Figure 59. Ramp-limiting control of the PV-BESS plant (100 k/min)

The ability of the PV-BESS to provide smoothing services was also tested when the BESS was
operating in grid-forming mode. The results of one interesting experiment are shown in Figure
60. In this case, we modulated the output of the PV plant to create a fully controlled variability
pattern to explore the response times of the BESS in grid-forming mode when controlled from
46
the PV-BESS plant controller. In this example, a 60-s averaging filter for PV power was applied
in the controller to operate the battery based on a smoothed PV time series. Figure 60 shows that
the BESS provided excellent smoothing under this highly variable PV production.
Figure 60. Example of a variability-smoothing test by the BESS in grid-forming mode
8.7 Oscillation Damping Controls

Power system oscillations triggered by disturbances such as generation outages or transmission
line faults can result in system separation because of large swings in the active power flow over
intertie lines, potentially leading to major blackout events [44]–[47]. It is critical for the reliable
operation of a power system that its interarea oscillation modes remain sufficiently damped
under all resource dispatch scenarios. Usually, the interarea modes are damped by installing
power system stabilizers (PSS) in the control system of synchronous generators. There is
increasing concern that the displacement of synchronous generators by IBRs might negatively
impact the damping of the interarea modes. For example, recent studies by ERCOT have shown
that it will be important for IBRs to provide PSS-type damping for the reliable operation of the
Texas Interconnection as the IBR penetration increases to higher levels [47], [48]. IBRs can
provide PSS-like functionality and provide damping to the interarea oscillation modes by
modulating their active power output. In this project, we tested the ability of the 430-kW First
Solar PV plant to modulate its active power output to evaluate the feasibility of incorporating the
PSS-like functionality in the PV plant. The details of this test are provided in the following.
To test the ability of the PV-BESS plant to modulate its active power output, the plant output
was slightly curtailed by generating an active power reference slightly less than the maximum
available power. In addition to the slow-varying active power reference, sinusoidal modulation
of different frequencies was added to the reference. A special control algorithm for testing such
modulation characteristics for PV-plus-storage plants was developed in National Instruments
LabVIEW software and implemented in a real-time PXI controller during this project (the main
user interface is shown in Figure 61).
47
Figure 61. Main control screen for the P-Q set points and oscillation control
The results of some sinusoidal active power modulation for both the PV plant (all inverters) and
the BESS inverter operating in grid-following mode are shown in Figure 62 and Figure 63,
respectively. The frequency sweep was conducted for the sinusoidal modulation frequencies in
ranging from 0. Hz–1.5 Hz.
Figure 62. PV power plant following the sinusoidal set point
Figure 63. BESS in grid-following mode following the sinusoidal set point
48
Figure 64 and Figure 65 show the active power reference and the plant output for different
frequencies of the sinusoidal modulation input individually for the PV and BESS plants. The
data captures shown in the figures were used to develop the transfer function gain from the
modulating active power reference to the plant output. A similar test was also conducted for a
GE 1.5-MW wind turbine.
Figure 64. Time-domain captures for fast Fourier transform analysis for First Solar’s 430-kW PV plant
Figure 65. Time-domain captures for the fast Fourier transform analysis for the BESS
Figure 66. Gain and phase response bode plot for different resources
49
Figure 66 shows the transfer function gain from the active power reference to the active power
output for the GE 1.5-MW wind turbine, First Solar’s 430-kW PV plant, and the 1-MW/1-MWh
BESS. It is evident that all three sources can provide active power modulation to near 1 Hz. The
PV plant exhibits the maximum bandwidth for the active power modulations—inverter controls
can be further tuned to improve performance. The phase response in Figure 66 exhibits the
typical behavior of a delay—the phase response can be further improved by reducing the number
and amount of delays from the active power reference to the active power output. Future work
will implement PSS-like functionality in a PV plant and a BESS, and the performance will be
tested using PHIL experiments and later in the field.
8.8 Modeling of POD by PV_BESS plants

Multiple approaches have been proposed for damping interarea oscillations. One method using
energy storage systems installed in areas that can oscillate against one another and with real
power injection at each location using a damping control system is described in [25]. Here, we
use the same approach to demonstrate the ability of a PV-BESS plant to provide interarea
oscillation damping using the modified two-area power system described in [26]. We show how
the PV component of PV-BESS plants can provide such control because the provision of such
control by energy storage was investigated in [25]. The damping controller located in the PV-
BESS plant in Area 1 controls the power of the curtailed PV plant proportional to the frequency
difference between areas 1 and 2 (Figure 67). Generation in both areas is dominated by
combined-cycle gas turbine generation.
Area 1 power, P1(t)
Area 1 f1 f2 Area 2
230 kV
GT1 GT3
GT2 GT4
PV-BESS
C1 C2
f1(t) 200 MVAR 350 MVAR f2(t)
Pod (t)
R-L Load 1 R-L Load 2

967 MW 1767 MW
100 mVAR 100 mVAR
Plant
POD Comm delay Td
control
Figure 67. Modified two-area system model

Using the same control law described in [25], the output power of the PV plant can be modulated
based on the following:
𝑃𝑃𝑝𝑝𝑝𝑝 (𝑡𝑡) = −𝐾𝐾𝑑𝑑 (𝑓𝑓1 (𝑡𝑡) − 𝑓𝑓2 (𝑡𝑡 − 𝑇𝑇𝑑𝑑 )) (8.1)
where 𝑃𝑃𝑝𝑝𝑝𝑝 (𝑡𝑡) is the commanded power of the PV plant, 𝑓𝑓1 and 𝑓𝑓2 are the frequencies measured in
areas 1 and 2, 𝐾𝐾𝑑𝑑 is the control gain [MW/mHz], and 𝑇𝑇𝑑𝑑 is the communication delay.
50
Without any damping control (no PSS control implemented in the gas generation models), after a
small change in load, the system quickly falls into an unstable oscillatory state, as shown in
Figure 68. The upper plot in Figure 68 shows frequency in areas 1 and 2, calculated from the
rotational speed of the generation plants. The second plot shows differences in frequency
between areas 1 and 2. The third plot shows instantaneous power flow through the tie-line, and
the bottom plot shows instantaneous power production of the 120-MW PV plant that was
curtailed to the 100-MW level (16% curtailment).
60.02
60
FREQUENCY (Hz)
59.98
59.96
59.94
59.92
59.9
59.88
0 5 10 15 20 25 30 35 40 45 50
TME (sec)
f1 f2 f3 f4
0.1
0.05
FREQUENCY (Hz)
0
0 5 10 15 20 25 30 35 40 45 50
-0.05
-0.1
TIEM (sec)
df1 df2
Tieline Power
800
600
400
POWER (MW)
200
0
-200 0 5 10 15 20 25 30 35 40 45 50
-400
-600
-800
TIME (sec)
PV plant
100.003
100.002
POWER (MW)
100.001
100
99.999
99.998
99.997
0 5 10 15 20 25 30 35 40 45 50
TIME (sec)
PV
Figure 68. Undamped oscillations triggered by a small change in load
51
The damping behavior of the system changes dramatically after the above control law is applied
to the PV plant controller in Area 1, as shown in Figure 69a. Even with a 200-ms communication
latency used in this case, the oscillations are effectively damped within 20 s–25 s after the event.
With less latency, the damping will be more efficient. If a similar PV plant located in Area 2 had
commanded the power set points in accordance with the same control law (with reversed sign),
then the damping would be even more efficient.
60.02
60
FREQUENCY (Hz)
59.98
59.96
59.94
59.92
59.9
59.88
59.86
0 5 10 15 20 25 30 35 40 45 50
TME (sec)
f1 f2 f3 f4
0.04
0.02
FREQUENCY (Hz)
0
0 5 10 15 20 25 30 35 40 45 50
-0.02
-0.04
-0.06
TIEM (sec)
df1 df2
Tieline Power
460
450
440
POWER (MW)
430
420
410
400
390
380
0 5 10 15 20 25 30 35 40 45 50
TIME (sec)
PV response
110
POWER (MW)
105
100
95
90
0 5 10 15 20 25 30 35 40 45 50
TIME (sec)
PV
Figure 69. Oscillations damped by POD control applied to the PV plant (Td=200 ms)
One important consideration about the provision of power oscillation damping (POD) by PV
plants is that they need to have headroom to modulate their output in both directions. To avoid
continuous curtailment, a method described in Figure 70 can be used. The PV plant can operate
without curtailment at peak power until the interarea oscillations are detected by the POD
52
controller. At this instance, the controller quickly curtails the plant to a desired level before it
starts modulating the output of the plant in accordance with the POD control law. This method
has not been implemented in the current work, but it will be demonstrated by the NREL team in
the near future.
Figure 70. PV plant providing POD without continuous curtailment
53
9 Reactive Power Control Tests
The voltage of the North American bulk system is normally regulated by generator operators,
which are typically provided, along with voltage schedules, by transmission operators [27]. The
increasing penetration levels of variable wind and solar generation have led to the need for them
to contribute to power system voltage and reactive regulation because, in the past, the bulk
system voltage regulation has been provided almost exclusively by synchronous generators.
According to the Federal Energy Regulatory Commission’s Large Generator Interconnection
Agreement [27], the generally accepted power factor requirement of a large generator is ±0.95.
In conventional power plants with synchronous generators, the reactive power range is normally
defined as dynamic, so synchronous generators must continuously adjust their reactive power
production or absorption within a power factor range of ±0.95. For PV power plants, the reactive
power requirements are not well defined. Federal Energy Regulatory Commission Order 661-A
is applicable to wind generators but sometimes applies to PV plants as well. It also requires a
power factor range of ±0.95 measured at the POI and requires that the plant provide sufficient
dynamic voltage support to ensure safety and reliability. (The requirement for dynamic voltage
support is normally determined during interconnection studies.) Utility-scale wind power plants
are designed to meet the ±0.95 power factor requirements. The common practice in the PV
industry is to configure PV inverters to operate at a certain power factor needed to cover the
reactive losses and then follow the utility’s reactive power or power factor set point (often at
unity power factor). It is expected that similar interconnection requirements for power factor
range and low-voltage ride-through will be formulated for PV in the near future. To meet this
requirement, PV inverters must have MVA ratings large enough to handle full active and
reactive current.
The steady reactive power capabilities of the 430-kW PV plant and 1-MW BESS were measured
in CGI-connected mode to avoid the voltage stability impact on NREL’s medium-voltage, Xcel
Energy-connected grid. The CGI provides full isolation for reactive power flows between the test
article bus and the regular grid bus; therefore, conducting such experiments on the CGI is a very
safe and controlled way to characterize the reactive capabilities of multimegawatt inverters. In
many cases, conducting such experiments in the field is not possible because of strictly defined
under- and overvoltage limits by utilities.
The measured steady-state reactive power capability of the 430-kW PV plant is shown in Figure
71. The voltage on the low-voltage terminals was set at a 600-V nominal level by the CGI.
Despite commanding a full two-quadrant reactive power profile (blue trace), the PV plant,
consisting of six string inverters, was capable of delivering only the reactive power profile
depicted by the orange trace. This is because of a -0.8…0.8 power factor limit set by the inverter
vendors.
54
Figure 71. Measured reactive power characteristics of the PV plant (all inverters)
The P-Q capability of NREL’s 1-MW/1-MWh BESS—consisting of LG Li-ion batteries and an
SMA 2.2-MVA, 400-V AC inverter/charger with a 1.1-MVA 13.2-kV/400-V transformer—was
verified using the experimental setup shown in Figure 19. The BESS inverter’s full four-
quadrant, steady-state P-Q characteristic was tested in CGI-connected mode to avoid impacts on
NREL’s medium-voltage, Xcel Energy-connected grid. The inverter was commanded to use
various combinations of active and reactive power set points to cover the whole range of P-Q
operation. The results of one such test are shown in Figure 72.
The P-Q characteristic of the BESS system was measured on the medium-voltage side (or CGI
side) of the BESS transformer as well. A comparison of both P-Q characteristics is shown in
Figure 73. The shift between the two is caused by a 6% impedance of the BESS transformer and
some reactive losses in the 100-meter underground collector line. NREL is developing a control
to compensate for these reactive losses so the reactive power can be accurately controlled on the
medium-voltage side of the BESS transformer. The inverter limits only Pmax and Qmax at the 1-
MW and 1-MVAR levels accordingly, but it does not limit the maximum apparent power, Smax,
which is expected to be 1 MVA. Instead, the measured the P-Q characteristic approaches the
square shape, as shown in Figure 72. Because of this characteristic, care must be taken not to
exceed the Smax set point for inverter transformer protection (the 400-V/132-kV step-up
transformer is rated at 1.1 MVA).
The overall combined ideal reactive power capability measured for the 430 kW PV plant and the
1-MW/1-MWh BESS is shown in Figure 74. As expected, the aggregate reactive power
capability of this hybrid PV-plus-storage plant is larger than the reactive power capability of
individual components. The ideal shape, assuming a low-impedance connection and steady grid
voltage, can change depending on the characteristics of each POI.
55
Figure 72. Measured reactive power capability of the BESS system
Figure 73. Reactive power capability measured on the low-voltage (SMA) and medium-voltage
(CGI) sides of the BESS transformer
Combined P-Q characteristic
Figure 74. Combined reactive power capability of the PV-BESS plant
56
10 Impedance Characterization of the PV-BESS Plant
Impedance-based stability analysis studies have proven effective for the evaluation and
mitigation of control interactions in wind and PV power plants, including subsynchronous and
super-synchronous resonance conditions [50]. Transmission system operators have started
demanding that manufacturers provide electromagnetic transient PSCAD models of IBRs to
evaluate different stability problems, including low- and high-frequency control interactions,
interarea oscillations, and frequency stability. The impedance measurement of wind turbines and
PV inverters can be used not only for stability studies but also for the high-fidelity validation of
electromagnetic transient models, which can be used for grid integration studies.
In this project, we performed the impedance characterization of a 2.2-MVA SMA inverter

interfaced with a 1-MW/1-MWh BESS. We also performed the impedance characterization of
inverters interfacing First Solar’s 430-kW PV plant. In addition to the impedance measurements,
we performed a PHIL experiment to simulate the operation of the First Solar plant under a weak
grid condition. Impedance measurements were used to predict the oscillatory instability observed
in the PHIL experiment for the operation of the PV plant under a weak grid condition. Details of
the impedance measurement and PHIL experiments for weak grid operation are discussed in this
section.
The impedance response of an inverter can be measured by injecting voltage or current

perturbations at different frequencies from its terminals superimposed on the voltages and
currents at the fundamental frequency. The 7-MW grid simulator, called the CGI, at NREL’s
Flatirons Campus was used for the impedance characterization of the 2.2-MVA battery storage
inverter and the PV inverters interfacing First Solar’s PV plant.
10.1 Impedance Measurement System

This subsection describes different systems used for the impedance characterization of the PV-
BESS plant. More details on the impedance measurement system can be found in [51].
10.1.1 Grid Simulator

The CGI grid simulator at NREL is rated for 13.8-kV, 7-MVA continuous power, and it is used
for injecting voltage perturbations at the terminals of the inverters at frequencies ranging up to 3
kHz for impedance characterization. The control system of the CGI is coupled with an RTDS,
which can provide dynamic voltage references for all three phases and the neutral wire for
different grid integration tests; the desired perturbations for the impedance measurement of the
2.2-MVA BESS inverter and the PV inverters of the 430-kW PV plant were configured inside
the RTDS.
57
10.1.2 GPS-Synchronized Medium-Voltage Measurements
Figure 75. Hardware in each medium-voltage DAS sensing node: (left) voltage and current
sensors; (right) data conditioning built around the National Instruments’ cRIO-9030 platform.
Photos by NREL
The admittance measurement test bed for utility-scale wind turbines and PV inverters at NREL
has a medium-votlage DAS, which is a distributed network of GPS-synchronized measurement
nodes sampling three-phase voltages and current waveforms at 50 kHz with 24-bit resolution.
Raw 50-kHz voltage and current waveform captures can be triggered by programmable
thresholds, supervisory control and automation systems, or asynchronously by user interface, and
they can be stored locally for post-event downloading, alignment, and processing. The system is
implemented on a National Instruments (NI) CompactRIO (cRIO) platform, and it uses onboard
field-programmable gate array hardware for synchronization and processing functions. Figure 75
shows the hardware in a sensor node, including capacitive voltage sensors 3; Rogowski coil-based
current sensors 4; and a GPS-synchronized, signal-conditioning unit built around the NI cRIO-
9030 platform. Medium-voltage DAS nodes are located at different points in the test bed, and
they allow for the GPS-synchronized measurement of voltages and currents during the injection
of perturbations to obtain impedance responses over a broad frequency range.
10.1.3 Impedance Measurement System

To enable the admittance measurement of an inverter, all the components previously described
are integrated, as shown in Figure 76. The CGI is controlled by an RTDS using a fast 2-Gbit/s
optical fiber interface. The high-bandwidth interface and delay compensation techniques
implemented in the CGI control system allow for the injection of perturbations over a wide
spectrum of frequencies, from 0.1 Hz to 3 kHz.
3
Jomitek – Power Solutions, “The Jomitek v3 voltage sensor,” Datasheet, 2012.
4
Powertek – RCTi current sensors, “RCTi single-phase current transducer,” Datasheet. [Online]
www.powertekuk.com/rcti-single-phase-current-transducer.pdf, accessed on July 17, 2019.
58
Figure 76. Impedance measurement system diagram
10.2 Impedance Measurements

Figure 77 shows the components of the PV-BESS plant whose sequence impedance responses
are measured using the impedance measurement system described in the previous subsection.
For the positive-sequence impedance measurement of a test article, the CGI injects positive-
sequence voltage perturbations superimposed on the nominal fundamental trajectory. The
measured positive-sequence impedance responses of the 1-MW/1-MWh BESS plant for different
magnitudes of the injected voltage perturbations are shown in Figure 78. The positive-sequence
impedance responses of the two groups of string inverters (4 kW x 40 kW and 2 kW x 125 kW)
at the front end of First Solar’s 430-kW PV plant are shown in Figure 79.
The impedance responses of the BESS and PV plants, shown in Figure 78 and Figure 79,
respectively, can disclose many important characteristics of inverters interfacing these plants that
can be used both for the stability and for the high-fidelity model validation. For example, the
impedance responses in Figure 78 and Figure 79 reveal that the current control bandwidth of the
BESS inverter is near 100 Hz and that of the string inverters in First Solar’s PV plant is near 2
kHz. This is inferred from the fact that the impedance response of an inverter exhibits inductive
and capacitive behavior, respectively, above and below the current control bandwidth of the
inverter. The capacitive behavior of the inverter’s output impedance below its current control
bandwidth is because of the integral gain of the current controller, whereas the inductive
behavior of the inverter’s impedance response above the current control bandwidth is because of
the phase reactors after the three-phase bridge in the inverter.
59
Figure 77. Inverters for impedance characterization: the (a) 2.2-MVA inverter interfacing the 1-
MW/1-MWh BESS and the (b) 430-kW PV plant with six string inverters. Photos by NREL
Figure 78. Positive-sequence impedance response of the 1-MW/1-MWh BESS plant measured for
different magnitudes of the voltage perturbation injected by the CGI for the impedance
measurement
60
Figure 79. Positive-sequence impedance response of two sets of string inverters (4 kW x 40 kW
and 2 kW x125 kW) interfacing First Solar’s 430-kW PV plant
Because the impedance measurement thoroughly characterizes the behavior of a device from the
grid perspective, it can be used for characterizing new technologies developed for improving grid
resilience. For example, recently there has been a lot of interest on grid-forming inverters to
substantially increase the penetration of IBRs. Different control methods—including droop, virtual
oscillators, and virtual synchronous machine-based approaches—have been investigated for
implementing the grid-forming functionality in inverters. Impedance characterization can serve as
an excellent platform for comparing these methods and evaluating their impacts on grid stability.
Figure 80 compares the positive-sequence impedance response of the 1-MW/1-MWh BESS plant
shown in Figure 77 (a) when it is operated in standard grid-following and the recently developed
grid-forming control modes. The peaking behavior of the impedance response near the
fundamental frequency for the grid-following mode shows that the inverter is being controlled as a
current source. On the other hand, the dipping of the impedance response near the fundamental
frequency for the grid-forming mode shows that the inverter is being controlled as a voltage
source. More importantly, the phase response of the impedance stays relatively closer to zero and
within ±90 degrees at most frequencies when the BESS inverter is operating in the grid-forming
mode; this signifies that the grid-forming mode provides higher positive damping than the grid-
following mode, which can avoid potential stability and control interaction problems.
10.3 Impedance-Based Prediction of Oscillatory Instability During

Weak Grid Operation
Fast and complex controls of inverters can lead to unstable interactions between a PV plant and
the grid as well as among the inverters of a PV plant. It is particularly difficult to maintain the
stability of an inverter when it operates under a weak grid condition. Different control systems
inside an inverter—including PLLs, current controllers, and DC-bus voltage controllers—can
interact in an undesired manner, leading to an unstable operation and damage or tripping of the
inverters [52], [53].
61
Figure 80. Comparison of positive-sequence impedance response of the 1-MW/1-MWh BESS plant
when it is operating in grid-following and grid-forming control modes
Based on the impedance response of the PV plant inverters shown in Figure 80, below the current
control bandwidth of 2 kHz, the impedance response is capacitive. The capacitive response of the PV
inverters can interact with a weak inductive grid and can create an unstable resonant mode. To
demonstrate the unstable operation of First Solar’s PV plant under a weak grid operation, a set of
PHIL experiments were performed by simulating a weak inductive grid using the CGI grid simulator.
Figure 81 shows the impedance-based stability analysis for the operation of the PV plant under a
weak grid condition. Figure 81 (a) compares the positive-sequence impedance responses of the
PV plant and the grid simulated inside the CGI. The impedance analysis predicts an unstable
super-synchronous resonance at 554 Hz. The inverter output currents in Figure 81(b) confirm the
presence of an unstable resonance predicted by the impedance analysis. This shows the value of
the impedance characterization of the PV inverters to evaluate their stability properties during
operation under different grid conditions.
a) b)
Figure 81. Impedance-based prediction of instability of First Solar’s 430-kW PV plant during
operation under a weak grid condition: the (a) impedance analysis comparing the impedance
responses of the PV inverters with the grid impedance and the (b) PV plant output current during
instability
62
11 Black Start of PV Plant with BESS
11.1 Black-Start Experiment with CGI
The first black-start experiment for the PV power plant was performed in 2019 using the CGI as
a black-start grid-forming resource. This was done because the black-start controls for the SMA
inverter of the BESS were not available until March 2020. In this first black-start experiment, the
CGI increased its terminal voltage from 0 kV to 13.2 kV with a slow ramp to avoid inrush
currents in the PV plant transformer. The total duration of the black-start process in this mode
was approximately 9 min. This was because the programmed delays in both the 45-kW and the
125-KW inverters were near 5 min.
A number of black-start experiments for the 430-kW PV array using the CGI as a black-start
resource were conducted in 2019. Results of a few such experiments are shown in Figure 83. The
PV plant demonstrated robust black-start performance during all experiments.
63
Slow increase of voltage to avoid inrush
currents in PV plant transformer
Two 125 kW inverters

reached full
production
7-min wait time by PV plant PPC after the voltage is

restored before starting production
20s 43s 40s
kW, KVAR
All four 45 kW
inverters reached full
production
Figure 82. Black start of PV system with the CGI
Figure 83. Series of black-start experiments using the CGI
64
11.2 Black-Start Experiment with BESS
After the completion and commissioning of the black-start controls, new experiments were
conducted using the BESS as a black-start resource for the PV plant. The experiment setup is
shown in Figure 84. In this setup, medium-voltage side transformers for both the PV plant and
BESS are wired as delta, so the whole islanded system does not have a ground reference. To
introduce a grounding reference, a grounding transformer was installed in the system, as shown
in Figure 84. The experiment was conducted with all assets connected to the CGI bus (the CGI
was de-energized and disconnected), so the system was fully isolated from the real grid.
Firs t Solar PV plant – 430 kV
Figure 84. Test setup for the black start of the PV plant with BESS
This experiment was conducted with the BESS set in two different voltage control modes:
• Soft black start: The BESS inverter ramped its voltage from zero to the full level during a
200-ms time period to minimize the inrush current in the PV plant transformer, which
was connected to the circuit during the whole process
• Hard start: The PV plant transformer was not connected until the BESS inverter reached
full voltage, and then the switch was closed to energize the transformer with the full
inrush current.
11.2.1 Soft Black Start

Results of the soft black start for the 430-kW PV plant using the 1-MW/1-MWh BESS as a
black-start resource are shown in Figure 85. The SMA inverter of the BESS increased its voltage
with a 200-ms ramp to minimize the inrush currents in the PV plant transformer, as shown in
Figure 86. The peak BESS current during the start was only at 5% of the inverter rated current.
After the transformer was energized, the initialization timers was triggered in the inverters (60 s
for the 45-kW inverters and 80 s for the125-kW inverters). Then the PV plant ramped it power
production to the maximum power point (MPP) level. The overall black-start process took
approximately 130 s.
65
Figure 85. Soft black start of the PV system using the BESS
Time (s)
Time (s)
Figure 86. (Top) Soft black start of the PV system using the BESS and (bottom) magnified view of
the voltage ramping by the inverter. The voltage ramp was performed within 200 ms.
66
11.2.2 Hard Black Start
Results of the hard black start are shown in Figure 87. In this case, the inverter first set the
voltage in the 13.2-kV collector system with the transformer breaker open. At t=70 s, the breaker
of the PV plant transformer was closed. This resulted in a much higher inrush current and
voltage transient, as shown in Figure 88; however, even this high inrush current is still within the
limits of the BESS inverter. After the transformer was energized, the initialization timers
triggered the inverter controls. The inverters came online, and the plant was fully operational
within approximately 320 s after the black start.
Figure 87. Hard black start of the PV system by closing the PV breaker on the live voltage bus
67
Figure 88. Hard black start of the PV system by closing the PV breaker on the live voltage bus—
magnified view of the voltage and current transient
11.3 Other Black-Start Configurations

The system restoration is a gradual process based on priorities as defined in restoration plans
until the totality of customers are re-energized. The system restoration strategies depend on the
type of systems being restored. For the bulk transmission system, the restoration can be in the
form of re-energizing the full network with black-start units before synchronizing the remaining
generators, or it can be in the form of island-by-island restoration with eventual interconnections
between them. Mixed strategies are also possible.
On lower voltage distribution levels, the system restoration is essentially a network

reconfiguration exercise. The distribution system operators try to transfer the maximum possible
loads from faulted feeders to healthy ones; however, in large-scale blackouts, the distribution
system might stay in the black condition until the bulk system is restored. From this perspective,
the ability of DERs and energy storage systems to provide black start and to participate in
restoration schemes can become a critical component of a future resilient grid.
A typical restoration plan for a bulk power system includes the following essential steps:
• System status identification: establishing blackout boundaries and locations with respect to
critical loads, the status of circuit breakers, the capacity of available black start-units, etc.
• Starting at least one black-start unit to supply critical loads, such as nuclear or large
thermal power plants
• Progressive restoration: providing the step-by-step supply of other loads, avoiding over-
and undervoltage conditions.
68
A restoration concept involving distributed generation can be developed if the distributed
resources are required to have black-start grid-forming capabilities to form islands with critical
loads (Figure 89). In some cases, distribution network reinforcement might be needed depending
on the distributed resource locations because more sectionalizing switches would be necessary
for feeder reconfiguration during a black start. The availability of distribution management
systems and communication networks for distributed generation is also important for successful
restoration services.
Another important aspect of black start and grid forming by a distributed inverter-coupled
system is power quality issues because interfacing converters generate voltage and current
harmonics transmitted to the loads. Harmonic interactions between voltage source inverters and
loads can vary significantly depending on the loading levels of the inverters, the types of
distribution networks (overhead or underground), and the types of loads (resistive or reactive). In
some cases, additional filtering must be installed with black-start-capable inverter units to ensure
a reliable restoration procedure because power supplies and auxiliary systems of some loads
might fail to start, thus jeopardizing the whole black-start process. Inverter filters are normally
designed for specific grid impedance levels, and network reconfiguration during black start
might cause undesirable consequences in the form of harmonic resonances or increased
distortions.
Figure 89. New black-start paradigm

The black-start capability by conventional generators requires first energizing the auxiliary
equipment to generate the initial voltage, which is normally characterized by long startup times
because of slow thermal or hydromechanical dynamics; however, voltage source converters of
modern variable-speed wind turbine generators and PV and energy storage inverters, once
equipped with black-start capability, can provide fast services for power system restoration.
Several configurations of PV-plus-storage systems can serve as a kick-starter for larger
conventional plants that are, in turn, designated as black-start resources. This can be achieved
69
either by PV-plus-storage plants that are colocated with conventional plants or by PV-plus-
storage plants that are not located in close proximity of a conventional plant, as shown in Figure
90. In addition, a PV-plus-storage plant can be configured as a fully functional black-start
resource. The number of PV-plus-storage plants can also be controlled to provide a collective
black-start resource.
Figure 90. Configurations of integrated PV-plus-storage plants as black-start resource for

conventional plants
A PV-plus-storage black-start resource that directly energizes high- and medium-voltage
transmission lines must be capable of managing the charging currents associated with energizing
unloaded transmission lines and must provide inrush current when energizing transformers.
Smaller black-start resources that might need to provide cranking power for larger plants might
also need to handle the inrush currents associated with the cold-load pickup of auxiliary
inductive loads, such as motors at the large plants. For systems similar to the one shown in
Figure 91, one main challenge during black start is to provide inrush current when energizing
transformers and feeders as well as starting auxiliary motors in power plants; therefore, grid-
forming inverters used in a black start need to provide necessary inrush currents. This can be
achieved by oversizing the inverters for inrush currents, equipping motors with soft-starters or
variable frequency drives, or employing soft-starting solutions in grid-forming inverters.
70
Figure 91. PV-plus-storage plant to start auxiliary loads in a thermal power plant
Simulation results for the case with a 1-MW grid-forming PV-BESS system operating at
constant frequency starting four 250-kW auxiliary motors are shown in Figure 92. One motor
starts at a time, and large inrush current is needed for each motor to start. The level of such
inrush current exceeds the rated current for the 1-MW PV-plus-storage plant, even during the
start of the first motor, making the black start of the whole plant impossible because the inverters
will either trip off because of the overcurrent protection or operate in current-limiting mode,
which will make the black start impossible because of voltage collapse.
Figure 92. Consecutive starts of auxiliary motors

Another method for a soft start is described in Section 11.2, using voltage ramping control for
the grid-forming inverter to minimize the inrush currents when energizing the PV pant
transformer. One advanced soft-start control method possible for grid-forming voltage source
inverters can be based on variable-frequency and variable-voltage operation during a black start.
Grid-forming inverters can start at zero voltage and very low frequency, and then increase both
the voltage and frequency to nominal levels at a constant volt/Hertz ratio to avoid both inrush
current and possible transformer saturation. Simulation results for such advanced “variable-
71
frequency-drive-like” functionality for the grid-forming inverter is shown in Figure 93. The
inverter is controlled to ramp up both voltage and frequency during the initial stage of the
transformer energization and startup of all four motors without exceeding the current limit during
the entire startup process.
Figure 93. Constant volt/Hertz startup

Today’s and future restoration strategies should align with the changing network paradigm.
Modern grid-forming inverters can contribute to a black start/restoration with superior reactive
power capabilities compared to conventional synchronous generators; however, the inherent
inverter current limit is one of the most important factors for black-start applications.
Recommendations for future grid-forming and black-start studies using PV-plus-storage at

NREL’s Flatirons Campus might include:
• Fault performance of grid-forming inverters—needs to be robust and standardized

• Robust seamless transition between grid-forming and grid-following modes
• Continued impedance characterization of grid-forming inverters
• Analysis of grid stability impacts of grid forming
• Validated grid-forming inverter models are needed for various renewable and storage
technologies for successful black-start studies.
• At-scale PHIL testing of black-start-capable renewable resources is an important tool to
discover potential issues, test mitigating solutions, and validate models.
72
12 Solar Forecast
In 2019, a solar resource forecasting was implemented for NREL’s Flatirons Campus by First
Solar that produces a PV forecast using a methodology based on National Oceanic and
Atmospheric Administration satellite data. The forecast data are being updated on a secure
NREL File Transfer Protocol server. Currently, several months of forecast data are available, and
more are coming, so long-term statistics on the accuracy and persistence of the solar forecast and
its impact on PV-plus-storage system operation for different market scenarios can be evaluated
in the future. All components of solar radiation—including global horizontal irradiance (GHI),
direct normal irradiance, and diffuse horizontal irradiance—along with ambient temperature are
provided as part of an hour-ahead, day-ahead, and 7-day-ahead forecast.
The estimated active power output of the 430-kW PV plant is calculated from forecast data using
various performance models of PV arrays, such as a widely accepted model that has been
developed and is continuously updated by Sandia National Laboratories [12]. A similar approach
was used in [13] for estimating the power production of the NREL 1-MW PV plant using
irradiance data. Examples of comparisons between forecasted irradiance time series and
estimated power are shown in the following figures. Figure 94 shows the correlation between
actual measured GHI at the NREL site and the hour-ahead forecast of GHI for a time period of 2
weeks of operation, and Figure 95 shows the recorded temperature at the site used in the power
calculations. There is a strong positive correlation between the measured and the forecasted GHI,
as shown in Figure 94. The accuracy of the estimated power time series compared to the actual
measured power is shown in Figure 96. The values of the absolute errors in the hourly power
production forecasts are quite high, reaching +138 W and -211 kW during certain days. NREL is
currently working with First Solar and a forecast provider to understand the nature of such
inaccuracies. The comparison between the estimated and the measured hourly power profiles is
shown in Figure 96 for a whole week of operation.
Figure 94. Correlation between the measured hourly and the forecasted GHI
73
Figure 95. Measured hourly average site temperature during a two-week period
Figure 96. Example of the forecasted and the measured PV plant hourly power production during
2 weeks of operation
Figure 97. Absolute error in the estimated power using an hour-ahead forecast
Figure 98. Example comparison of an hour-ahead and a week-ahead forecast
74
13 Accurate Method for Real-Time Power Reserve
Estimation for Utility-Scale PV Power Plant
13.1 Importance of Accurate Estimation of Available PV Power
All around the world, system operators and utilities are continually adapting their grid codes,
interconnection requirements, operational practices, and market mechanisms to make the
integration of shares of fast-growing variable renewable generation both reliable and economic
[4]. As power systems continue to evolve, the Federal Energy Regulatory Commission noted that
there is an increasing need for a refined understanding of the services necessary to maintain a
reliable and efficient system. In orders 755 and 784, the Federal Energy Regulatory Commission
required improving the mechanisms by which frequency regulation service is procured and
enabling compensation by fast-response resources, such as energy storage. The NERC
Integration of Variable Generation Task Force made several recommendations for requirements
for variable generators (including solar) to provide their share of grid support, including active
power control capabilities [16], [17]. Similar requirements for renewable energy plants have
been introduced in Europe at both the transmission and distribution levels [18], [19]. In 2018,
Federal Energy Regulatory Commission Order No. 842 amended the pro forma interconnection
agreements to include certain operating requirements, including maximum droop and deadband
parameters in addition to sustained response provisions [23].
NERC’s standard BAL-003-1, Frequency Response and Frequency Bias Setting, establishes
target contingency protection criteria for each North American interconnection and individual
balancing authorities within interconnections [24]. Balancing authorities need to meet a
minimum frequency response obligation, so the generating resources that are operated in a mode
and range to meet their frequency response obligation need to have adequate headroom to
respond to frequency transients and load-frequency control set points. Establishing this
headroom is not a problem for the conventional generation fleet, but the varying nature of solar
and wind generation makes it challenging to set and maintain adequate headroom for these
varying resources. In general, all system operators have processes and procedures in place to
ensure grid reliability by monitoring market participant operation. For example, provisions of the
CAISO tariff [28] set penalties for deviations from the dispatch and the regulation capacity for
market participants that fail to comply. The permitted area of variation for the performance
requirements of resources used for various purposes is provided in the CAISO tariff [28]. The
tolerance band is expressed in terms of energy (MWh) for generating units and imports from
external dynamic system resources for each settlement interval, and it equals the greater of the
absolute value calculated using either of the following methods: (1) 5 MW divided by the
number of settlement intervals per settlement period; or (2) 3% of the relevant generating unit’s
maximum output (Pmax), as registered in the master file, divided by the number of settlement
intervals per settlement period.
This CAISO tariff and similar requirements from other system operators imply that the accurate
real-time estimation of available maximum power from a curtailed PV plant is important for
avoiding excessive penalty payments if utility-scale PV plants become market participants for
energy and various reliability services related to active power controls.
75
A typical modern utility-scale PV power plant is a complex system of large PV arrays and
multiple power electronic inverters, and it can contribute to mitigating the impacts on grid
stability and reliability through sophisticated, automatic “grid-friendly” controls [28]. To provide
active power reserves (or a headroom margin) for up-regulation that can be automatically
dispatched as needed, a PV plant needs to operate below its MPP; however, evaluating that MPP
in curtailed mode is not a trivial task, especially for large PV power plants during various types
of variable conditions caused by clouds. One paper [29] proposed an experimentally validated
maximum power point estimation (MPPE) method, which operates in real time using irradiance
and cell temperature measurements to ensure that sufficient reserve power is available. Another
paper [30] proposes an advanced real-time MPP estimation algorithm by applying curve fitting
on voltage and current measurements obtained during inverter operation. Some previously
proposed MPPE methods used offline prediction and employed regression analysis or neural
networks [31], [32]. These methods seem to be accurate but might require excessive processing
power. Others have proposed methods for real-time calculation ([33]–[35]) by making
assumptions that reduce the accuracy of the PV model or, in some cases, require knowledge that
is not typically available on PV module data sheets [33], [36]. Another important limitation of
MPPE estimation methods is that modifications are needed based on inverter types and
topologies. For example, in single-stage inverters (no DC/DC conversion), the power reserve
capability can be achieved by inverter control modifications [30]–[37]; however, in two-stage
systems (inverter and DC/DC converter), the DC/DC converter control needs to be modified
[38]. This makes the use of maximum peak power estimation for curtailed PV systems
challenging and highly dependent on inverter make and topology, the types of modules used in
PV plants, and accurate knowledge of the inverter and PV module parameters.
The variability of PV output in the regulation reserve time frame among various arrays within a
large-scale (~50-MW) PV plant in the southwestern United States was explored in [39].
Although the distributions of changes in aggregate power output throughout all timescales
considered were clustered around a strong peak at zero, the distributions at all timescales
exhibited significant instances of higher magnitude ramps in the tails of the histograms. The
results achieved in [39] were very important because the method presented in this report was also
tested using the data from the same PV power plant [50].
In 2015, a demonstration project was conducted in the U.S. territory of Puerto Rico using a 20-
MW grid-connected PV power plant [40]. This plant was controlled to provide different types of
reliability services to the island’s grid, including various types of active power controls. Testing
on this plant provided “real-world” data on levels of uncertainty that can be introduced by
traditional MPPE estimation methods based on irradiance and temperature measurements as well
as inverter I-V characteristics. One example of such uncertainty is shown in Figure 99 during
operation when the PV plant was responding to an AGC signal sent from the system operator.
The AGC system assumed that there was still some available headroom for up-regulation
because its evaluation was based on the available plant power value that was communicated by
the PPC; however, the calculated available power was overly optimistic, and the inverters were
not able to produce that much power because they were already operating at the maximum peak
power point.
76
Figure 99. Example of inaccurate maximum peak power estimation
Figure 100. Example of maximum peak power estimation using a single reference inverter
A different method for estimating the maximum power was used during a demonstration test of a
300-MW PV power plant in California [4]. In this case, a single 4-MVA inverter was taken from
the AGC scheme and set to operate at the power level determined by its MPPT algorithm. The
measured AC power of this inverter was used as an indicator of available power for the other 79
inverters (80 inverters total), so the plant was able to operate with a fixed 30-MW headroom
(Figure 100). This method also has inherent uncertainties because it assumes uniform solar
irradiation conditions across the whole 300-MW plant. Fortunately, cloud conditions were
favorable for this method to be acceptable because there was a clear sky above the plant during
most of the day of testing.
The method proposed in this project is also based on using dedicated reference inverters within a
curtailed PV power plant for estimating the maximum available power; however, it is based on
using multiple reference inverters to achieve high levels of real-time, maximum peak power
estimation under extreme variability conditions. The method was first validated using data from
an approximately 50-MW PV plant in the western United States [50], and it showed a high level
of accuracy for large utility-scale PV plants. In this report, the method was further improved to
provide accurate estimation of available peak power for smaller distributed PV plants that have
larger per-unit magnitudes of fast changes in plant outputs compared to larger utility-scale plants
because of smaller footprints.
The concept of active power reserve estimation is explained in Figure 101 for a curtailed PV
plant. The estimated available headroom should be accurate to ensure that the plant can avoid
penalties for not meeting its dispatch and regulation capacity, and deviations are within tolerance
bands. For example, if a curtailed plant is requested to release its estimated spinning reserve,
then care must be taken to ensure that the PV plant can reach the commanded power set point
with a high level of precision.
77
Available power underestimated –
Available power overrestimated –
insufficient reserve
excessive curtailment
Can be corrected by Accurate estimation of

energy storage
Available power available power
POWER
Reserve
commitment
Curtailed power Reserve deployed
TIME
Figure 101. Estimation of available power for a PV power plant
13.2 Proposed Method

For utility-scale PV power plants to be able to maintain the desired regulation range or spinning
reserve levels, the plant controller must be able to estimate the available aggregate peak power
that all the plant’s inverters can produce at any point in time. The available power is normally
estimated by an algorithm that considers solar irradiation, the PV modules’ I-V characteristics
and temperatures, inverter efficiencies, etc.; however, this method has many uncertainties,
among them a dependence on the availability of accurate system models, and it does not account
for other factors, such as panel soiling because of dust. The proposed method can determine the
available peak power of the PV plant and maintain the desired reserves with high levels of
accuracy without the use of external sensors or devices. The existing plant hardware and controls
can perform this task after the addition of the new optimized control algorithm in the PPC
software.
Under clear-sky conditions, a single PV inverter can be used as a reference for the whole plant to
determine the available power at any point in time (Figure 102, upper graph); however, under
variable cloud-cover conditions, a single inverter method will not be accurate enough for large
PV power plants. Instead, we propose a concept that can accurately allocate reserves for PV
power plants by:
• Creating dynamic virtual control zones in the PV plant

• Determining which PV inverter will serve as a reference by operating at MPP
• Determining the optimal dispatch interval for the reference MPP inverters based on the
rate of change of power in each zone (indicator of cloud movement)
• Determining the optimized combinations of curtailment set points for participating
inverters in each zone for maximum aggregate inverter efficiency (or minimum electric
losses in the plant) for every control interval.
78
The idea for this method is shown in the lower graph in Figure 102. The plant controller allocates
virtual dynamic control zones consisting of two or more inverters depending on the cloud
conditions over the plant. Then, a single inverter in each virtual group is operated at MPP and is
used as a reference for determining the maximum available power for the zone so that
appropriate curtailment set points can be sent to all participating inverters within that zone. After
a certain time interval, the process is repeated, ensuring accurate reserve allocation by the whole
plant and avoiding excess curtailments.
In this report, we use the following abbreviations:
• 𝑁𝑁𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡 : total number of inverters in a PV power plant

• 𝑁𝑁𝑧𝑧𝑧𝑧𝑧𝑧𝑧𝑧𝑧𝑧 : number of control zones (same as the number of reference inverters, 𝑁𝑁𝑟𝑟𝑟𝑟𝑟𝑟 )
𝑁𝑁𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡
• 𝑁𝑁𝑖𝑖𝑖𝑖𝑖𝑖 = 𝑁𝑁𝑧𝑧𝑧𝑧𝑧𝑧𝑧𝑧𝑧𝑧
: number of inverters in each control zone
• 𝑃𝑃𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡.𝑚𝑚𝑚𝑚𝑚𝑚 : total power production of the plant when all inverters are operating at MPP
• 𝑃𝑃𝑚𝑚𝑚𝑚𝑚𝑚,𝑖𝑖 : power production of zone 𝑖𝑖 when all inverters in the zone are operating at MPP
• 𝑃𝑃𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚,𝑖𝑖 : power production by a single inverter operating at MPP in zone 𝑖𝑖.
Figure 102. Large PV power plant divided into control zones during cloud conditions
79
The estimated maximum available power from the plant using the instantaneous power produced
by single MPPT inverters in each zone (blue inverters shown in Figure 102) is obtained by:
𝑒𝑒𝑒𝑒𝑒𝑒 𝑁𝑁𝑧𝑧𝑧𝑧𝑧𝑧𝑧𝑧𝑧𝑧
𝑃𝑃𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡.𝑚𝑚𝑚𝑚𝑚𝑚 = ∑𝑖𝑖=1 𝑁𝑁𝑖𝑖𝑖𝑖𝑖𝑖 ∙ 𝑃𝑃𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚,𝑖𝑖 (13.1)
The plant curtailment set point as a percentage of the estimated maximum available power from
the plant is determined by:
𝑠𝑠𝑠𝑠𝑠𝑠 𝑒𝑒𝑒𝑒𝑒𝑒
𝑃𝑃𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡 = (1 − ∆𝑃𝑃) ∙ 𝑃𝑃𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡.𝑚𝑚𝑚𝑚𝑚𝑚 (13.2)
where ∆𝑃𝑃 (in per units) is a curtailment set point. For example, ∆𝑃𝑃 = 0.1 means that the plant is
𝑠𝑠𝑠𝑠𝑠𝑠
expected to operate with a 10% active power reserve margin or at 90% of 𝑃𝑃𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡 . The power set
points to all individual inverters in zone 𝑖𝑖 that are participating in the curtailment scheme (orange
inverters in Figure 102) can then be calculated:
𝑁𝑁𝑖𝑖𝑖𝑖𝑖𝑖 (1−∆𝑃𝑃)−1
𝑃𝑃𝑖𝑖𝑖𝑖𝑖𝑖,𝑖𝑖 = 𝑃𝑃𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚,𝑖𝑖 ∙ 𝑁𝑁𝑖𝑖𝑖𝑖𝑖𝑖 −1
(13.3)
Therefore, the power production of zone 𝑖𝑖 can be calculated as:
𝑃𝑃𝑧𝑧𝑧𝑧𝑧𝑧𝑧𝑧,𝑖𝑖 = 𝑃𝑃𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚,𝑖𝑖 + (𝑁𝑁𝑖𝑖𝑖𝑖𝑖𝑖 − 1) ∙ 𝑃𝑃𝑖𝑖𝑖𝑖𝑖𝑖,𝑖𝑖 (13.4)
And the total power production of the plant operating with the curtailment set point, ∆𝑃𝑃, is:
𝑁𝑁 𝑁𝑁
𝑃𝑃𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝 = ∑𝑖𝑖=1
𝑧𝑧𝑧𝑧𝑧𝑧𝑧𝑧𝑧𝑧
𝑃𝑃𝑧𝑧𝑧𝑧𝑧𝑧𝑧𝑧,𝑖𝑖 = [𝑁𝑁𝑖𝑖𝑖𝑖𝑖𝑖 (1 − ∆𝑃𝑃) − 1] × ∑𝑖𝑖=1
𝑧𝑧𝑧𝑧𝑧𝑧𝑧𝑧𝑧𝑧
𝑃𝑃𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚,𝑖𝑖 (13.5)
As part of this project, the validity of this method for utility-scale PV power plants was
demonstrated by using field measurement data from an approximate 50-MW PV plant in the
western United States The results of such analysis were published in a separate NREL report in
2019 [51]. In this report, we explore the possibility of using the same method when applied to
smaller distributed PV power plants (similar to the 430-kW plant at NREL’s Flatirons Campus)
that are exposed to faster output power ramps caused by moving clouds because of the smaller
footprint of such plants compared to large utility-scale projects. The accurate estimation of
available power in PV plants that happened to be curtailed for any reason is also important for
integrated PV-plus-storage operation so that the plant controller can have precise information on
the available spinning reserve from PV and can dispatch energy storage accordingly.
13.3 Implementation of Reserve Estimation Method on the 430-kW PV

Power Plant
In this work, we examined the applicability of the proposed method using PV output power data
from different arrays in a single utility-scale 430-kW PV plant at NREL. The plant consists of 6
individual inverters, two rated at 125 kW and four rated at 45 kW. We used 1-s power data from
each individual inverter collected from the plant, allowing us to analyze the accuracy of the
proposed method under different resource variability scenarios. Example production profiles for
the plant are shown in Figure 103 for 2 days of operation with different variability patterns for
the whole plant and individual inverters (the output of the 45-kW inverters is shown in
aggregate). The 1-s changes in total plant power are shown in Figure 104, indicating the
80
magnitudes of the 1-s variability, which is quite high compared to larger utility-scale solar power
plants, as observed in the analysis conducted in [4] and [51]. The maximum 1-s change for this
430-kW plant can reach more than ±10% of capacity per second.
Figure 103. Examples of 430-kW plant daily production profiles
Figure 104. Measured 1-s total power ramps

Because the 430-kW PV plant has only 6 string inverters, the evaluation of the available power
estimation method was done with only one 125-kW reference inverter, unlike up to the 48
reference inverters used in the larger plant [50].
We evaluated the accuracy of the method using a simple equation for calculating the available
total plant power derived from the real-time power output of a single 125-kW operating at MPP:
𝑃𝑃𝑒𝑒𝑒𝑒𝑒𝑒 (𝑡𝑡) = 𝑃𝑃125 (𝑡𝑡) ∙ 𝐾𝐾𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠 (13.5)
where 𝑃𝑃𝑒𝑒𝑒𝑒𝑒𝑒 (𝑡𝑡) is the estimated maximum total plant power [kW], 𝑃𝑃125 (𝑡𝑡) is the measured output
power of a single 125-kW inverter operating at MPP [kW], and 𝐾𝐾𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠 is a constant scaling
coefficient defined as 430 kW / 125 kW = 3.44.
After applying the active power estimation method described in the previous section (henceforth,
we refer to it as the “simple” method), a larger than expected estimation error was observed,
81
especially during the periods of high variability, as shown in Figure 105 and Figure 106. For
comparison, the same simple method applied to the 50-MW plant produced estimation errors
within ±1.5% of the rated plant capacity. For this much smaller 430-kW plant, however, the
estimation errors exceeded ±10% of rated plant capacity, which might not be acceptable for the
accurate evaluation of available spinning reserves from distributed PV plants.
Figure 105. Comparison of estimated and measured power using simple method
Figure 106. Peak power estimation error

This error was caused by using a constant value for the 𝐾𝐾𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠 scaling coefficient, which did not
account for fast changes in power across the various sections of the plant because of its small
footprint. In fact, it was observed from further analysis that the estimation error had a strong
linear correlation with the 1-s changes in power output of the reference inverter, as shown in
Figure 107. This fact allows introducing a dynamic scaling coefficient that accounts for such fast
variability in plant power.
82
Figure 107. Correlation between the estimation error and the 1-s variability in the power output of
the reference inverter and linear regression
The slope, a, and intercept, b, of the linear regression estimation were used to introduce a
dynamic scaling factor as a function of 1-s changes in power for the reference inverter, defined
as:
𝑅𝑅𝑟𝑟𝑟𝑟𝑟𝑟 (𝑡𝑡𝑖𝑖 ) = 𝑃𝑃𝑟𝑟𝑒𝑒𝑓𝑓 (𝑡𝑡𝑖𝑖 )−𝑃𝑃𝑟𝑟𝑟𝑟𝑟𝑟 (𝑡𝑡𝑖𝑖−1 ) (13.6)
where 𝑅𝑅𝑟𝑟𝑟𝑟𝑟𝑟 (𝑡𝑡𝑖𝑖 ) is the 1-s ramp rate of the reference inverter (% of capacity/s) measured at time
instance 𝑡𝑡𝑖𝑖 ; and 𝑃𝑃𝑟𝑟𝑟𝑟𝑟𝑟 (𝑡𝑡𝑖𝑖 ) is the measured power output of the reference inverter operating at
MPP.
The dynamic scaling factor then can be defined as:
𝐾𝐾𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠 (𝑡𝑡𝑖𝑖 ) = 𝑎𝑎 ∙ 𝑅𝑅𝑟𝑟𝑟𝑟𝑟𝑟 (𝑡𝑡𝑖𝑖 ) + 𝑏𝑏 (13.7)
Therefore, the improved estimated power at each time step, 𝑡𝑡𝑖𝑖 , can be defined as:
𝑃𝑃𝑒𝑒𝑒𝑒𝑒𝑒 (𝑡𝑡𝑖𝑖 ) = 𝑃𝑃125 (𝑡𝑡𝑖𝑖 ) ∙ 𝐾𝐾𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠 (𝑡𝑡𝑖𝑖 ) (13.8)
This is a very simple approach but one that does not require complex calculations in the PV plant
controller and provides substantial improvement in the estimated maximum available power for
the entire plant using real-time power measurements from a single reference inverter that
operates at MPP. The results of such improvements are shown in the estimation error histograms
depicted in Figure 108. For both the less variable Day 1 and more variable Day 2 profiles, the
estimation error of the improved method helps keep the maximum 1-s peak power estimation
error within ±5% of the rated plant capacity, which is adequate for a plant of this size.
The accuracy of the method will improve with time if long-term plant data are analyzed to derive
a more precise regression for a dynamic scaling factor that can also be scheduled based on the
season. This will be a subject of future research.
83
Figure 108. Distribution of peak power estimation errors
13.4 Evaluation of Peak Power Estimation Method for a Large 300-MW

PV Power Plant
During this project, we continued evaluating the reference inverter method for estimating the
available peak power in large PV plants with curtailed headroom. We used 1-s production data
from one approximately 300-MW PV plant located in the western United States. The plant
consists of 79 inverters, each rated at 4 MW, with PV arrays covering large geographic area.
Figure 109. Example of the daily operation profile of a 300-MW PV plant

Figure 109 shows an example of 1 day of production of the plant, with the plant total power and
total 1-s changes in power (upper graph) and a more detailed view of the performance of each
individual inverter during that day (lower heat map). The horizontal axis of the heat map
84
represents the time of day, the vertical axis indicates the inverter number, and the color of the
heat map indicates the power production level of each individual inverter during every second of
the day. We analyzed several similar days of production with different variability patterns. The
total plant 1-s production data and data from the 79 individual inverters for four variability cases
were used to evaluate the ability of the proposed method to accurately predict the maximum available
power from the entire plant for different numbers of reference inverters.
The following numbers of control zones with a single reference inverter in each zone were used in
the analysis of the number of reference inverters: 𝑁𝑁=1, 2, 4, 8, 12, 24, 38. A simple algorithm was
developed to select the locations of the reference inverters using an equidistant approach. Another
innovative approach used in this analysis was to implement dynamic control zones in the PV plant.
The designated reference inverters can be selected dynamically by the plant controller based on
various factors. One of these factors can be the array temperatures of curtailed zones. By cycling the
reference inverters every specific time interval, excessive heating of the array sections can be
avoided to reduce losses. Such operation is shown in Figure 110. In this example, the reference
inverters are cycled every 10 min, which explains the checkers-like heat map because the same
inverters are switching between peak power and curtailed operation every 10 min.
Figure 110. Dynamic reference inverters (light green circles represent reference inverters)
The accuracy of the peak power estimation under high-variability conditions for different
numbers of reference inverters is shown in Figure 111 in the form of histogram plots. As
expected, the highest accuracy of estimation is achieved for a larger number of reference
inverters (N=38). The maximum positive estimation error is 7.82 MW (or 2.6% of the plant
capacity), and the minimum negative estimation error is -7.61 MW (or 2.53% of the plant
capacity).
85
Figure 111. Peak power estimation errors
The examined peak power estimation method for curtailed PV power plants is based on using
dedicated reference inverters within a plant. The proposed technique does not require deploying
any additional equipment or sensors and is based only on the addition of new control logic to the
existing PPC. Based on our calculations using measured 1-s power production data from the
entire PV power plant and individual inverters, the method has the potential to produce highly
accurate real-time estimates of available aggregate peak power that all the plant’s inverters can
produce at any point in time and can ensure that the control error stays within the desired
tolerance bands at all times. The method, by default, accounts for external factors, such as dust
accumulation on portions of the modules and the presence of snow or ice patches on the
modules, without any additional computations when estimating the available peak power.
The NREL-developed method for active power reserve estimation in curtailed PV plants has
been successfully implemented in several utility-scale PV projects worldwide. One such example
is the Luz del Norte 141-MW PV power plant owned by First Solar in Chile’s Atacama region
[42]. This PV plant was used as a pilot to enable PV generation to provide ancillary services to
the SEN grid in Chile.
86
14 Test Platform for PV-BESS Systems in Distribution
Grids
During this project, we developed a test bed for testing the PV-BESS system in a distribution
system and for microgrid applications. The test bed was developed in collaboration with Pacific
Gas & Electric Company using a model of their typical distribution circuit. The distribution
system model was provided by Pacific Gas & Electric Company, and the NREL team carried out
the RSCAD implementation of the distribution system model and implemented it in RTDS (a
simplified diagram of the system is shown in Figure 112).
The CGI POI is on the 13.2-kV side of the PV-BESS transformers. An additional voltage-
matching transformer was placed in the RSCAD model to match the PV and BESS with a
voltage level of 12.47 kV. This point, shown in Figure 112, is the virtual POI of the PV-BESS
system with the RSCAD distribution model. Various types of one, two, and three-phase voltage
faults can be introduced in the RSCAD model on the 13.2-kV terminals, so the CGI can emulate
the exact voltage waveforms on the 12.47-kV bus under such fault conditions. The distribution
system model also has submodels of other distributed PV plants, the distribution grid, and
various types of loads. A real hardware relay (SEL-351) is connected to the RTDS model as
well. The distribution circuit can operate as a stand-alone microgrid after opening Circuit
Breaker 1 (CB1). This way, the ability of the grid-forming control of the PV-BESS plant can be
tested for islanded operation. One example of such a test event is shown in Figure 113, with the
BESS operating in grid-forming mode. After Circuit Breaker 1 trips, the BESS starts tracking the
load and automatically injects power into the island. The measured voltage waveforms do not
show any transient behavior during this process (upper plot in Figure 113).
Another useful outcome of the test bed is the ability to evaluate the impacts of PV-BESS systems
on currents and voltages in the microgrid during transient events. One example of such an
evaluation is shown in Figure 114. It combines results on both peak current and minimum
voltage levels in different measurement points in the circuit during a line-to-line faut in one of
the system buses. NREL developed an automated test and visualization system that allows for
producing these types of plots for any desired fault and transient event scenarios, identifying the
most “stressed” nodes in the system, and determining the appropriate settings for protection and
coordination between various protection devices under each scenario.
This test bed is a unique tool for conducting “at-scale” integration testing of various types of PV-
BESS configurations and associated controls in distribution grids and islanded systems. NREL
plans on using this platform in future SETO-funded research and collaborative projects with
industry.
87
PHIL
FS PV Plant Island forms after CB1 trip

=
~ Additional voltage matching
SEL-351 CGI POI transformer in RSCAD model
NREL BESS 1MW/1MWh
T3 13.2/12.47kV T2 T1 230kV
= x
Z2 Z1
~ 400V/13.2kV CB4 CB1 12.47/70kV 70/230kV

12.47kV, Old River
N SEL-351
(# of BESS units)
CHIL
CB2
F1 Z3
Z4
Island 650 kVA

PF=0.6893
Z5
Z6
F2
Z7 PV
924 kW
Z8
SW1 800 kVAR
Z9
R
F3 Z10
Z11 SW2 800 kVAR
F3
560 kVA Z12 Z14 Z15
PF=0.6825
Z16
CB3 PV
T4 12.47kV/480V 150 kW 150 kW 150 kW
M
1000HP
PF=0.85
Figure 112. Test bed for PV-BESS systems in distribution microgrids and islanded grids
Time (s)
Figure 113. Measured response of grid-forming BESS after Circuit Breaker 1 trips in the PHIL
model
88
Figure 114. Minimum voltages and maximum currents recorded in all buses during a line-to-line
low-impedance fault in Bus 691
89
15 Conclusions
This NREL-First Solar cooperative research-and-development agreement project leveraged and
benefited from the prior collaboration of large-scale demonstration projects for the provision of
reliability grid service by PV plants, knowledge and field experience accumulated from NREL’s
previous work with First Solar and other partners (CAISO, ERCOT, AES, PREPA, etc.), and
advanced prior work by creating new capabilities at NREL’s Flatirons Campus for the validation
and testing of advanced controls for PV-BESS plants and the demonstration testing of such
controls under both real and controllable grid conditions. Specific accomplishments of this
project are:
• Constructed, interconnected, and commissioned a real 430-kV PV plant with string

inverters. The plant was designed for easy expansion to larger power capacities (up to 1
MW), if needed in the future.
• Integrated the PV plant into NREL’s PHIL validation platform based on the CGI and
RTDS systems
• Combined the PV and BESS controls under the hybrid plant controller framework to
provide existing and evolving ancillary service products for various balancing authorities
and system operators in the U.S. interconnections and island systems
• Developed controls with increased complexity because of the multi-technology nature of
the PV-BESS operation
• Enhanced the PV PPC based on the SEL RTAC platform by adding new algorithms to
include BESS controls and provide many types of essential reliability services
• Demonstrated the operation of the PV-BESS plant with a robust controller interacting
with three types of inverters from different vendors
• Demonstrated new PV-BESS controls that can be used by existing and future ancillary
service markets (inertial response, PFR, FFR, down-ramp control, provision of spinning
reserve, dispatchable operation, voltage control, different types of ramp controls, etc.)
• Demonstrated the ability of PV-BESS plants to provide wide-area stability services in the
form of interarea oscillation damping
• Demonstrated that grid-forming controls by BESS can provide resilience services (black
start, islanded operation, system restoration) to hybrid PV-plus-storage plants
• Developed and demonstrated methods for the impedance characterization of PV-BESS
systems using the 7-MVA CGI coupled with the RTDS
• Demonstrated the black-start capabilities of grid-forming BESS as part of the PV-BESS
plant
• Developed a new accurate method for the real-time estimation of available active power
reserve in curtailed PV plants. The method has been implemented in serval utility-scale
projects worldwide.
The project resulted in the development of a new valuable test and validation asset—a grid-scale
PV-plus-storage hybrid system test bed that can be used by the industry and research
communities for the validation and demonstration of new control concepts, stakeholder
engagement, workforce education, and as a validation platform for the future standardization of
hybrid technologies. It will lead to the development of both technology- and vendor-neutral,
plug-and-play hybridization guidelines and control architectures openly available at all segments
of stakeholder communities.
90
This project was the first research effort that demonstrated the multi-technology aspect of grid
integration research possible at NREL’s Flatirons Campus, and it facilitated the shift toward the
development of new research platforms, such as IESS and ARIES. Capabilities developed under
this project were also foundational for continued hybrid systems research conducted at NREL
and were used in other DOE -funded research work, such as the Grid Modernization Laboratory
Consortium FlexPower hybrid plants project.
91
16 Future Opportunities
The emergence of new integrated energy systems provides many opportunities, and it also
presents urgent challenges that must be addressed to ensure that these new integrated energy
systems meet evolving consumer expectations and continue to be economic, reliable, resilient,
and secure. From this perspective, integrated PV-BESS systems can play a crucial role for power
systems in the transition to high shares of IBRs. New, upcoming capabilities of NREL’s
Flatirons Campus will establish new, unique research opportunities for PV-plus-storage systems
in terms of storage technology diversity, increased capacity, and increased voltage levels for
integrated systems, allowing for the study of at-scale interactions among PV generation, storage,
advanced controls, and cybersecurity in hybrid energy systems—at both the integration and
planning stages. A new 19.9-MVA CGI will be a major addition to the ARIES research platform,
enhancing NREL’s capabilities to conduct more advanced PHIL experiments for integrated PV-
plus-storage systems involving not only BESS but also other storage technologies, such as
hydrogen storage, flow batteries, flywheels, and ultracapacitors. The main technical
characteristics of the new CGI are shown in Figure 115. There is currently no other research
platform like the Flatirons Campus anywhere in the world that can support research at the scale
and integration necessary to make a global impact in the marketplace.
Figure 115. Main characteristics of the 19.9-MVA CGI
92
References
[1] https://www.cpuc.ca.gov/rps/
[2] P. Denholm, R. Margolis, “Energy Storage Requirements for Achieving 50% Solar
Photovoltaic Penetration in California”. NREL report, August 2016
[3] V. Gevorgian, B. O’Neill Advanced Grid-Friendly Controls Demonstration Project
for Utility-Scale PV Power Plants. NREL report, 2016.
www.nrel.gov/docs/fy16osti/65368.pdf
[4] V. Gevorgian, C. Loutan, M. Morjaria, et.al. “Demonstration of Advanced Ancillary
Service Controls by a Large Utility-Scale PV Power Plant in California”. NREL draft
report in review, December 2016
[5] P. Denholm, J. Eichman, R. Margolis, “Evaluating the Technical and Economic
Performance of PV Plus Storage Power Plants” NREL technical report, August 2017
[6] R. Fu, T. Remo, R. Magolis, “2018 U.S. Utility-scale PV-plus-Energy Storage
System Costs Benchmark”, NREL technical report, November 2018
[7] W. Gorman, A. Mills, M. Bolinger, R. Wiser, N. Singhal, E. Ela, E. O’Shaughnessy,
“Motivations and options for deploying hybrid generator-plus-battery projects within
bulk power system”, The Electricity Journal, 33/2020.
[8] M. Bortolini et al., “Technical and economic design of photovoltaic and battery
energy storage system,” Energy Conversion and Management, vol. 86, pp. 81 – 92,
2014.
[9] Y. Khawaja, D. Giaouris, H. Patskis, M. Dahidah, “Optimal Cost-Based Model for
Sizing Grid-Connected PV and Battery Energy System”, 2017 IEEE Jordan
Conference on Applied Electrical Engineering and Computing Technologies
(AEECT), October 11-13, 2017
[10] C.S. Lai, M. D. Mcculloch, : LCOE for PV and Grid Scale Energy Storage Systems,
Applied Energy, Elseiver, volt 190©, pages 191-203
[11] V. Gevorgian et. al. “WGRID-49 GMLC Project Report”, NREL, August 2019,
https://www.osti.gov/biblio/1558905
[12] Kratochvil, J.A.; Boyson, W.E.; King, D.L. Photovoltaic Array Performance Model;
Technical Report; Sandia National Laboratories: Albuquerque, NM, USA, 2004.
[13] L. Petersen, F. Iov, G. C. Tarnowski, V. Gevorgian, P. Koralewicz, D.-I. Stroe,
“Validating Performance Models for Hybrid Plant Control Assessment”, Energies
2019, 12, 4330
[14] “EPIC 2.05: Inertia Response Emulation of DG Impact Improvement”, EPIC Final
Report, Pacific Gas and Electric Company, February 20, 2019
[15] X.-Q. Guo, H.-R. Gu, “Phase locked loop and synchronization methods for grid
interfaced converters”, PRZEGLĄD ELEKTROTECHNICZNY (Electrical Review),
ISSN 0033-2097, R. 87 NR 4/2011
[16] NERC, Integration of Variable Generation Task Force Report (Technical Report)
(Washington, D.C.: 2012).
[17] NERC, Essential Reliability Services Working Group and Distributed Energy
Resources Task Force Measures (Technical Report) (Washington, D.C.: 2012),
93
[18] ENTSO-E, “ENTSO-E network code for requirements for grid connection applicable
to all generators,” ENTSO-E, Brussels Belgium, Tech. Guideline, March 2013
[19] Technical Requirements for the Connection to and Parallel Operation with Low-
Voltage Distribution Networks, VDE-AR-N 4105
[20] H. Villegas Pico, V. Gevorgian, “Ultra-fast Freqeucny Response of Converter-
dominant Grids using PMUs”, Conference paper presented at Wind and SOalr
Integration Workshop in Dublin, Ireland, October 2019
[21] WECC Governor Droop Seeting Criterion, PRC-001-WECC-CRT-1.2
[22] P. Denholm, E. Ela, B. Kirby, “The Role of Energy Storage with Renewbae Electricity
Generaion”, NREL technical report, January 2010
[23] “Essential Reliability Services and the Evolving Bulk-Power System – Primary
Frequency Response”, Federal Energy Regulatory Commission’s Order No. 842
[24] “BAL-003-1 Frequency Response and Frequency Bias Setting Standard”, NERC, 2016
[25] J. C. Neely, R. H. Byrne, R.T. Elliott, C. Silva-Monroy, D. Schoenwald, D.
Trudnowski, M, Donnely, “Daming of Inter-area Oscillations using Energy Storage”,
IEEE PES GM meeting paper, July 2013
[26] P. Kundur, Power System Stability and Control. New York: McGraw-
[27] Hill, Inc., 1993
[26] P. Kundur, Power System Stability and Control, New York, KcGraw-Hill, Inc., 1993
[27] Federal Energy Regulatory Commission (FERC), “Large Generator Interconnection
Agreement,” www.ferc.gov/industries/electric/indus-act/gi/stnd-gen.asp.
[28] M. Morjaria, D. Anichkov, V. Chadliev, and S. Soni, “A Grid Friendly Plant,” IEEE
Power and Energy Magazine (May/June 2014)
[29] CAISO 5th Replacement Electronic Tariff, Appendix A: Master Definition Supplement,
February 1, 2017
[30] A. Hoke, E. Muljadi, D. Maksimovic, “Real-time Photovoltaic Plant Maximum Power
Point Estimation for Use in Grid Frequency Stabilization”, 2015 IEEE 16th Workshop
on Control and Modeling for Power Electronics
[31] E. Batzelis, G. Kampitsis, S. Papathanassiou, “Power Reserves Control for PV Systems
with Real-time MPP Estimation via Curve Fitting”, IEEE Transactions on Sustainable
Energy, Vol. 8, No.3, July 2017
[32] T. Hiyama and K. Kitabayashi, “Neural network-based estimation of maximum power
generation from PV module using environmental information,” Energy Convers. IEEE
Trans. On, vol. 12, no. 3, pp. 241–247, 1997.
[33] M. Taherbaneh and K. Faez, “Maximum power point estimation for photovoltaic
systems using neural networks,” in Control and Automation, 2007. ICCA 2007. IEEE
International Conference on, 2007, pp. 1614–1619.
[34] J.-C. Wang, Y.-L. Su, J.-C. Shieh, and J.-A. Jiang, “High-accuracy maximum power
point estimation for photovoltaic arrays,” Sol. Energy Mater. Sol. Cells, vol. 95, no. 3,
pp. 843–851, 2011.
94
[35] A. Garrigós, J. M. Blanes, J. A. Carrasco, and J. B. Ejea, “Real time estimation of
photovoltaic modules characteristics and its application to maximum power point
operation,” Renew. Energy, vol. 32, no. 6, pp. 1059–1076, 2007.
[36] J. Ma, K. L. Man, T. O. Ting, N. Zhang, S.-U. Guan, and P. W. Wong, “Dem: direct
estimation method for photovoltaic maximum power point tracking,” Procedia
Comput. Sci., vol. 17, pp. 537–544, 2013.
[37] G. Kumar and A. K. Panchal, “Geometrical prediction of maximum power point for
photovoltaics,” Appl. Energy, vol. 119, pp. 237–245, 2014.
[38] B.-I. Craciun, T. Kerekes, D. Sera, and R. Teodorescu, “Frequency support functions
in large PV power plants with active power reserves,”IEEE J. Emerg. Sel. Topics
Power Electron., vol. 2, no. 4, pp. 849–858, Dec. 2014.
[39] V. A. K. Pappu, B. Chowdhury, and R. Bhatt, “Implementing frequency regulation
capability in a solar photovoltaic power plant,” in Proc. North Amer. Power Symp.
2010, Arlington, TX, USA, Sep. 2010, Art. no.5618965
[40] S. Shedd, B.-M. Hodge, A. Florita, and K. Orwig, “A Statistical Characterization of
Solar Photovoltaic Power Variability at Small Timescales”, 2nd Annual International
Workshop on Integration of Solar Power in Power Systems Conference, Lisbon,
Portugal, Nov 12-13, 2012
[41] V. Gevorgian, B. O’Neill, “Advanced Grid-Friendly Controls Demonstration Project
for Utility-Scale PV Power Plants”, NREL technical report, January 2016.
[42] V. Gevorgian, “Highly Accurate Method for Real-Time Active Power Reserve
Estimation for Utility-Scale Photovoltaic Plants”, NREL report, February 2019
[43] S. Michels Alvaro, T. Avila Araya, G. Ortiz Mercado, N. Milan, “Demonstration of
Ancillary Service Provision Capabilities of PV Power Plants: Study Case Luz Del
Norte”, LABORELEC report, April 2019
[44] “Innovative Ancillary Services”, IRENA report, 2019
[45] M. Singh, A. J. Allen, E. Muljadi, V. Gevorgian, Y. Zhang, and S. Santoso, “Interare
oscillation damping controls for wind power plants,” IEEE Trans. Sustain. Energy,
vol. 6, no. 3, pp. 967-975, July 2015.
[46] “Oscillation Event 03.12.207 – System Protection and Dynamics WG,” ENTSO-E,
Brussels, Belgium, 2018.
[47] F. R. Schleif and J. H. White, “Damping for the northwest – southwest oscillations—
an analog study,” IEEE Trans. Power App. And Syst., vol. 85, no. 2, pp. 1239-1247,
Dec. 1966.
[48] E. Rehman, M. Miller, J. Schmal, S. H. Huang, “Dynamic stability assessment if high
penetration of renewable generation in the ERCOT grid,” ERCOT, Taylor, TX, USA,
2018. [Online]. Available:
http://www.ercot.com/content/wcm/lists/144927/Dynamic_Stability_Assessment_of_
High_Penetration_of_Renewable_Generation.pdf
[49] E. Rehman, M. G. Miller, J. Schmall, S. H. Huang, and J. Billo, “Stability assessment
of high penetration of inverter-based generation in the ERCOT grid,” in Proc. 2019
IEEE Power and Energy Soc. Gen. Meeting, Atlanta, GA.
95
[50] S. Shah, W. Yan, V. Gevorgian, and W. Gao, “Power-domain impedance theory for
the analysis and mitigation of interarea oscillations,” in Proc. 2020 IEEE Power and
Energy Soc. Gen. Meeting. [Online]. Available:
https://www.nrel.gov/docs/fy20osti/75273.pdf
[51] S. Shah, V. Gevorgian, P. Koralewicz, R. Wallen, and W. Yan, “Impedance methods
for analyzing stability impacts of inverter-based resources,” in Proc. NSF Workshop
Power Electron. Enabled Operation of Power Syst., Chicago, IL, Nov. 2019.
[52] S. Shah, P. Koralewicz, V. Gevorgian, and R. Wallen, “Impedance measurement of
wind turbines using a multimegawatt grid simulator,” in Proc. 18th Wind Integr.
Workshop, Dublin, Ireland. [Online]. Available:
https://www.nrel.gov/docs/fy20osti/74890.pdf
[53] S. Shah and L. Parsa, “Impedance modeling of three-phase voltage source converters
in dq, sequence, and phasor domains,” IEEE Trans. Energy Conv., vol. 32, no. 3, pp.
1139-1150, Sep. 2017.
[54] S. Shah, P. Koralewicz, V. Gevorgian, and L. Parsa, “Small-signal modeling and
design of phase-locked loops using harmonic signal-flow graphs,” IEEE Trans.
Energy Conv., vol. 35, no. 2, pp. 600-610, June 2020.
96

Photovoltaic Plant and BESS Integration

Uploaded by

Copyright:

Available Formats

Photovoltaic Plant and BESS Integration

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

Photovoltaic Plant and BESS Integration

Uploaded by

Copyright:

Available Formats

Photovoltaic Plant and Battery Energy

Storage System Integration at NREL’s

With collaboration from First Solar

NREL is a national laboratory of the U.S. Department of Energy Technical Report

Contract No. DE-AC36-08GO28308

With collaboration from First Solar

NREL is a national laboratory of the U.S. Department of Energy Technical Report

This report is available at no cost from the National Renewable

U.S. Department of Energy (DOE) reports produced after 1991

NREL prints on paper that contains recycled content.

Figure ES-1. Colocated AC-coupled PV-storage power plant

Figure 1. CAISO generation breakdown for April 21, 2019.

Figure 3. Available and curtailed PV power on April 21, 2019

1.2 Configurations of PV-Storage Systems

1.3 Structure of This Report

1. Develop a world-class research facility consisting of an at-scale PV-BESS system at

This cooperative research-and-development agreement project conducted by NREL and First

Figure 5. ARIES virtual emulation platform at NREL’s Flatirons Campus

Figure 6. Physical layout of First Solar’s PV plant at NREL’s Flatirons Campus

Figure 8. Aerial photo of First Solar’s PV plant at NREL’s Flatirons Campus.

Figure 9. Layout of First Solar’s PV plant site at NREL’s Flatirons Campus

Table 1. Parameters of First Solar’s PV Plant

Power rating 430 kW AC/480 kW DC

Figure 10. Interconnection diagram for six string inverters

Figure 12. Circuit diagram of a 45-kW string inverter

Figure 15. Aerial photo of BESS. Photo by NREL

40 MVA for 2 sec DC

NR RTAC CGI RTDS

First Solar PPC

Solar forecast and price signals

Figure 22. Flatirons Campus main SCADA display

Figure 23. PV plant SCADA display

Selectable single BESS Selectable number of

Qlimit Qset with

ΔV deadband ± 1MVAR limit

Voltage at POI Qset

Figure 27. Measured PV plant response times

Figure 28. Measured BESS response time in grid-following mode

Figure 29. Measured BESS response time in grid-forming mode

Figure 30. BESS response to active power set points

Table 2. PHIL Scenarios

PV-BESS Level of IBRs (% of Load)

8.1 Inertial Response Tests

Figure 35. Calculated range of programmable H for NREL’s 1-MW/1-MWh BESS

8.2 Fast Frequency Response Tests

8.2.2 PMU-Based FFR Concept

Figure 37. PMU networks embedded in the power system

Figure 38. Concept of PMU-based wide-area reliability service by PV-BESS plants

Figure 39. PMU-controls test bed developed at NREL's Flatirons Campus

Figure 40. PMU-based system characterization test setup

8.3 Frequency Droop Tests

Figure 46. Frequency droop by a PV-BESS system

Figure 47. PV plant providing 5% droop response

Figure 48. PV plant providing 3% droop response

8.3.2 Comparison of Frequency Response Methods

• No PV-BESS present in the PHIL test case

Figure 50. Comparison of impacts by inertial, FFR, and PFR controls

8.4 AGC Tests

Cloud impact on AGC performance