WEBINAR OUTLINE

Jay Jonhson and

Adam Summers Mathieu Haineault
Introduction Hardware-in-the-Loop Simulations OPAL-RT Product
of PV Inverter Distribution Voltage Announcement
Regulation

1 2 3 4 5 46

Dr. Sudipta Chakraborty François Tempez

Focus on European PHIL
What is Power Hardware-in-the- Question Period
projects from OPAL-RT
Loop (PHIL)
users
Hardware-in-the-Loop Simulations of PV
Inverter Distribution Voltage Regulation

Jay Johnson and Adam Summers

(with support from Rachid Darbali-Zamora, Javier Hernández-Alvídrez, Matt Reno, and Cliff Hansen)

March 7, 2019

Sandia National Laboratories is a multimission laboratory managed and operated by National Technology and Engineering Solutions of Sandia, LLC, a wholly owned subsidiary of
Honeywell International, Inc., for the U.S. Department of Energy’s National Nuclear Security Administration under contract DE-NA0003525. SAND2019-2407 PE.
 BACKGROUND 4

• Context
• Total installed capacity of PV is growing fast
• Large growth expected in distribution systems
• Problem
• Grid is slow to evolve, we encounter technical challenges with
voltage/frequency regulation, protection, etc.
• Unless mitigated, these challenges will make it increasingly
difficult and costly to continue integrating renewable energy
• Solution: advanced inverters
• Actively support voltage and frequency by modulating output
...Faster than a tap changer
• Have high tolerance to grid disturbances ...More powerful than a rotating
machine
• Interact with the system via communications
...Able to leap deep voltage sags in a
• Research questions: single bound
Courtessy of B. Lydic, Fronius
• What is the best technique for providing voltage regulation?
• How can the methods be evaluated with physical devices prior
to field implementation?
DISTRIBUTION VOLTAGE REGULATION 5

Voltage regulation on a feeder without Voltage regulation on a feeder with

distributed generation. distributed generation.

Solution: Use DER grid-support functions with reactive power capabilities.

- Cost-effective: no additional equipment required
- Logical: employs devices which are causing voltage rise to mitigate the problem

Images: B. Palmintier, et al., On the Path to SunShot: Emerging Issues and Challenges in Integrating Solar with the Distribution
System, NREL/TP-5D00-65331, May 2016.
ENERGISE ProDROMOS Project 6

Programmable Distribution Resource Open Management Voltage Regulation Power

Optimization System (ProDROMOS)1 Simulations

Real-Time Voltage
Goal: create an Advanced Distribution Management System Regulation Power
(ADMS) that captures distribution circuit telemetry, performs Simulations
state estimation, and issues optimal DER setpoints based on PV
production forecasts.
Power Hardware-in-the-
Loop Voltage Regulation
Team is using PHIL experiments to gain confidence in control Power Simulations
algorithms, verify communication interfaces, and predict
performance prior to deploying the ADMS on a live feeder in
Field Demonstrations on
Massachusetts.
Live Power Systems

1Prodromos is Greek for "forerunner" and the prodromoi were a

light cavalry army unit in ancient Greece used for scouting missions.
OPTIONS FOR VOLTAGE REGULATION USING GRID-SUPPORT
FUNCTIONS 7

Volt-Var Mode
Distributed Autonomous Control
Q
• Function: volt-var or volt-watt
V
• Pros: Simple, requires little or no communications, DER locations not needed
• Cons: does not reach global optimum
Extremum Seeking Control (ESC)
ESC

• Function: new grid-support function

• Pros: can achieve global optimum
• Cons: requires fitness function broadcast; ideally with new inverter function
Optimized Power Factor Control
• Function: power factor or reactive power commands
• Pros: direct influence over DER equipment to achieve objective
Optimal Power Factor
• Cons: requires telemetry, knowledge of DER locations, and state t
estimator/feeder model V I
 ESC 8

Extremum Seeking Control will be used as a

comparison to the PF optimization technique
Steps in ESC:
A. Centralized control center
collects data from the power system
B. Control center calculates the objective
function, e.g.,𝐽𝐽 = 1/n∗Σ[ Vi − Vn /Vn ]2
C. Control center broadcasts objective function to all inverters.
D. Individual inverters extract their frequency-specific effect
on the objective function and adjust output to trend toward
the global optimum.

• D. B. Arnold, M. Negrete-Pincetic, M. D. Sankur, D. M. Auslander and D. S. Callaway, "Model-Free Optimal Control of VAR Resources in Distribution Systems: An
Extremum Seeking Approach," IEEE Transactions on Power Systems, vol. 31, no. 5, pp. 3583-3593, Sept. 2016.
• J. Johnson, R. Darbali, J. Hernandez-Alvidrez, A. Summers, J. Quiroz, D. Arnold, J. Anandan, "Distribution Voltage Regulation using Extremum Seeking Control with
Power Hardware-in-the-Loop," IEEE Journal of Photovoltaics, vol. 8, no. 6, pp. 1824-1832, 2018.
• J. Johnson, S. Gonzalez, and D.B. Arnold, "Experimental Distribution Circuit Voltage Regulation using DER Power Factor, Volt-Var, and Extremum Seeking Control
Methods," IEEE PVSC, Washington, DC, 25-30 June, 2017.
• D. B. Arnold, M. D. Sankur, M. Negrete-Pincetic and D. Callaway, "Model-Free Optimal Coordination of Distributed Energy Resources for Provisioning Transmission-
Level Services," in IEEE Transactions on Power Systems, vol. 33, no. 1, pp. 817-828, 2017.
• Code: https://github.com/sunspec/prodromos/blob/master/optimization/extemum_seeking_control.py
PF OPTIMIZATION
Objective Function: 9
• Optimization occurs every minute over a 15-min horizon
min w0δ violation (V ) + w1σ (V − Vbase ) + w2 C ( PF )
• OpenDSS simulation is instantiated with PV production PF
forecast and current feeder status (which is assumed to
violation (V )
δ= 1 if any V > Vlim
persist)
• State-estimation determines current feeder loads σ (V − Vbase ) is standard deviation of V − Vbase
• Forecasting tool estimates PV power production C ( PF
= ) ∑1 − PF
• Particle Swarm Optimization (PSO) is used to determine the
optimal PF settings for the DER devices because of nonconvex Cost minimized when voltage = Vbase and PF=1
fitness landscape

Optimal Power Factor

Settings for Time Period

-0.85

-0.88 0.013

-0.91
0.012
-0.94

-0.97 0.011

PV 1 Power Factor

Voltage RMSE
1
0.01
0.97

0.94
0.009

0.91

0.88 0.008

0.85
-0.85 -0.88 -0.91 -0.94 -0.97 1 0.97 0.94 0.91 0.88 0.85

IEDs DER
PV 2 Power Factor
 SIMULATED AND PHYSICAL DER HARDWARE 10

The Electric Power Research Institute (EPRI)-developed PV/ESS Simulator Sandia’s Distributed Energy Technologies
software includes several grid-support functionalities and communication Laboratory (DETL) includes 100
interfaces. Any number of DER devices, with different nameplate capacities and microinverters with independent power
phasing configurations, can be instantiated with independent irradiance profiles. factor settings (±0.85) issued through
The Opal-RT system exchanged P, Q, V, and f points through a TCP/IP Data Bus SunPower PV Supervisors via Power Line
(DBus) interface. DER controls and measurements were issued/gathered from a Carrier. PF commands to PVSs are sent via a
commercial ADMS software vendor via IEEE 1815 (DNP3). Secure Shell (SSH).
 DISTRIBUTION CIRCUIT MODELS 11
Phase A
Phase B
Phase C

OpenDSS models were converted to reduced-order* RT-Lab Substation

PV PCC

models
Switching Capacitor
Relay
Recloser

• The PNM feeder has ~440% PV penetration because of large

utility-scale PV systems PNM Model
• The NG model had distributed PV added to reach 50%
penetration 258 kW 10.0 MW

Before Reduction After Reduction

1.0 MW
Lines Xfmrs Loads Buses Lines Xfmrs Loads Buses
PNM 196 20 20 217 12 2 14 15
NG 1579 4 656 1582 13 3 43 15

Voltage Peak Load Power PV Power

(V) Active (kW) Reactive (kVAR) (kVA)
PNM 7200/277 2568.63 1418.71 11258.00
NG 8000 9494.76 318.10 5495.36

*Circuit reduction methodology:

NG Model
• M. J. Reno, K. Coogan, R. J. Broderick, and S. Grijalva, "Reduction of Distribution Feeders for Simplified PV Impact Studies," in IEEE PVSC, Tampa, FL, 2013.
• Z. K. Pecenak, V. R. Disfani, M. J. Reno, and J. Kleissl, “Multiphase Distribution Feeder Reduction,” IEEE Transactions on Power Systems, 2017.
• Z. K. Pecenak, V. R. Disfani, M. J. Reno, and J. Kleissl, “Comprehensive Reduction of Multiphase Distribution Feeder Models,” IEEE Transactions on Power Systems, 2018.
 IMPLEMENTATIO
N 12

Opal-RT Communication Interfaces

• PMU C37.118 to state estimator
• Opal DataBus Interface receives
P/Q values for EPRI PV
Simulators and transmits bus
voltages and frequency

Information Flow
• The State Estimator ingests PMU
data to produce current/voltage
estimates for the distribution
system
• State estimation data and PV
generation forecasts populate an
OpenDSS model.
• PSO wraps the OpenDSS model to
calculate the optimal PF setpoints
for each of the DER devices.
• DER PF settings are issued through
proprietary SSH commands and
IEEE 1815 (DNP3) commands
 IRRADIANCE PROFILES FOR THE RT SIMULATIONS 13

EPRI provided 3 correlated, highly variable irradiance These data were smoothed using the
measurements from a PV site on the East Coast. Wavelet Variability Model1 based on each
plant size.

1https://pvpmc.sandia.gov/applications/wavelet-variability-model/
 VV VS BASELINE SIMULATION RESULTS 14

The VV function was tested by

programming VV curves into the EPRI
PV simulators
More aggressive slopes on the VV
curve produced oscillations in the
power system because of delays in
the DBus communications
• Oscillations aren’t expected in
physical systems because of fast
voltage measurement speeds and
DER reactive power ramp rates

VV setpoints
P1 = (92, 25)
P2 = (99, 0)
P3 = (101, 0)
P4 = (108, -25)
 ESC VS BASELINE RESULTS 15

ESC produced a sizable

(0.01 pu) ripple on the
voltage from the
probing signals, but did
well to track nominal
voltage.
ESC has an integrator
(“inertia”) so it cannot
respond instantly to the
irradiance changes.
Additional tuning of the
ESC parameters is
expected to tighten the
voltage envelope.
 PF OPTIMIZATION VS BASELINE RESULTS (Preliminary Results) 16

Preliminarily solutions to
the PSO OPF simulations Erroneous
result reasonable PF PSO Solutions
settings most of the time
• PSO converges on
nonoptimal solutions
during some of the
solves - the large PV
system is injecting
reactive power in those
cases
• Team is refining the
optimization and
forecasting
implementation to
ensure consistent
results
Comparison of Methods 17
FUTURE WORK 18

The team will be running Volt-Var, ESC, and

Optimal PF control techniques on the live
National Grid feeder in Grafton, MA.
• 28 PV inverters will be controlled at a
672 kVA PV site.
• A feeder monitor located at a separate
location on the feeder will be used to
collect feeder voltages
• Data will be collected for multiple days
for each control technique to compare
the techniques
DIGITAL TWIN CONCEPT 19

Problem
• Not enough Intelligent Electronic Devices (IEDs, i.e.,
PMUs, DERs, meters, etc.) to make state estimation
observable for the field demonstration
• Short-term load forecasts or historical data is often
used as “pseudo-measurements” to get a solution, but
the team doesn’t have access to this data

Proposal
• Use a real-time digital twin of the feeder to estimate
the system operations
• If general behavior of digital twin is similar to the
physical feeder, the “optimal” PF settings should
support feeder voltages
• PV PF setpoints are sent to the physical and virtual PV
system

Challenges
• This does not account for the current load (only pre-
recorded versions)
CONCLUSION 20

PHIL and RT experiments are effective at providing confidence for field deployments and
demonstrations because:
• Challenges with real-time operations are faced when operating in a PHIL environment.
• Errors or oversights in control algorithms can be remedied early.
• Communications between measurement equipment, ADMS controllers, and DER devices can be
verified.

Many voltage regulation techniques exist—each with their own pros and cons
• Volt-var functionality is an effective method of performing DER voltage regulation without
communications.
• In low communication environments, extremum seeking control is a viable means to control a
fleet of DER devices to track toward optimal PF setpoints, but it is relatively slow and the system
must be tolerant of probing signal ripple.
• State estimation-fed, model-based DER optimization will be a viable approach in the future.
• Optimal DER PFs based on state-estimation is currently impractical because there are not enough feeder
measurements to solve for all distribution system states, even with a reduced-order model
• In the future, ubiquitous DER equipment, µPMUs, and other measurement devices will enable greater degrees of
optimization
 Thank You!

Jay Johnson Adam Summers

Renewable and Distributed Systems Integration Electric Power Systems Research
Sandia National Laboratories Sandia National Laboratories
P.O. Box 5800 MS1033 P.O. Box 5800 MS1140
Albuquerque, NM 87185-1033 Albuquerque, NM 87185-1140
Phone: 505-284-9586 Phone: 505-401-1562
jjohns2@sandia.gov asummer@sandia.gov

This work was completed under the Department of Energy Solar Energy
Technologies Office-funded ENERGISE project, “Voltage Regulation and Protection
Assurance using DER Advanced Grid Functions”