Synthetic Power Grid Models:
What are They,
How They’re Made, and Why
They Matter
Tom Overbye
University of Illinois at Urbana-Champaign
(overbye@illinois.edu)
PSERC Webinar
March 15, 2016
Acknowledgments and Thanks
• Work presented in these slides is based on the
results of several projects including
• PSERC S-62G (Seamless Bulk Electric Grid
Management with EPRI)
• PSERC T-57 (High Impact)
• BPA project TIP 353 (Improving Operator Situation
Awareness by PMU Data Visualization
• ARPA-E Grid Data Synthetic Data for Power Grid R&D
• Support is gratefully acknowledged!
• Thanks also to Adam Birchfield, Kathleen Gegner, Ti Xu,
Komal Shetye, Richard Macwan, Profs Bob Thomas,
Anna Scaglione, Zhifang Wang and Ray Zimmerman
2
Presentation Overview
• Access to data about the actual power grid is often
restricted because of requirements for data
confidentiality (e.g., critical energy infrastructure)
• Focus here is on high voltage power flow, optimal
power flow, transient stability models, SCADA, PMUs
• Some data is public, some is available by NDAs, and
some is essentially unavailable to those outside of
power system control centers
• Focus of talk is on the creation of synthetic
(fictional) models that mimic the complexity of the
actual grid cases but will contain no confidential
data and can be publicly available
3
A Few Initial Thoughts
• The reason why this matters is to help spur
innovation in the electric grid software
• Algorithms tested on synthetic models applied to actual
• In 2000 the NAE named Electrification (the vast
networks of electricity that power the developed
world) as the top engineering technology of the
20th century
• automobiles (2), airplanes (3), water (4), electronics (5)
• Our challenge in this century is to develop a
sustainable and resilient electric infrastructure for
the entire world
4
A Few Initial Thoughts
• "All models are wrong but some are useful,“
George Box, Empirical Model-Building and Response
Surfaces, (1987, p. 424)
• “The use of nondisclosure agreements or NDA’s
to obtain data, while useful in many instances, is
not useful if the world community is to engage in
research that adheres to the scientific principle
of reproducibility of results by other qualified
researchers and to use important findings to
advance their own work“
PSERC Founding Director Bob Thomas, 2015
5
Overall Goals
• The development of entirely synthetic transmission
system models and scenarios that match the
complexity and variety of the actual grid
• Models that incorporate both the average characteristics
and outlier characteristics of the actual grid
• Models and scenarios suitable for security constrained
optimal power flow (SCOPF) studies; they will also be
set for use in transient stability and geomagnetic
disturbance analysis
• All models will have embedded geographic coordinates
• Scenarios will be SCOPF validated
• We want to partner with industry!
6
The Need
• Few, if any, of the existing public models (such as
the IEEE 300 bus) match the complexity of the
models used for actual large-scale grids
• Issues include size, with the
Eastern Interconnect models
now more than 70,000 buses,
and also model complexity
3
176 MW
105 Mvar
3
3
100 MW
75 Mvar
96 MW
42 MW
14 Mvar
3
3
3
3
3
16 Mvar
22 MW
77 MW
33 Mvar
3
77 Mvar
3
3
MW
361
30 Mvar
3
-114 MW
3
7 Mvar
47 MW
26 Mvar
8 MW
3 Mvar
3
131 MW
96 Mvar
3
38 MW
13 Mvar
3
3
171 MW
3
61 MW
30 Mvar
70 Mvar
3
328 MW
188 Mvar
3
3
3
3
3
3
404 MW
212 Mvar
3
3
29 MW
14 Mvar
3
3
3
35 MW
12 Mvar
157 Mvar
3
489 MW
53 Mvar
3
3
10 MW
3 Mvar
96 MW
46 Mvar
12 MW
2 Mvar
3
1
58 MW
14 Mvar
2
2
78 MW
1 Mvar
2
276 MW
59 Mvar
2
-100 MW
2
2
34 Mv ar
515 MW
381 MW
37 Mvar
14 MW
650 Mvar
2
1200
2
60 MW2
24 Mvar
60 MW
24 Mvar
2
2
2
2
2
137 MW
17 Mvar
17 MW
9 Mvar
2
85 MW
2
2
24 MW
14 Mvar
110 Mvar
1
1
155 MW
18 Mvar
1
1
86 MW
0 Mvar
-5 MW
1
5 Mvar
2
2
1
1
6 Mvar
5 MW
0 Mvar
9 Mvar
83 MW
0 Mvar
1
112 MW
0 Mvar
25 MW
-1 Mvar
50 MW
0 Mvar
1
63 MW
0 Mvar
1
1
1
1 MW
0 Mvar
1
1
2 MW
1
1
1
1
1
-4 Mvar
66 MW
0 Mvar
1
1
1
1
85 MW
32 Mvar
2
1
1
9 MW
0 Mvar
1
1
46 MW
-21 Mvar
1 Mvar
-11 MW
-1 Mvar
1
41 MW
19 Mvar
2
30 MW
1
13 Mvar
1
208 MW
107 Mvar
16 MW
0 Mvar
1
1
37 MW
74 MW
28 Mvar
1
2
74 MW
29 Mvar
1
1
1
32 MW
0 Mvar
1
176 MW
83 Mvar
56 MW
25 Mvar
1
69 MW
3 Mvar
55 MW
1
19 Mvar
1
67
1 MW
0 Mvar
1
39 MW
1
2
1
0 Mvar
1
57 MW
1
10 MW
1 Mvar
20 MW
3 Mvar
1
1
1
92 MW
26 Mvar
1
14 MW
1
1
1
1
1
1
116 MW
38
1 Mvar
48 MW
14 Mvar
28 MW
1
-20 Mvar
56 MW
20 Mvar
1
1
-15 MW
26 Mvar
1
2
2
224 MW
71 Mvar
1
40 MW
1
2
49 Mvar
5 MW
4 Mvar
2
89 MW
35 Mvar
1
0 Mvar
1
78 MW
0 Mvar
2
2
427 MW
174 Mvar
2 MW
69
28 MW
12 Mvar
2
2
163 MW
43 Mvar
218 MW
227 MW 106 Mvar
2
2
25 Mvar
2
2
26 MW
0 Mvar
1
2
2
145 MW
58 Mvar
1
1
116 MW
-24 Mvar
1
50 MW
1 17 Mvar
1
1
60 MW
0 Mvar
1 Mvar
24 Mvar
115 Mvar
2
2
2
1
14 MW
1
2
63 MW
2
58 MW
12 Mvar
1
7 Mvar
1
21 MW
7 Mvar
1
22 MW
10 Mvar
61 MW
28 Mvar
388 MW
35 MW
15 Mvar
241 MW
89 Mvar
595 MW
83 Mvar
1
1
100 Mvar
40 MW
4 Mvar
2
2
28 MW
1
29 Mvar
1
1
75 MW
2
274 MW
55 MW
18 Mvar
2
195 MW
1
1
200 MW
50 Mvar
2
0 MW
20 Mvar
2
2
1
26 MW
0 Mvar
1
145 MW
-35 Mvar
50 Mvar
2
73 MW
0 Mvar
2
2
2
169 MW
42 Mvar
2 70 MW
5 Mvar
1
98 MW
20 Mvar
1
2 MW
-13 Mvar
12 Mvar
1
1
1
2
58 MW
5 Mvar
2
124 MW
-24 Mvar
17 Mvar
1
17 MW
0 Mvar
30 Mvar
1
135 Mvar
2
2
33 MW
• Public models also lack extra
data like transient stability
183 MW
44 Mvar
2
827 MW
2
229 MW
2
12 Mvar
2
70 MW
45 MW
1
1
1
2
2
2
83 Mvar
2
1190
2
764 MW
291 Mvar
2
1
31 MW
0 Mvar
1
2
29 Mva r
1
49 Mvar
55 Mvar
100 MW
90 MW
1
44 MW
0 Mvar
1
28 MW
9 Mvar
1
1
1
1
300 MW
96 Mvar
535 MW
69 MW
1 13 Mvar
18 MW
0 Mvar
1
0 Mvar
20 MW
1
1
285 MW
100 Mvar
1
1
127 MW
23 Mvar
23 Mvar
-21 MW
-14 Mvar
15 Mvar
1
2
2
81 MW
96 MW
1
43 Mvar
56 MW
1
1
2
1
1
1
160 MW
60 Mvar
1
353 MW
130 Mvar
1
120 MW
41 Mvar
1
561 MW
220 Mvar
1
41 MW
14 Mvar
205 Mvar
482 MW
77 MW
1 Mvar
1
21 Mvar
3
72 MW
24 Mvar
-29 Mvar
2
1
1
3
-33 MW
2
2
1
83 MW
3
1
58 MW
10 Mvar
148 MW
33 Mvar
3
3
3
3
595 MW
120 Mvar
7 MW
2 Mvar
0 MW
-5 Mvar
3
369 Mvar
3
3
3
38 MW
13 Mvar
3
3
3
3
3
3
3
159 MW
107 Mvar
538 MW
64 MW
21 Mvar
7 MW
2 Mvar
3
148 Mvar
3
448 MW
143 Mvar
-23 MW
-17 Mvar
3
3
572 MW
244 Mvar
223 MW
3
43 MW
14 Mvar
3
27 MW
12 Mvar
3
269 MW
410 MW
40 Mvar
3
800 MW
72 Mvar
29 MW
14 Mvar
3
255 MW
149 Mvar
428 MW
232 Mvar
173 MW 3
99 Mvar
4 MW
1 Mvar
1
30 MW
23 Mvar
2 MW
1 Mvar
3 MW
1
1
2
1 Mvar
1
1
1
3 MW
1 Mvar
• Innovation is hindered by not
being able to compare results for complex models
1
1
1 MW
0 Mvar
1
1
1
1
0 MW
0 Mvar
1
1
1
1 MW
0 Mvar
1
1
1
3 MW
1 MW
0 Mvar 1 Mvar
1
1
1
1
1
1 MW 2 MW 2 MW 2 MW 2 MW
0 Mvar 1 Mvar 1 Mvar 1 Mvar
1 Mvar
1
1
1 MW
0 Mvar
5 MW
2 Mvar
1
2 MW
1 Mvar
1
2 MW
1 Mvar
1
1 MW
0 Mvar
1
1 MW
0 Mvar
0 MW
0 Mvar
Image: IEEE 300 Bus case downloaded from http://icseg.iti.illinois.edu/ieee-300-bus-system/
7
What Makes a Model Real?
• The challenge is to capture the essence of what
makes actual grid models different
• Actual grid models are
quite diverse
• Statistics can be used
to quantify some of the
characteristics
• topology, parameters for
buses, generators, loads,
transmission lines, transformers, switched shunts,
transient stability and GMD parameters
• System-wide metrics are also needed
8
Complexity Examples
• A recent 76,000 bus Eastern Interconnect (EI)
power flow model has 27,622 transformers
including 98 phase shifters
• Impedance correction tables are used for 351, including
about 2/3 of the phase shifters; tables can change the
impedance by more than two times over the tap range
• The voltage magnitude is controlled at about
19,000 buses (by Gens, LTCs, switched shunts)
• 94% regulate their own terminals with about 1100 doing
remote regulation. Of this group 572 are regulated by
two or more devices, 277 by three or more, twelve by
eight or more, and three by twelve devices!
9
How to Make Realistic, GeographicallyBased, Synthetic Models
• Our approach is to make models that look real
and familiar by siting these synthetic models in
North America, and serving a population density
the mimics that of North America
• The transmission grid is, however, totally fictitious
• Goal is to leverage widely available public data:
•
•
•
•
Geography
Population density (easily available by post office)
Load by utility (FERC 714) and state-wide averages
Existing and planned generation: Form EIA-860
contains information about generators 1 MW and
larger; data includes location, capacity and fuel type
10
Example: 2100 Bus Texas Case
Frequency Response
11
EIA-860 Generator Data
• Online at www.eia.gov/electricity/data/eia860/
Since our goal is to make entirely synthetic models, no
existing company names will be used. We may be
changing the actual generator capacity values as well.
12
How to Make Realistic Synthetic Models
• First step is to select a desired size (bus count)
and geographic footprint
• These are two independent parameters: for example,
geographically large with a small number of buses
• Our approach does not require that we use actual
geography; however most, if not all, of our models will
• Requires an assumption on underlying load density
• Nominal transmission voltages need to be selected
(e.g., 500/230/115 kV); we will allow multiple levels
• On larger models the geographic footprint is divided into
balancing authority areas and fictitious owners
13
How to Make Realistic Synthetic Models:
Substation Selection
• The next step is to site the substations
• Buses are located in substations; number of buses in a
substation can vary widely
• Most substations have load and/or generation; number
of buses can depend on model assumptions, such as
whether generator step-up transformers are modeled
• Substation are sited geographically primarily in
order to meet load and generation requirements
• One approach for the assumed load density is mimic
population density as given by zip code information
• Number of substation depends on the desired model
size; in actual models the amount of substation load can
widely vary (from 1 MW to more than 500 MW)
14
How to Make Realistic Synthetic Models:
Substation Selection
• In our approach substations are placed
geographically at post offices
• The load is proportional to
population, taking into
account state variation
• Hierarchical clustering is
used to reduce the number
of substations as needed
• Load is usually attached at
lowest-voltage bus
15
Generator Substation Placement
• Based on actual model statistics, some
generation is located at existing load
substations
Statistics derived from
• Other plants are combined into real
power system case
generator-only substations
• Generator parameters,
including reactive power
limits and cost information,
are derived from statistics
• Transient stability models
are added
16
Substation Voltage Levels
• Each substation now has load/generation defined
• Statistically about 90% in actual grid have load or gen
• Different system voltage levels are chosen
• E.g., 500/161, 765/345/138, 500/230/115
• Almost all substations have lower voltage bus
• A percent of substations (e.g.,15%) also include
higher voltage buses and transformers
• Higher-voltage substations are iteratively selected
with probabilities proportional to load
• All large (> 250 MW) generators are placed at the
higher voltage level, but with a GSU
17
Adding Transmission Lines
• Substations are connected together by
transmission lines, matching characteristics of
actual models
• Builds on pioneering work done by PSERC researchers
Thomas, Wang and Scaglione
• Z. Wang, R.J. Thomas, A. Scaglione, “Generating random
topology power grids,” HICSS-41, HI, Jan 2008
• Z. Wang, A. Scaglione and R.J. Thomas, “The Node Degree
Distribution in Power Grid and Its Topology Robustness under
Random and Selective Node Removals”, the 1st IEEE
International Workshop on Smart Grid Communications, Cape
Town, South Africa, May 2010
• Z. Wang, R.J. Thomas, “On Bus Type Assignments in Random
Topology Power Grid Models”, HICSS-48, Jan. 2015
18
Substation Node Degree
(Number of Neighbors)
• Need to match statistics for number of connected
substations at each voltage level
• Average nodal degree
, nearly constant
with for single-voltage networks in EI
• Number of lines
• Node degree
distribution appears
to be exponential.
.
(except for k=1 and 2)
19
Adding Transmission Lines
• Graph theory considerations are used to
determine which substations are connected
•
An approach is to do Delaunay triangulation along with minimum
spanning tree (MST) analysis
Image shows Delaunay triangulation
of 42,000 North America substations;
statistics only consider single voltage
levels; computationally fast (order n ln(n))
20
Adding Transmission Lines
• In general, transmission line topologies are totally
connected, and remain so with one node removed
• Typical actual power system contains 60% of its
substations’ minimum spanning tree (MST) at
each nominal voltage level (percent varies by
voltage level)
• Approach is to match the MST percentage
• Then other lines are added to match the typical
average (
edges per bus)
21
Using Delaunay Triangulation to Add
Additional Lines
• Delaunay triangulation
• No triangle’s circumcircle
contains another point
• Nearest few neighbors are
connected
• Statistics
and
match
regular lattice and actual grid
• Contains 70% of real lines
on average, and 98%
separated by 3 hops or less
• We select subset out of
Delaunay’s
segments
22
Transmission Line Parameters
• Transmission line parameters from EPRI &
ACSR guides
• Different configurations for each voltage level:
These
parameters are
validated
against real
transmission
lines
23
Iterative Updates to Obtain a Feasible
DC Power Flow Solution
• A connected graph allows dc power flow solutions
• Iteratively add lines to obtain a dc power flow with
no line flow violations
• Candidate lines are segments of the Delaunay
triangulation or near neighbors
• Place total of
lines per voltage level
• Select lines based on:
•
•
•
•
Voltage angle gradient, indicating likely power flow
Avoid radial substations
Encourage parallel circuits to overloaded lines
Forbid lines exceeding a maximum length
24
Example: Transmission Line Placement
• Based on voltage angle gradient, this might be a
good location for a transmission line
25
Reactive Compensation and Additional
Model Complexity
• The next step is the specification of the generator
PV bus setpoints, the inclusion of additional
reactive power control devices such as switched
shunts and LTC control, and the inclusion of
additional complexities such tap dependent
impedances (XF correction tables)
• Realistic remote generator PV control will be modeled,
including reactive power sharing among a number of
generator
• A hypothesis we are considering is that the difficulties
encountered with actual models compared to public
models, such as the IEEE 118 bus case, are due to
these complexities
26
Model Creation Methodology:
Inclusion of Additional Parameters
• The final step in the creation of
the models themselves will be
the inclusion of the models
necessary to do transient stability
and GMD analysis
• As with the other models, parameters
will be set to
match the statistics
of the actual grid
Images show example
transient stability models
and parameters
27
Example: 150 Bus Network for
GMD Analysis
TN
E L I Z A B E T H T O N _ 3 7 6 4 3
5 0 0 . 0 0
k V
A
Amps
A
TN
J ohn Se vie r
5 0 0 .
0 0
TN
k V
Amps
J O N E S B O R O U G H _ 3 7 6 5 9
5 0 0 . 0 0
k V
TN
C L A R K S V I L L E _ 3 7 0 4 2
5 0 0 . 0 0
k V
A
TN
Amps
A
C L A R K S V I L L E _ 3 7 0 4 0
5 0 0 .
0 0
k V
A
Amps
Amps
TN
G a la t in
( T N )
5 0 0 . 0 0
k V
A
Amps
TN
A
P A R I S _ 3 8 2 4 2
J E F F E R S O N
A
C I T Y _ 3 7 7 6 0
5 0 0 . 0 0
Amps
A
Amps
k V
TN
C O O K E V I L L E _ 3 8 5 0 1
Amps
2
5 0 0 . 0 0
k V
TN
TN
2
5 0 0 . 0 0
A
K N O X V I L L E _ 3 7 9 2 1
A
N A S H V I L L E _ 3 7 2 1 5
5 0 0 . 0 0
k V
Amps
k V
TN
A
5 0 0 . 0 0
Amps
A
k V
Amps
A
K in g s t o n
5 0 0 . 0 0
A
Amps
TN
A
Amps
TN
k V
Amps
L O U D O N _ 3 7 7 7 4
M
This is a synthetic power
system model that does NOT
represent the actual grid. It is
intended for algorithm testing
and education use. It contains
no CEI information.
A R Y V I L L E _ 3 7 8 0 4
2
5 0 0 . 0 0
k V
Amps
TN
J o h n s o n v ile
7
5 0 0 .
0 0
k V
A
TN
M
TN
U R F R E E S B O R O _ 3 7 1 2 9
5 0 0 . 0 0
k V
M
U R F R E E S B O R O _ 3 7 1 3 0
Amps
5 0 0 .
0 0
k V
2
A
A
Amps
A
Amps
A
Amps
Amps
D A Y T O N _ 3 7 3 2 1
A
Amps
TN
D U N L A P _ 3 7 3 2 7
A T H E N S _ 3 7 3 0 3
5 0 0 . 0 0
k V
TN
A
TN
L a g o o n
5 0 0 . 0 0
J A C K S O N _ 3 8 3 0 5
5 0 0 .
0 0
k V
Amps
C r e e k
k V
7
A
Amps
TN
T U L L A H O M
2
A _ 3 7 3 8 8
5 0 0 . 0 0
k V
TN
S equ oya h
A
A
A
Amps
Amps
7
A
5 0 0 .
0 0
k V
A
Amps
Amps
Amps
T im
s
TN
F o r d
H I X S O N _ 3 7 3 4 3
2
5 0 0 .
0 0
k V
TN
C O R D O V A _ 3 8 0 1 6
5 0 0 . 0 0
k V
A
P ic k w ic k
Amps
TN
L a n d in g
5 0 0 . 0 0
D a m
k V
TN
M
E M
P H I S _ 3 8 1 1 1
5 0 0 . 0 0
k V
TN
E L I Z A B E T H T O N _ 3 7 6 4 3
5 0 0 . 0 0
k V
A
Amps
A
TN
J ohn Sev ier
5 0 0 . 0 0
TN
k V
J O N E S BO R O U G H_ 3 7 6 5 9
Amps
5 0 0 . 0 0
k V
TN
C L AR K S V I L L E _ 3 7 0 4 2
2
5 0 0 . 0 0
A
k V
Amps
TN
A
C L A R KS V I L L E _ 3 7 0 4 0
5 0 0 . 0 0
k V
Amps
A
Amps
TN
Gallat in ( TN)
5
5 0 0 . 0 0
k V
A
Amps
M
A R T I N _ 3 8 2 3 7
TN
A
P A R I S _ 3 8 2 4 2
L E B A N ON _ 3 7 0 8 7
J E F F E R S O N
A
2
CI T Y _ 3 7 7 6 0
5 0 0 . 0 0
k V
Amps
A
Amps
TN
C O O K E VI L L E _ 3 8 5 0 1
2
Amps
5 0 0 . 0 0
k V
TN
2
5 0 0 . 0 0
A
K N O X V I L L E _ 3 7 9 2 1
TN
A
N A S HV I L L E _ 3 7 2 1 5
2
5 0 0 . 0 0
k V
k V
Amps
Amps
TN
K N O X VI L L E _ 3 7 9 1 9
A
5 0 0 . 0 0
A
k V
Amps
A
A N T I O C H _ 3 7 0 1 3
A
5 0 0 . 0 0
A
Amps
TN
Kin gst on
Amps
TN
k V
L O U D O N_ 3 7 7 7 4
Amps
M
A R Y V I L L E _ 3 7 8 0 4
5 0 0 . 0 0
S M
Amps
Y R N A _ 3 7 1 6 7
TN
J ohns onv ile
7
5 0 0 . 0 0
k V
A
TN
M
F R A N K L I N _ 3 7 0 6 4
TN
U R F R E E S B O R O _ 3 7 1 2 9
5 0 0 . 0 0
k V
M
Amps
U R F R E E S B O R O _ 3 7 1 3 0
2
5 0 0 . 0 0
k V
A
A
Amps
Amps
A
A
Amps
Amps
D A Y T O N _ 3 7 3 2 1
A
Amps
TN
D U N L A P _ 3 7 3 2 7
A T H E N S _ 3 7 3 0 3
5 0 0 . 0 0
A
TN
J A C K S O N_ 3 8 3 0 5
5 0 0 . 0 0
TN
L a g o o n
7
5 0 0 . 0 0
C r e e k
k V
Amps
k V
A
Amps
B R I G H T ON _ 3 8 0 1 1
TN
T U L L A H O M
5 0 0 . 0 0
A _ 3 7 3 8 8
k V
k V
k V
This is a synthetic power
system model that does NOT
represent the actual grid. It is
intended for algorithm testing
and education use. It contains
no CEI information.
Images
show
a synthetic
150 bus
model placed
geographically
in Tennessee;
bottom image
shows
response to
an assumed
GMD.
TN
Sequoy ah
A
A
Amps
A
7
5 0 0 . 0 0
k V
A
A
Amps
Amps
Amps
Amps
T im
s
TN
F o r d
H I X S O N _ 3 7 3 4 3
2
5 0 0 . 0 0
k V
S A VA N N A H _ 3 8 3 7 2
TN
C OR D O V A _ 3 8 0 1 6
2
5 0 0 . 0 0
k V
A
TN
Amps
P ic k w ic k
L a n d in g
5 0 0 . 0 0
D a m
k V
TN
M
E M
P HI S _ 3 8 1 1 1
5 0 0 . 0 0
k V
28
1500 Substation, 2100 Bus Texas Example
• Texas geographic footprint
• No relationship to actual
transmission grid
• Nominal 345/115 kV grid
• 1500 substations,
2092 buses, 282 gens,
2857 branches
• Automatic line placement
takes about 70 seconds
• Currently we are supplementing with
manual adjustment for voltage control
29
Example Case: Initial Generation Dispatch
• System divided into 8 areas
• Two areas have more load
than generation capacity
• Transactions set up
from other areas
• Generators dispatched
proportionally to meet
load + transaction commitments
• This is done before lines are placed, so that the
algorithm’s dc power flow reflects realistic
generation dispatch
30
Example Case: Voltage Phase Angle Contour
• Gradual voltage
angle gradient
• All branches less
than 90% loaded
• Average branch is
28% loaded,
matching real cases
• These properties are
direct result of automatic
line placement without
manual intervention
31
Example Case: Voltage Control
• All voltages within
0.97-1.05 pu in
base case
• After line placement
algorithm voltages were
within 0.9 to 1.1 pu
• Adjustment of generator
set points and insertion
of 33 shunt capacitor
banks in urban areas
32
Simulating High Impact, Low Frequency
Events: Results can be Exchanged!
40
35
30
E3(t) (V/km)
25
20
15
10
5
0
-5
-10
0
10
1
2
10
10
3
10
Time (s)
First 60 seconds of
IEC 61000-2-9
33
Synthetic Model Validation
• Key to this research is to demonstrate synthetic
models have similar properties of actual grids
• Synthetic models are not meant to represent the
actual grid, so direct comparison is not appropriate
• Useful metrics are
• Topological properties, which we meet by design
• Individual model parameters, which we meet by design
• Solution algorithm properties, such as power flow
convergence
• Solution results, such as LMPs, amount of congestion,
transient stability damping, etc.
34
Driving Innovation!
• Goal is to publicly release synthetic models of
various sizes and complexities
• Algorithm results from synthetic models can be
published without restriction; algorithms can be used
confidentially on real models
• Fully public, anyone can make derivative models; some
models will be standardized for comparisons purposes
• Large-scale models can be used to compare
software packages
• Customers and researchers can compare results
• Visualization research not hindered by confidentiality
35
Thank You!
Questions?
overbye@illinois.edu