Data Challenges for Wide-Area Modeling, Analysis and
Control using Synchrophasors
Aranya Chakrabortty
North Carolina State University, Raleigh, NC
Carnegie Mellon University, PA
14st March, 2012
Main trigger: 2003 Northeast Blackout
NYC before blackout
New England
Ohio
Power flow
INTER-AREA
STABLE
NYC after blackout
Hauer, Zhou & Trudnowsky, 2004
Kosterev & Martins, 2004
Lesson learnt:
1. Wide-Area Dynamic Monitoring is important
2. Clustering and aggregation is imperative
INTER-AREA
UNSTABLE
Advanced Transmission & Distribution
Wid A / Global
Wide-Area/
Gl b l View
Vi
The electric power network is a large interconnected graph
Whatever I do is going to affect my neighbors &
propagate the disturbance
Must be able to measure transient signals at limited
number
b off buses
b
• Phasor Measurement Units
• High sampling rate digital
devices
• GPS synchronized
• 30-60 samples per second
Phadke et al., 1981
Zhou, Pierre & Trudnowsky, 2006
Phadke & Thorpe, 2007
Arbiter
ABB
Schweitzer
Global behavior must be understood from the limited local measurements
JF
KR
MF
RK
OR
OS
JF
KR
PMU Voltage Magnitudes in AEP - Richview Event
MF
RK
OR
OS
Frequencies in AEP - Oconee Event
60.01
1.04
60
59.99
Frequency (H
Hz)
1.02
1.01
1
59.97
59.96
59.95
59 94
59.94
0 99
0.99
0.98
59.98
59.93
0
500
1000
1500
2000
2500
3000
Time in seconds (sec.) with origin at: 12:38:00 EST
3500
1700
1750
1800
1850
Time in seconds (sec.) with origin at: 16:24:00 EST
Swing Component of Potential Energy ( VA-s )
Voltage Magnitud
de (pu)
1.03
3.5
NASPI – North American Synchrophasor Initiative
(www.naspi.org)
- jointly held by NERC and DOE 3 times a year
3
Kinetic Energy ( MW-s )
• System wide visibility
• Real time streaming information
• Rapid assessment of actual system conditions through
visualization of information
• Event triggered alarms
• Performance Metrics
2.5
2
1.5
1
0.5
0
50
100
150
200
Time(sec)
250
300
1900
3.5
3
2.5
2
1.5
1
0.5
0
50
100
150
200
Time (sec)
250
300
Disturbance events
Maximum swing energy in
Transfer Path 1
Maximum swing energy in
Transfer Path 2
1
3 MW-s
0.198 MW-s
2
3.5 MW-s
17 MW-s
Distributed Architecture for PMU Applications
Two vulnerable
points where
data can be
hacked into, and
manipulated
Ongoing
g g PMU Projects
j
Wide-area Modeling
Wid
Wide-area
C
Control
l
Wide-area Monitoring
Distributed State Estimation
JF
KR
MF
RK
OR
OS
JF
Frequencies in AEP - Oconee Event
KR
MF
RK
OR
OS
PMU Voltage Magnitudes in AEP - Richview Event
60.01
1.04
60
1.03
Voltage Magnitud
de (pu)
Frequency (H
Hz)
59.99
59.98
59.97
59.96
59.95
59.94
64 Bus New England Power System Model
Security if imperative for all three applications
(real-time and offline)
1.01
1
0.99
59.93
0.98
1700
•
1.02
1750
1800
1850
Time in seconds (sec.) with origin at: 16:24:00 EST
1900
0
500
1000
1500
2000
2500
3000
Time in seconds (sec.) with origin at: 12:38:00 EST
3500
Derive Dynamic Mathematical Model of the Clusters
System-wide Oscillation Detection - How power generation clusters oscillate
Winnipe
g, MB,
CA
0.22-Hz Mode
Shape
Bismarc
k, ND
St. Paul,
MN
Ames, IA
Grand
Rapids,
MI
Chicago,
IL
Toronto,
CA
Troy,
NY Holyoke,
MA
Princeton
, NJ
Blacksburg,
VA
Huntsvill
e, AL
Tallahassee, FL
•
•
•
How do different oscillatingg zones of the system
y
oscillate with
respect to each other after a disturbance just like an interconnected
pendulum
Dynamic PMU data are ideal for answering these questions
Hi hl useful
Highly
ef l ffor system
te le
level
el studies
t die for
f operation
e ti & planning
l i
Model Aggregation using distributed PMU data
6-machine, 30 bus, 3 areas
Problem Formulation:
1. Model Reduction
PMU
• How to form an aggregate model
from the large system
Area 1
PMU
PMU
Aggregate
Transmission
Network
Area 2
Area 3
4
Two-Dimensional Models
More than Two Areas: Pacific AC Intertie
Salient Points
Vn  J1 ( x ) 1 (t )  J 2 ( x )  2 (t )
• Current in each branch is different
• No single spatial variable a
• Derivations need to be done piecewise
(each edge of the star)
• Two interarea
i
modes/
d / relative
l i states – δ1 & δ2
Vn
J1 ( x)1 (t*)  J 2 ( x) 2 (t*)

4
4
4
Vn
J1 ( x)1 (t*)  J 2 ( x) 2 (t*)
3
3
3
Time-space separation property lost!
20
Application to WECC Data
• 2000 gens • 11,000 lines
• 22 areas, 6500 loads
Colstrip
Grand
Coulee
M li
Malin
SVC
Vincent
13
Application to WECC Data
1.1
Colstrip
Grand
Coulee
1.08
1.06
Bus Voltage (pu)
• 2000 gens • 11,000 lines
• 22 areas, 6500 loads
Bus 1
Bus 2
Midpoint
1.04
1.02
1
0.98
0.96
0.94
0
M li
Malin
50
100
150
200
Time (sec)
250
300
SVC
Needs processing to get usable data
Vincent
• Sudden change/jump
• Oscillations
• Slowly varying steady-state (governer
effects)
14
Application for Stability Assessment
PMU
PMU
Full Model
Aggregated Model
PMU
Signal
Processing
PMU 1
PMU 3
PMU 3
Aggregated
Model
Modes, Amplitude,
Residues, Eigenvectors
Metrics indicating the dynamic interaction
between the areas
16
Energy Functions for Two-machine System
Gen1
~
V2 2
V11
P
Gen2
~
jxe
P
Load
S  S1  S 2 
n ( n 1) / 2 z j

j 1
 j (k ) dk 
 ij *
n
Mj
j 1
2

   ,
2 H   Pm 
E1E2 sin  
xe'
E1E2 sin  
xe'
 j2
E1 E 2

[cos(  op )  cos (δ )  sin(  op )(  op   )]  H 2
x e
Potential Energy
Kinetic Energy
Using IME algorithm: x,e ' E1, E2, δ = δ1- δ2, δop, ω = ω1-ω
ω2 & H are computable from
xe, V1, V2, θ = θ1- θ2, θop, ν=ν1-ν2 & ωs
• Note
N t : θop = pre-disturbance
di
b
angular
l separation
i
17
Energy Functions for WECC Disturbance Event
Angular Difference  ((deg)
74
IME
70
66
62
58
0
50
Sending End and Receiving End
Bus Angles
2.5
x 10
100
150
Time (sec)
200
250
300
Angle difference between
machine internal nodes
-3
Machine Speed D
Difference (pu)
2
1.5
1
0.5
0
-0.5
-1
-1.5
1
-2
0
Machine speed difference
50
100
150
Time (sec)
200
250
300
18
Energy Functions for WECC Disturbance Event
3.5
Total Energy = Kinetic Energy + Potential Energy
2.5
3.5
2
1.5
1
0.5
0
50
100
150
200
Time(sec)
250
300
Sw
wing Component o
of Potential Energy
y ( VA-s )
Kinetic Energy
3.5
3
2.5
2
Swiing Energy Fu
unction V E ( M
MW-s )
Kin
netic Energy ( MW--s )
3
3
2.5
2
1.5
1
0.5
0
15
1.5
1
50
100
150
200
Time (sec)
250
300
• Total energy decays exponentially – damping stability
0.5
0
50
100
150
200
Time (sec)
250
300
osc.
• Total energy does not oscillate – Out - of - phase osc
– Damped pendulum
Potential Energy
19
WatchDog: A Software Visualization Tool for Wide- Area Monitoring Wide
Area Monitoring
- Joel Anderson and A. Chakrabortty
* Phase Angle Contour Videos
* Power
Power-angle
angle Curves
Curves
* Modal Analysis through ERA
* Inter-area response Visualization
* Statistical Analysis/baselining
* Streaming real
Streaming real-time
time data
data
from 7 PMU stations
* 3D baselining plots
* PMU Placement Algorithms
P-Delta
P
DeltaCurves
Curves
Mode Extraction Information
ModeExtractionInformation
The Wide-area Damping Control
• Computational complexity sharply increases with number of areas
Area Identification
Border Buses
Control AllocationAllocationInverse Problem
Network Reduction *
PMU
PMU
PMU
PMU
• Please see CDC 2011, ACC 2012 papers by Chakrabortty on model-reference control (MRC) approach,
Distributed Architecture for PMU Applications
State-of-art Centralized Processing
Chakrabortty, Michailidis and Xin, CDC 2012
Distributed PMU Networks
30
O ti i ti variables:
Optimization
i bl
Data
D
t exporting
ti rates
t off PMU
PMUs or virtual
i t l PMUs
PMU
Information update rate between PDCs – local and inter-area
Network bandwidth
Constraints: Network delays,
y , Processingg delays,
y , Congestion,
g
, Dynamic
y
routingg
•
Define an execution metric M for each application of the PMU-net
•
p
g centralized Mc
Minimize the error between distributed Md and corresponding
Dynamic Rate
Control Problem
(DRCP)
Dynamic Link
Assignment
Problem (DLAP)
31
Phasor Lab
1.
2.
3
3.
Real PMU Data from WECC (NASPI data)
RTDS-PMU Data (Schweitzer PhasorLab)
FACTS TNA with
FACTS-TNA
ith NI CRIO PMU
PMUs
We can provide all three data via our new
PMU and RTDS facilities at the FREEDM
center
How about setting up an
intra-campus local PMU
communication network
with RENCI/UNC and Duke Univ.?
Real Time Digital Simulators
NI PMU Rack
PMU Data Visualization
SEL/ABB PMU Rack with PDC
Conclusions
1. WAMS is a tremendously promising technology for smart grid researchers
2 Communications
2.
C
i ti
andd Computing
C
ti mustt merge with
ith power engineering
i
i
3. Plenty of new research problems – EE, Applied Math, Computer Science
4. Cyber-security is essential
5 Right time to think mathematically – Network theory is imperative
5.
6. Right time to pay attention to the bigger picture of the electric grid
7. Needs participation of young researchers!
8. Promises to create jjobs and provide
p
impetus
p
to the ARRA
Acknowledgements:
1. Financial support provided by NSF & SCE
2. PMU data and software provided by SCE, BPA, and EPG