Oscillation Monitoring System...Interarea Mode Analysis • Identify interarea modes and related...
Transcript of Oscillation Monitoring System...Interarea Mode Analysis • Identify interarea modes and related...
1
PSERC
Mani V. Venkatasubramanian
Washington State University
Oscillation Monitoring System
PSERC S29 Project Tele-seminar
October 28, 2008
2
PSERC
Overview of S29 project
• S19 project from June 2002 to May 2005
• S29 project from June 2006 to July 2008
• Detection, Prevention and Mitigation of Cascading Events
• Three tasks:
• Task I: Detection: Mladen, Texas A&M
� Advanced warning
• Task II: Prevention: Mani, Wash. State
� Wide-area monitoring and controls
• Task III: Mitigation: Vijay, Iowa State/ASU
� Adaptive islanding
• S29 Focus on Prototype Implementations
3
PSERC
Task II: Mitigation
• Implementation of Wide-area Small-signal Stability controller, Wash. State University
• Two subtasks:
� Reliable Oscillation detection:• Multi-input Prony, Matrix Pencil and HTLS algorithms
• Rules for real-time analysis of data
• Noise? Linear versus nonlinear? Switching events?
� Real-time design of damping controls:• Which control to trigger? What design?
• Not part of the prototype testing
4
PSERC
Problem Overview
• Low frequency electromechanical oscillations
� Local or inter-area oscillations
� Insufficient damping
� Example - Aug 10, 1996 WECC blackout
• Detection and control of small signal stability problem in power systems
• Research supported by PSERC, TVA, Entergy, BPA and EPG (CERTS)
5
PSERC
Task II Project Team• WSU:
• Guoping Liu, Qiang Zhang, Jaime Quintero, Mani V. Venkatasubramanian
• TVA:
� Ritchie Carroll, Gary Kobet, Lisa Beard• Entergy:
� Floyd Galvan, Sujit Mandal, Sharma Kolluri• BPA:
� Bill Mittelstadt, Dmitry Kosterev
• EPG: � Manu Parashar
6
PSERC
Oscillation Monitoring System (OMS)
• Goal of Oscillation Monitoring System (OMS)
� Early detection of poorly damped oscillations as they appear
� Trigger warning or control signals
• OMS is made possible by Wide Area PMU Measurements
� Growing numbers of PMUs across the power grid
� Fast algorithms available for online measurements
� Rule based automatic analysis of PMU measurements
� Prototype implementation at TVA
SVC SVC
SVC
LOAD
LOAD
LOAD
LOAD
PDC
P
PP
7
PSERC
OMS Flowchart
Event
Analysis Engine
Damping
Monitor Engine
8
PSERC
TVA Cumberland event
Four minutes of oscillations
Source: Gary Kobet/TVA
9
PSERC
TVA Cumberland Event
• Recent oscillatory event at TVA:� Oscillations at Cumberland plant 9/18/2006� PMU recordings enabled the analysis� Local 1.2 Hz mode changed from +1.5% damping
to –0.2% damping and back to +1.5% damping during the event
� PSS installed at the plant subsequently� PMU based real-time alarm coded by TVA into TVA
PDC as back-up measure – uses standard deviation thresholds – plant operators to reduce MW output when alarm received.
10
PSERC
Oscillation Monitoring System
• Software Engines built into TVA PDC
• Real-time streaming data input to the engines
• Fast detection of poorly damped oscillatory modes: mode frequency, damping and mode shape
• Multiple algorithms integrated by expert system like rules
• Focus on Redundancy and Reliability
11
PSERC
OMS Engines
• Event Analysis Engine
� Automated Prony type analysis of oscillatory ringdown responses
� Five seconds of PMU data analyzed every one
second
• Damping Monitor Engine
� Automated analysis of ambient noise data
� Three minutes of PMU data analyzed every ten
seconds
� Provisional Patent application filed by WSU
12
PSERC
Results from Two Engines
820 840 860 880 900 920 940
5.09
5.1
5.11
5.12
5.13
5.14
5.15
x 105
Time (s)
Bus Voltage Magnitude at Cumberland
Event
Analysis
1.2 Hz at +1.5% damping. Local Mode.
Ambient Noise Analysis
1.2 Hz at +1.8% damping. Local Mode.
Nov. 29th 2007 TVA event
13
PSERC
Mode Shape – Local Mode
-1 -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8 1-1
-0.8
-0.6
-0.4
-0.2
0
0.2
0.4
0.6
0.8
1
Real Part
Imagin
ary
Part
Mode Shape Identified by FDD at 1.224 Hz
TVA_COLN_MAIN
TVA_CUMB_BUS1
TVA_CUMB_BUS2
TVA_HEND_BUS1
TVA_LOWN_161K
TVA_LOWN_500K
TVA_MARS_CUMB
TVA_RIDG_BUS1
TVA_RIDG_BUS2
CumberlandOther PMUs
Cumberland oscillating against rest of system
14
PSERC
Basics of Prony Analysis
uDxCy
uBxAx
+=
+=&
• Assumptions: Linear Time-Invariant System, Distinct Eigenvalues, Step changes
in input, …
• Any output is a linear combination of fundamental modal responses
• Well-suited for Prony type curve fitting methods. Estimate oscillatory frequency,
damping ratio and mode shape.
• Estimates should be consistent:
• Moving time-windows (Linearity of responses)
• Different groupings of outputs (Superposition)
ta
j
ijjdj
t
j
iji
jnjj ectecty−−
∑∑ +−= )cos()( ϕωωζ
15
PSERC
Power System Prony Analysis
• Nonlinear Large Scale System
• In theory, Prony Analysis works well for analyzing “Small-disturbance responses”
• Nonlinearity dominant just after large disturbances
• Switching of lines and cap banks in the middle of analysis windows
• Noise effect on results if disturbance “fades away”
• How to get reliable estimation automatically?
16
PSERC
Rules for Real-time Prony Analysis
Three types of Consistency Crosscheck rules
� Different Curve-fitting Methods (Redundancy)
� Different Signal Groups (Superposition)
� Moving Window Analysis (Linearity of Reponses)
60 80 100 120 140 160 180
534
536
538
540
542
544
546
548
550
Time (s)
MALN_Malin
_Bus_Voltage_VMag
MALN_Malin_Bus_Voltage_VMag vs time
17
PSERC
Event Analysis Engine
Local
PMU
Analysis
Inter-area
Mode
Analysis
18
PSERC
Local PMU Analysis
• Signals from one PMU used at a time
• Parallel implementation of multiple PMU analysis
• Parallel implementation for multiple algorithms
• Check for consistency using rules:
� Crosscheck results from Prony, Matrix Pencil and HTLS
� Crosscheck results among moving time-windows
19
PSERC
Interarea Mode Analysis
• Identify interarea modes and related PMUs from local analysis
• Grouping of signals from relevant multiple PMUs
• Check for consistency using three sets of rules:
� Crosscheck results from Prony, Matrix Pencil and HTLS
� Crosscheck results among moving time-windows
� Crosscheck results from different groupings
• Parallel implementation for different interareamodes
20
PSERC
Event Analysis – Inter-area Example
60 80 100 120 140 160 180
534
536
538
540
542
544
546
548
550
Time (s)
MA
LN
_M
alin
_B
us_V
oltage_V
Mag
MALN_Malin_Bus_Voltage_VMag vs time
• WECC Aug. 4, 2000
• Alberta system separated at 19:56 GMT
• 0.27 Hz oscillation is poorly damped
21
PSERC
Case Study 1 – Local PMU Analysis
60 70 80 90 100 110 120 130
0.1
0.2
0.3
0.4
0.5
Time (s)
Fre
qu
en
cy
(H
z)
Frequency estimates
Prony
MatrixPencil
HTLS
60 70 80 90 100 110 120 130
-0.2
-0.1
0
0.1
0.2
Time (s)
Da
mp
ing
ra
tio
Damping ratio estimates
Prony
MatrixPencil
HTLS
60 70 80 90 100 110 120 130
0.1
0.2
0.3
0.4
0.5
Time (s)
Fre
qu
en
cy
(H
z)
Frequency estimates
Prony
MatrixPencil
HTLS
60 70 80 90 100 110 120 130
-0.2
-0.1
0
0.1
0.2
Time (s)
Da
mp
ing
ra
tio
Damping ratio estimates
Prony
MatrixPencil
HTLS
Grand Coulee Malin
22
PSERC
Inter-Area Oscillation Mode
60 70 80 90 100 110 120 130
1
2
3
4
5
6
7
8
Time (s)
Consistent local estimates vs time
BE50
COLS
GC50
MALN
JDAY
KEEL
VNCT
DVER
Consistent estimate
at 106 sec
Oscillation frequency =
0.286 Hz
Mean damping ratio =
+2.77%
23
PSERC
Event Monitor – Local Mode
• Consistent estimate at +9 sec
• Frequency = 1.1785 Hz. Damping at 0.04%
320 325 330 335 340 345 350 355 360 365
2.945
2.95
2.955
2.96
2.965
2.97
2.975
x 105
Time (s)
Vo
lta
ge
at
PM
U 2
(V
)
PMU 2
335 340 345 350 355 360
1
2
3
4
5
6
7
8
Time (s)
Consistent local estimates vs time
PMU
PMU
PMU
PMU
PMU
PMU
PMU
PMU
24
PSERC
Case Study 3 – Growing Oscillations
• Consistent estimate at 809 sec; Local mode.
• Frequency = 0.6930 Hz. Damping ratio = -0. 12%
720 740 760 780 800 820 840 860 880
5.225
5.23
5.235
5.24
5.245
5.25
5.255
5.26
5.265
5.27
5.275
x 105
Time (s)
Voltage a
t P
MU
1 (
V)
PMU 1
740 750 760 770 780 790 800 810 820
0.5
0.6
0.7
0.8
0.9
Time (s)
Fre
quency (
Hz)
Frequency estimates
Prony
MatrixPencil
HTLS
740 750 760 770 780 790 800 810 820
-0.2
0
0.2
0.4
Time (s)
Dam
pin
g r
atio
Damping ratio estimates
Prony
MatrixPencil
HTLS
25
PSERC
Damping Monitor Engine
• Frequency Domain Decomposition (FDD) algorithm proposed for off-line analysis in other areas
• Extended for real-time PMU analysis by Guoping
• Can detect damping ratio as well as mode shape of poorly damped oscillatory modes using short spans of ambient PMU data
• Mode shape information critical for correctly identifying problematic mode towards control actions
• Excellent results for modes with damping ratio up to +10%
26
PSERC
FDD for Ambient Analysis
• FFT and Power Spectra from multiple signals
• Clean up spectra using Singular Value Decomposition procedure near dominant modes
• Prony type damping analysis after Inverse FFT around dominant modes
• ISCAS 2008 paper
• Simultaneous extraction of mode damping and mode shape from ambient data
27
PSERC
Example of Damping Monitor Engine
300 400 500 600 700 8001000
1100
1200
1300
1400
1500
1600
1700
Time (s)
Activ
e P
ow
er (M
W)
Active Power from Malin to Round Mountain #1
Keeler-Allston Line trips
Ross-Lexington Line trips
28
PSERC
Before Keeler-Allston Trip
150 200 250 300 350 4000
0.5
1
1.5
Time (s)
Fre
quency (H
z) Frequency Estimates
150 200 250 300 350 4000
2
4
6
8
Time (s)Dam
pin
g R
atio
(%
)
Damping Ratio Estimates
Dominant mode is McNary local mode
29
PSERC
Before Keeler-Allston trip
150 200 250 300 350 4000
0.5
1
1.5
Time (s)
Fre
quency (H
z) Frequency Estimates
150 200 250 300 350 4000
2
4
6
8
Time (s)Dam
pin
g R
atio
(%
)
Damping Ratio Estimates
Second Dominant mode is COI interarea mode
30
PSERC
After Allston-Keeler Trip
600 620 640 660 680 700 7200
0.5
1
1.5
Time (s)
Fre
qu
ency (
Hz) Frequency Estimates
600 620 640 660 680 700 7200
2
4
6
8
Time (s)Dam
pin
g R
atio
(%
)
Damping Ratio Estimates
Dominant mode is COI interarea mode
31
PSERC
After Keeler-Allston Trip
600 620 640 660 680 700 7200
0.5
1
1.5
Time (s)
Fre
quency (H
z) Frequency Estimates
600 620 640 660 680 700 7200
2
4
6
8
Time (s)Dam
pin
g R
atio
(%
)
Damping Ratio Estimates
Second Dominant mode is McNary local mode
32
PSERC
Complementary Engines
• Event Analysis Engine
� Three algorithms: Prony, Matrix Pencil and HankelTotal Least Square. Crosscheck Rules.
� Aimed at events resulting in sudden changes in damping
• Damping Monitor Engine
� Ambient noise based. Continuous.
� Frequency Domain Decomposition Algorithm. (Prony). Crosscheck Rules.
� Provides early warning on poorly damped modes
33
PSERC
Example of results for TVA
+3.0%
+1.8%
No data
+1.7%
Damping Monitor
PSS offline (1U)+4.0%Feb. 5, 2008
PSS offline (2U)+1.5%Nov. 29, 2007
PSS installed (1U)+7.2%Dec. 16, 2006
No PSS (2U)+1.7%Sept. 18, 2006
PSS StatusEvent Analysis
Damping history of 1.2 Hz mode
PSS status and effectiveness from the damping level of the local mode.
PSS not effective for two units in service. PSS hardware problem
detected and fixed (June 2008).
34
PSERC
Eastern System Interarea Mode• Interarea mode frequency varies between 0.4 Hz to 0.5 Hz
depending on season.
• Damping Monitor (ambient noise) showed the mode to be
poorly damped around +3% to +5% seasonally.
• 0.47 Hz Interarea mode clearly visible in Event Analysis of
Feb. 26th 2008 Florida blackout event.
• Mode involves many eastern control areas
• Frequency ~ 0.47 Hz, damping ~ +7%, on Feb. 26th 2008
• Likely not related to the blackout. Mode damping at +7% is
comparable to the interarea modes in the western system.
35
PSERC
Possible Control Actions
• Operator Actions
� Reduce Critical Tie-line Transfers
� Switch Damping Enhancement Controls at Specific Thyristor Devices – SVC, HVDC
• Automatic Control Actions
� Switching of Damping Controls: Series Capacitors, Shunt Capacitors, Thyristor Devices
� Generation Rescheduling
36
PSERC
SVC Damping Control
• Stressed Operating Condition
• Near-by tie line active power-flow used as control input
• Sending end => Phase Lag Compensator
• Receiving end => Phase Lead Compensator
SVCSVC
P
37
PSERC
HVDC Damping Control
• Stressed Operating Condition
• Phase Angle Difference used as Control Input
• Phase Lead Compensator for HVDC Modulation
• Effective Improvement in Damping of InterareaMode for Diverse Levels of AC Power Transfer
PAC
PDC
38
PSERC
• Successful implementation of real-time code into TVA PDC
• Advanced signal processing algorithms for oscillation analysis of events and ambient noise
• Automatic detection of poorly damped electromechanical modes and their mode shape
• Operator alerts, Operator alarms, Control actions, …
• Provides early warning on emerging oscillatory problems
• Can validate effectiveness and status of PSS at generators when PMU near generator
OMS Summary
39
PSERC
Future Work
• Testing and tuning at TVA
• Conversion of OMS code to 64 bit architecture
• New dedicated eight processor machine with 32 GB dynamic memory at TVA
• OMS Engines for Eastern Grid?
• Operator Alerts and Alarms? Operator Actions?
• Implementation and testing at BPA and California ISO in collaboration with EPG, Operator Actions?
• Implementation at Entergy