Yang, M.-J., B. J.-D. Jou , S.-C. Wang, J.-S. Hong, P.-L. Lin, J.-H .
Integration of Large Data Sets - University of Wisconsin ......sj S P S P rr s j s t j s j s t j s j...
Transcript of Integration of Large Data Sets - University of Wisconsin ......sj S P S P rr s j s t j s j s t j s j...
![Page 1: Integration of Large Data Sets - University of Wisconsin ......sj S P S P rr s j s t j s j s t j s j s j P D D P E T J T ¦¦ ¦¦ ¦¦ Spatio-Temporal Wind Speed Forecast • The](https://reader036.fdocuments.in/reader036/viewer/2022071101/5fdab73d97c245084a1cc924/html5/thumbnails/1.jpg)
Integration of Large Data Sets for Improved Decision-Making in Bulk
Power Systems: Two Case Studies
PSERC WebinarOctober 20, 2015
Le XieTexas A&M University
![Page 2: Integration of Large Data Sets - University of Wisconsin ......sj S P S P rr s j s t j s j s t j s j s j P D D P E T J T ¦¦ ¦¦ ¦¦ Spatio-Temporal Wind Speed Forecast • The](https://reader036.fdocuments.in/reader036/viewer/2022071101/5fdab73d97c245084a1cc924/html5/thumbnails/2.jpg)
Presentation Outline
• Brief Overview of PSERC T-51
• Spatio-temporal Correlations among Data Sets
• Case Study 1: Renewable Forecast
• Case Study 2: Synchrophasor dimensionality
reduction and early anomaly detection
• Concluding Remarks
2
![Page 3: Integration of Large Data Sets - University of Wisconsin ......sj S P S P rr s j s t j s j s t j s j s j P D D P E T J T ¦¦ ¦¦ ¦¦ Spatio-Temporal Wind Speed Forecast • The](https://reader036.fdocuments.in/reader036/viewer/2022071101/5fdab73d97c245084a1cc924/html5/thumbnails/3.jpg)
T-51 Project Objectives
• Improved decision-making by utilization of large
data sets – “Big Data”
• The integration of emerging large-data sets to
support advanced analytics, to enhance electricity
system management (planning, operations and
control)
• Correlation of data in time and space and assuring
consistent data semantic and syntax
3
![Page 4: Integration of Large Data Sets - University of Wisconsin ......sj S P S P rr s j s t j s j s t j s j s j P D D P E T J T ¦¦ ¦¦ ¦¦ Spatio-Temporal Wind Speed Forecast • The](https://reader036.fdocuments.in/reader036/viewer/2022071101/5fdab73d97c245084a1cc924/html5/thumbnails/4.jpg)
Big Data in Bulk Power Systems:
Opportunities and Challenges
4
[4] M. Kezunovic, L. Xie, and S. Grijalva, “The role of big data in improving power system operation and protection,” Bulk
Power System Dynamics and Control - IX Optimization, Security and Control of the Emerging Power Grid (IREP), 2013
IREP Symposium.
![Page 5: Integration of Large Data Sets - University of Wisconsin ......sj S P S P rr s j s t j s j s t j s j s j P D D P E T J T ¦¦ ¦¦ ¦¦ Spatio-Temporal Wind Speed Forecast • The](https://reader036.fdocuments.in/reader036/viewer/2022071101/5fdab73d97c245084a1cc924/html5/thumbnails/5.jpg)
T-51 Project Summary
• Project duration: 2013-2015, final report available at
PSERC
• Use of Big Data for Outage and Asset Management
(Kezunovic, lead PI)
• Distributed Database for Future Grid (Grijalva and
Chau)
• Spatio-Temporal Analytics for Renewables (Xie)
• Wind power prediction and its economic benefits
• Solar power prediction and quantified economic benefits
5
![Page 6: Integration of Large Data Sets - University of Wisconsin ......sj S P S P rr s j s t j s j s t j s j s j P D D P E T J T ¦¦ ¦¦ ¦¦ Spatio-Temporal Wind Speed Forecast • The](https://reader036.fdocuments.in/reader036/viewer/2022071101/5fdab73d97c245084a1cc924/html5/thumbnails/6.jpg)
Growth of Renewable Generation (and Data)
6
Renewable Growth in US
In 2014, renewable energy sources account for 16.28% of total installed U.S.
operating generating capacity.
Solar, wind, biomass, geothermal, and hydropower provided 55.7% of new
installed U.S. electrical generating capacity during the first half of 2014 (1,965
MW of the 3,529 MW total installed).
http://www.renewableenergyworld.com
http://www.eia.gov/
http://www.triplepundit.com/
![Page 7: Integration of Large Data Sets - University of Wisconsin ......sj S P S P rr s j s t j s j s t j s j s j P D D P E T J T ¦¦ ¦¦ ¦¦ Spatio-Temporal Wind Speed Forecast • The](https://reader036.fdocuments.in/reader036/viewer/2022071101/5fdab73d97c245084a1cc924/html5/thumbnails/7.jpg)
7
Growth of Synchrophasors (PMU)
Reported by NASPI*
• By March 2012, 500
networked PMUs installed.
• >1700 PMUs installed by
2015.
*NASPI: North American SynchroPhasor Initiative.
North America
• More than 2000 PMU
[Beijing Sifang, 2013].
China
• http://www.eia.gov/todayinenergy/detail.cfm?id=5630
• Beijing Sifang Company, “Power grid dynamic monitoring and disturbance identification,”
in North American SynchroPhasor Initiative WorkGroup Meeting, Feb. 2013, 2013.
PMU map in North America as of Oct. 2014.
![Page 8: Integration of Large Data Sets - University of Wisconsin ......sj S P S P rr s j s t j s j s t j s j s j P D D P E T J T ¦¦ ¦¦ ¦¦ Spatio-Temporal Wind Speed Forecast • The](https://reader036.fdocuments.in/reader036/viewer/2022071101/5fdab73d97c245084a1cc924/html5/thumbnails/8.jpg)
Spatio-temporal Correlations at Multi Scales
8
Renewable Data
Renewable Data
Renewable Data
Renewable Data
Renewable Data
Renewable Data
Improved Forecast & Dispatch
(5 minutes to hours ahead)
Renewable Data
![Page 9: Integration of Large Data Sets - University of Wisconsin ......sj S P S P rr s j s t j s j s t j s j s j P D D P E T J T ¦¦ ¦¦ ¦¦ Spatio-Temporal Wind Speed Forecast • The](https://reader036.fdocuments.in/reader036/viewer/2022071101/5fdab73d97c245084a1cc924/html5/thumbnails/9.jpg)
Spatio-temporal Correlations at Multi-scale
9
Synchrophasor
Synchrophasor
SynchrophasorSynchrophasor
Synchrophasor
Synchrophasor
SynchrophasorEarly Anomaly Detection & Mitigation
(milliseconds to seconds)
![Page 10: Integration of Large Data Sets - University of Wisconsin ......sj S P S P rr s j s t j s j s t j s j s j P D D P E T J T ¦¦ ¦¦ ¦¦ Spatio-Temporal Wind Speed Forecast • The](https://reader036.fdocuments.in/reader036/viewer/2022071101/5fdab73d97c245084a1cc924/html5/thumbnails/10.jpg)
Wind Variability and Spatial Correlation
10• Source: http://www.caiso.com/1c9b/1c9bd3a394f0.pdf
![Page 11: Integration of Large Data Sets - University of Wisconsin ......sj S P S P rr s j s t j s j s t j s j s j P D D P E T J T ¦¦ ¦¦ ¦¦ Spatio-Temporal Wind Speed Forecast • The](https://reader036.fdocuments.in/reader036/viewer/2022071101/5fdab73d97c245084a1cc924/html5/thumbnails/11.jpg)
11
Spatio-Temporal Wind Forecast
91 mile 24 mile
West East
Communication and Information exchange:
• Produce good wind generation forecast
• Reduce the system-wide dispatch cost
• Lower the ancillary services cost
• Incorporate more wind generation
• Related work by M. He and L. Yang and J. Zhang and V. Vittal, "A Spatio-temporal Analysis Approach for Short-
term Wind Generation Forecast,” IEEE Transactions on Power Systems, 2014.
![Page 12: Integration of Large Data Sets - University of Wisconsin ......sj S P S P rr s j s t j s j s t j s j s j P D D P E T J T ¦¦ ¦¦ ¦¦ Spatio-Temporal Wind Speed Forecast • The](https://reader036.fdocuments.in/reader036/viewer/2022071101/5fdab73d97c245084a1cc924/html5/thumbnails/12.jpg)
• Persistence (PSS) Model
• Assume the future wind speed is the same as the current
one:
• Autoregressive (AR) Model:
• Estimate μrs,t+k as a linear combination of the previous
wind speed at the same location
where μrs,t+k is the residue term of center parameter of wind speed.
12
Existing Methods
, ,ˆ
s t k s ty y
, 0 ,
0
pr r
s t k i s t i
i
![Page 13: Integration of Large Data Sets - University of Wisconsin ......sj S P S P rr s j s t j s j s t j s j s j P D D P E T J T ¦¦ ¦¦ ¦¦ Spatio-Temporal Wind Speed Forecast • The](https://reader036.fdocuments.in/reader036/viewer/2022071101/5fdab73d97c245084a1cc924/html5/thumbnails/13.jpg)
, 0 , ,
1 0
, , , ,
1 0 1 0
cos sin
S Pr r
s t k s j s t j
s j
S P S Pr r
s j s t j s j s t j
s j s j
Spatio-Temporal Wind Speed Forecast
• The scale parameter is modeled as
• Where b0, b1>0 and vs1,t is the volatility value:
• The residual term modeled as a linear function of current
and past (up to time lag h) wind speed and trigonometric
functions of wind direction.
, 0 1 ,s t k s tb b v
1
2
, , , 1
1 0
1
2
Sr r
s t s t i s t i
s i
vS
13
![Page 14: Integration of Large Data Sets - University of Wisconsin ......sj S P S P rr s j s t j s j s t j s j s j P D D P E T J T ¦¦ ¦¦ ¦¦ Spatio-Temporal Wind Speed Forecast • The](https://reader036.fdocuments.in/reader036/viewer/2022071101/5fdab73d97c245084a1cc924/html5/thumbnails/14.jpg)
14
Spatio-Temporal Wind Forecast
West Texas Case Study [1]
[1] L. Xie, Y. Gu, X. Zhu, and M. G. Genton, “Short-Term Spatio-Temporal Wind Power Forecast in
Robust Look-ahead Power System Dispatch,” IEEE Tran. Smart Grid, 2014.
![Page 15: Integration of Large Data Sets - University of Wisconsin ......sj S P S P rr s j s t j s j s t j s j s j P D D P E T J T ¦¦ ¦¦ ¦¦ Spatio-Temporal Wind Speed Forecast • The](https://reader036.fdocuments.in/reader036/viewer/2022071101/5fdab73d97c245084a1cc924/html5/thumbnails/15.jpg)
15
Forecast Model Performance
MAE VALUES OF THE 10-MINUTE-AHEAD, 20-MINUTE-AHEAD AND UP TO 1-HOUR-AHEAD
FORECASTS ON 11 DAYS’ IN 2010 FROM THE PSS, AR, TDD AND TDDGW MODELS AT THE FOUR
LOCATIONS (SMALLEST IN BOLD)
![Page 16: Integration of Large Data Sets - University of Wisconsin ......sj S P S P rr s j s t j s j s t j s j s j P D D P E T J T ¦¦ ¦¦ ¦¦ Spatio-Temporal Wind Speed Forecast • The](https://reader036.fdocuments.in/reader036/viewer/2022071101/5fdab73d97c245084a1cc924/html5/thumbnails/16.jpg)
16
Wind Generation Forecast Distribution
One Hour-ahead Wind Generation Forecast Uncertainty
Hour ahead Wind generation forecast uncertainties of Jayton (JAYT), Texas
under various days
![Page 17: Integration of Large Data Sets - University of Wisconsin ......sj S P S P rr s j s t j s j s t j s j s j P D D P E T J T ¦¦ ¦¦ ¦¦ Spatio-Temporal Wind Speed Forecast • The](https://reader036.fdocuments.in/reader036/viewer/2022071101/5fdab73d97c245084a1cc924/html5/thumbnails/17.jpg)
17
Total Operating Cost
![Page 18: Integration of Large Data Sets - University of Wisconsin ......sj S P S P rr s j s t j s j s t j s j s j P D D P E T J T ¦¦ ¦¦ ¦¦ Spatio-Temporal Wind Speed Forecast • The](https://reader036.fdocuments.in/reader036/viewer/2022071101/5fdab73d97c245084a1cc924/html5/thumbnails/18.jpg)
Spatio-temporal Solar Forecast [2]
18
[2] C. Yang, A. Thatte, and L. Xie, “Multi time-Scale Data-Driven Spatio-Temporal Forecast of Photovoltaic Generation,” IEEE
Tran. Sustainable Energy, 2015.
![Page 19: Integration of Large Data Sets - University of Wisconsin ......sj S P S P rr s j s t j s j s t j s j s j P D D P E T J T ¦¦ ¦¦ ¦¦ Spatio-Temporal Wind Speed Forecast • The](https://reader036.fdocuments.in/reader036/viewer/2022071101/5fdab73d97c245084a1cc924/html5/thumbnails/19.jpg)
ARX Model for Solar Irradiance Forecast [2]
We compared our model (ST) to persistence (PSS),
auto regression (AR), and back-propagation neural
network (BPNN) forecast models
19
Local
Neighbor Sites
[2] C. Yang, A. Thatte, and L. Xie, “Multi time-Scale Data-Driven Spatio-Temporal Forecast of Photovoltaic Generation,” IEEE
Tran. Sustainable Energy, 2015.
![Page 20: Integration of Large Data Sets - University of Wisconsin ......sj S P S P rr s j s t j s j s t j s j s j P D D P E T J T ¦¦ ¦¦ ¦¦ Spatio-Temporal Wind Speed Forecast • The](https://reader036.fdocuments.in/reader036/viewer/2022071101/5fdab73d97c245084a1cc924/html5/thumbnails/20.jpg)
Results [2]
20
Performance for 1 hour ahead
Performance for 2 hour aheadShorter time-scale:
Spatio-temporal is
worse than PSS
![Page 21: Integration of Large Data Sets - University of Wisconsin ......sj S P S P rr s j s t j s j s t j s j s j P D D P E T J T ¦¦ ¦¦ ¦¦ Spatio-Temporal Wind Speed Forecast • The](https://reader036.fdocuments.in/reader036/viewer/2022071101/5fdab73d97c245084a1cc924/html5/thumbnails/21.jpg)
Part I: Summary
• Spatio-temporal correlation among renewable
generation sites (wind and photovoltaic) could
be leveraged for improved near-term forecast
• The economic benefit from spatio-temporal
forecast vary at different time scales
• Possible extensions:
• Large ramp forecast
• Distributed PV forecast
21
![Page 22: Integration of Large Data Sets - University of Wisconsin ......sj S P S P rr s j s t j s j s t j s j s j P D D P E T J T ¦¦ ¦¦ ¦¦ Spatio-Temporal Wind Speed Forecast • The](https://reader036.fdocuments.in/reader036/viewer/2022071101/5fdab73d97c245084a1cc924/html5/thumbnails/22.jpg)
22
Barriers to Using PMU
for Real-time Operations
Large sets of PMU data
Efficient
real-time
analysis
Missing
data
prediction
Data
storage
Dimensionality Reduction
Related Work:
[5] M. Wang, J.H. Chow, P. Gao, X.T. Jiang, Y. Xia, S.G. Ghiocel, B. Fardanesh, G. Stefopolous, Y. Kokai, N. Saito, M.
Razanousky, "A Low-Rank Matrix Approach for the Analysis of Large Amounts of Power System Synchrophasor Data,"
in System Sciences (HICSS), 2015.
[6] N. Dahal, R. King, and V. Madani, IEEE, “Online dimension reduction of synchrophasor data,” in Proc. IEEE PES
Transmission and Distribution Conf. Expo. (T&D), 2012.
![Page 23: Integration of Large Data Sets - University of Wisconsin ......sj S P S P rr s j s t j s j s t j s j s j P D D P E T J T ¦¦ ¦¦ ¦¦ Spatio-Temporal Wind Speed Forecast • The](https://reader036.fdocuments.in/reader036/viewer/2022071101/5fdab73d97c245084a1cc924/html5/thumbnails/23.jpg)
0 100 200 300 400 500 600 700 800 900 1000 11001.01
1.015
1.02
1.025
1.03
1.035Voltage Magnitude Profile for ERCOT Data
Time (Sec)
Voltage M
agnitude (
p.u
.)
V1
V2
V3
V4
V5
V6
V7
Raw PMU Data from Texas
Bus Frequency Profile Voltage Magnitude Profile
0 100 200 300 400 500 600 700 800 900 1000 11000.995
0.996
0.997
0.998
0.999
1
1.001Bus Frequency Profile for ERCOT Data
Time (Sec)
Bus F
requency (
p.u
.)
1
2
3
4
5
6
7
No system topology, no system model.
Total number of PMUs: 7. 23
![Page 24: Integration of Large Data Sets - University of Wisconsin ......sj S P S P rr s j s t j s j s t j s j s j P D D P E T J T ¦¦ ¦¦ ¦¦ Spatio-Temporal Wind Speed Forecast • The](https://reader036.fdocuments.in/reader036/viewer/2022071101/5fdab73d97c245084a1cc924/html5/thumbnails/24.jpg)
PMU Data from Eastern Interconnection
24
0 50 100 150 200 250 300 350 400 450 5001
1.05
1.1
1.15
1.2
1.25
1.3
1.35
1.4Voltage Magnitude Profile for PJM Data
Time (Sec)
Voltage M
agnitude (
p.u
.)
V1
V2
V3
V4
V5
V6
V7
V8
0 100 200 300 400 500 600 700 800 900 1000 11000.9994
0.9995
0.9996
0.9997
0.9998
0.9999
1
1.0001
1.0002Bus Frequency Profile for PJM Data
Time (Sec)
Bus F
requency (
p.u
.)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
Bus Frequency Profile Voltage Magnitude Profile
Total number of PMUs: 14 for frequency analysis
8 for voltage magnitude analysis.
![Page 25: Integration of Large Data Sets - University of Wisconsin ......sj S P S P rr s j s t j s j s t j s j s j P D D P E T J T ¦¦ ¦¦ ¦¦ Spatio-Temporal Wind Speed Forecast • The](https://reader036.fdocuments.in/reader036/viewer/2022071101/5fdab73d97c245084a1cc924/html5/thumbnails/25.jpg)
PCA for Texas Data
Cumulative variance for bus frequency and voltage magnitude for Texas data.
1 2 3 4 5 6 799.7
99.75
99.8
99.85
99.9
99.95
100
Number of PCs
Perc
enta
ge (
%)
(a) Cumulative Variance for Bus Frequency in Texas Data
1 2 3 4 5 6 720
30
40
50
60
70
80
90
100
Number of PCs
Perc
enta
ge (
%)
(b) Cumulative Variance for Voltage Magnitude in Texas Data
25
5 3, BY
2 1 7, , V
BY V V V
![Page 26: Integration of Large Data Sets - University of Wisconsin ......sj S P S P rr s j s t j s j s t j s j s j P D D P E T J T ¦¦ ¦¦ ¦¦ Spatio-Temporal Wind Speed Forecast • The](https://reader036.fdocuments.in/reader036/viewer/2022071101/5fdab73d97c245084a1cc924/html5/thumbnails/26.jpg)
26
2 4 6 8 10 12 1499.4
99.5
99.6
99.7
99.8
99.9
100
Number of PCs
Perc
enta
ge (
%)
(a) Cumulative Variance for Bus Frequency in PJM
1 2 3 4 5 6 7 855
60
65
70
75
80
85
90
95
100
Number of PCs
Perc
enta
ge (
%)
(b) Cumulative Variance for Voltage Magnitude in PJM
PCA for Eastern Interconnection
11 6, BY
8 5 2, , V
BY V V V
Cumulative variance for bus frequency and voltage magnitude for PJM data.
![Page 27: Integration of Large Data Sets - University of Wisconsin ......sj S P S P rr s j s t j s j s t j s j s j P D D P E T J T ¦¦ ¦¦ ¦¦ Spatio-Temporal Wind Speed Forecast • The](https://reader036.fdocuments.in/reader036/viewer/2022071101/5fdab73d97c245084a1cc924/html5/thumbnails/27.jpg)
27
Scatter Plot of Bus Frequency
2D Scatter plot for bus frequency. 3D Scatter plot for bus frequency.
Normal
Condition
Abnormal
Condition
Back to Normal
Condition
![Page 28: Integration of Large Data Sets - University of Wisconsin ......sj S P S P rr s j s t j s j s t j s j s j P D D P E T J T ¦¦ ¦¦ ¦¦ Spatio-Temporal Wind Speed Forecast • The](https://reader036.fdocuments.in/reader036/viewer/2022071101/5fdab73d97c245084a1cc924/html5/thumbnails/28.jpg)
28
Scatter Plot of Voltage Magnitude
2D Scatter plot for voltage magnitude. 3D Scatter plot for voltage magnitude.
Normal
Condition
Abnormal
Condition
Back to Normal
Condition
![Page 29: Integration of Large Data Sets - University of Wisconsin ......sj S P S P rr s j s t j s j s t j s j s j P D D P E T J T ¦¦ ¦¦ ¦¦ Spatio-Temporal Wind Speed Forecast • The](https://reader036.fdocuments.in/reader036/viewer/2022071101/5fdab73d97c245084a1cc924/html5/thumbnails/29.jpg)
Observations
• High dimensional PMU raw measurement data
lie in an much lower subspace (even with linear
PCA)
• Scattered plots suggest that
Change of subspace -> Occurrence of events !
• But, what is the way to implement it?
• Is there any theoretical justification?
Data-driven subspace change Indication of physical
events in wide-area power systems
29
![Page 30: Integration of Large Data Sets - University of Wisconsin ......sj S P S P rr s j s t j s j s t j s j s j P D D P E T J T ¦¦ ¦¦ ¦¦ Spatio-Temporal Wind Speed Forecast • The](https://reader036.fdocuments.in/reader036/viewer/2022071101/5fdab73d97c245084a1cc924/html5/thumbnails/30.jpg)
30
Corporate PDCData
Storage
Synchrophasor Data
Dimensionality Reduction
Data
Storage
Early Event Detection
: Phasor measurement unit
PDC: Phasor data concentrator
: Raw measured PMU data
: Preprocessed PMU data
Local PDC Local PDCLocal PDC
Early Event Detection
Pilot PMUs
![Page 31: Integration of Large Data Sets - University of Wisconsin ......sj S P S P rr s j s t j s j s t j s j s j P D D P E T J T ¦¦ ¦¦ ¦¦ Spatio-Temporal Wind Speed Forecast • The](https://reader036.fdocuments.in/reader036/viewer/2022071101/5fdab73d97c245084a1cc924/html5/thumbnails/31.jpg)
Dimensionality Reduction Algorithm [3]
: Measurement matrix 1. PMU Data Collection ],,[ 1 N
Nn yyY
Tin
ii yyy ],,[ 1
N measurements. Each has n samples in the time history.
2. PCA-based Dimensionality Reduction
(1) Eigenvalues and eigenvectors of covariance matrix .
(2) Rearrange eigenvalues in decreasing order to find principal components (PCs).
(3) Select top m out of N PCs based on predefined threshold.
(4) Project original N variables in the m PC-formed new space.
(5) m’ variables are kept and chosen as orthogonal to each other as possible.
(6) Basis matrix .
(7) Predict in terms of
( )Cov Y
YyyY mbbB ],,[: '1
iy BY
.ˆ
'
1
iB
m
j
jb
ij
i vYyvy
.:1 iT
BBTB
i yYYYv
[3] L. Xie, Y. Chen, and P. R. Kumar, “Dimensionality Reduction of Synchrophasor Data for Early Event Detection:
Linearized Analysis,” IEEE Tran. Power Systems, 2014. 31
![Page 32: Integration of Large Data Sets - University of Wisconsin ......sj S P S P rr s j s t j s j s t j s j s j P D D P E T J T ¦¦ ¦¦ ¦¦ Spatio-Temporal Wind Speed Forecast • The](https://reader036.fdocuments.in/reader036/viewer/2022071101/5fdab73d97c245084a1cc924/html5/thumbnails/32.jpg)
PCA for Bus Frequency
Cumulative variance for bus frequency.
0 5 10 15 20 2599.975
99.98
99.985
99.99
99.995
100
Number of PCs
Perc
enta
ge (
%)
Cumulative Variance for Bus Frequency in PSSE
0 0.05 0.1 0.15 0.2 0.25-0.4
-0.3
-0.2
-0.1
0
0.1
0.2
0.3
0.4
0.5
0.6
PC1
2D Loading Plot for Bus Frequency
PC
2
206
102
2 PCs. Basis matrix 206 102,BY
2D Loading plot for bus frequency.
32
![Page 33: Integration of Large Data Sets - University of Wisconsin ......sj S P S P rr s j s t j s j s t j s j s j P D D P E T J T ¦¦ ¦¦ ¦¦ Spatio-Temporal Wind Speed Forecast • The](https://reader036.fdocuments.in/reader036/viewer/2022071101/5fdab73d97c245084a1cc924/html5/thumbnails/33.jpg)
Possible Implementation
33
Adaptive Training
PCA-basedDimensionality Reduction
Robust Online Monitoring
Online Detection
PMU Measurement sfdfffa
Covariance Matrix ga
Reorder va Eigenvalues
Select fa PCs, ggagga
Project jfj in m-D Space
Define Base Matrix ags
Calculate sh
Approximate grffs
Approximation error gfsgf
Event indicator grss
Alert to
System
Operators
0n NY t
YC
BY
iv
N
m m N
ˆi
y t
i
e t
i
t
YES
1t t
NO
Y
Early Event Detection Algorithm
?i
t
NEED
0?t t UT T
NO
YES
Update
![Page 34: Integration of Large Data Sets - University of Wisconsin ......sj S P S P rr s j s t j s j s t j s j s j P D D P E T J T ¦¦ ¦¦ ¦¦ Spatio-Temporal Wind Speed Forecast • The](https://reader036.fdocuments.in/reader036/viewer/2022071101/5fdab73d97c245084a1cc924/html5/thumbnails/34.jpg)
Theorem for Early Event Detection
Using the proposed event indicator , a system event can
be detected within 2-3 samples of PMUs, i.e., within 100 ms,
whenever for some selected non-pilot PMU i, the event
indicator satisfies
i
t
where is a system-dependent threshold and can be
calculated using historical PMU data.
Theoretically justified.
34
![Page 35: Integration of Large Data Sets - University of Wisconsin ......sj S P S P rr s j s t j s j s t j s j s j P D D P E T J T ¦¦ ¦¦ ¦¦ Spatio-Temporal Wind Speed Forecast • The](https://reader036.fdocuments.in/reader036/viewer/2022071101/5fdab73d97c245084a1cc924/html5/thumbnails/35.jpg)
Sketch of the Proof
• Power system DAE model
• Discretization
• Using back substitution, explicitly express output
(measurement) y[k] in terms of initial condition x[1],
control input u[k], noise e[k]
35
![Page 36: Integration of Large Data Sets - University of Wisconsin ......sj S P S P rr s j s t j s j s t j s j s j P D D P E T J T ¦¦ ¦¦ ¦¦ Spatio-Temporal Wind Speed Forecast • The](https://reader036.fdocuments.in/reader036/viewer/2022071101/5fdab73d97c245084a1cc924/html5/thumbnails/36.jpg)
Sketch of the Proof (cont.’d)
• Normal conditions: training errors are small
• U0 and x[1] can be theoretically calculated by TRAINING
data.
• Any changes in control inputs and initial conditions will
lead to large prediction error.
• If system topology changes, and will change,
resulting in a large prediction error.
36
xc uc
![Page 37: Integration of Large Data Sets - University of Wisconsin ......sj S P S P rr s j s t j s j s t j s j s j P D D P E T J T ¦¦ ¦¦ ¦¦ Spatio-Temporal Wind Speed Forecast • The](https://reader036.fdocuments.in/reader036/viewer/2022071101/5fdab73d97c245084a1cc924/html5/thumbnails/37.jpg)
Case Study 1: PSS/E Data
• 23-bus system
• 23 PMUs.
• Outputs of PMUs: ω, V.
• Siemens, “PSS/E 30.2 program operational manual,” 2009.37
![Page 38: Integration of Large Data Sets - University of Wisconsin ......sj S P S P rr s j s t j s j s t j s j s j P D D P E T J T ¦¦ ¦¦ ¦¦ Spatio-Temporal Wind Speed Forecast • The](https://reader036.fdocuments.in/reader036/viewer/2022071101/5fdab73d97c245084a1cc924/html5/thumbnails/38.jpg)
38
Time Sampling
PointsEvent
0-100s 1-3000 Normal Condition
100.03-150s
3001-45000Bus Disconnection
(206)
150.03-250s
4501-7500Voltage set point changes (211)
Oscillation Event
time / sec250 0 100 150
Bus 206
disconnectedTraining
Stage
211
ref 0.1V
![Page 39: Integration of Large Data Sets - University of Wisconsin ......sj S P S P rr s j s t j s j s t j s j s j P D D P E T J T ¦¦ ¦¦ ¦¦ Spatio-Temporal Wind Speed Forecast • The](https://reader036.fdocuments.in/reader036/viewer/2022071101/5fdab73d97c245084a1cc924/html5/thumbnails/39.jpg)
39
Early Event Detection
![Page 40: Integration of Large Data Sets - University of Wisconsin ......sj S P S P rr s j s t j s j s t j s j s j P D D P E T J T ¦¦ ¦¦ ¦¦ Spatio-Temporal Wind Speed Forecast • The](https://reader036.fdocuments.in/reader036/viewer/2022071101/5fdab73d97c245084a1cc924/html5/thumbnails/40.jpg)
• How EARLY is our algorithm?
Our Method: potentially within a few samples (<0.1
seconds)
• Most Oscillation monitoring system (OMS) needs
10 sec to detect the oscillation.
40
Early Event Detection
• V. Venkatasubramanian, “Oscillation monitoring system, ” PSERC Report, 2008.
![Page 41: Integration of Large Data Sets - University of Wisconsin ......sj S P S P rr s j s t j s j s t j s j s j P D D P E T J T ¦¦ ¦¦ ¦¦ Spatio-Temporal Wind Speed Forecast • The](https://reader036.fdocuments.in/reader036/viewer/2022071101/5fdab73d97c245084a1cc924/html5/thumbnails/41.jpg)
Case Study 2: Unit Tripping Event
• No system topology, no system model.
• Total number of PMUs: 7.
• 2 unit tripping events.
• Sampling rate: 30 Hz.
41
![Page 42: Integration of Large Data Sets - University of Wisconsin ......sj S P S P rr s j s t j s j s t j s j s j P D D P E T J T ¦¦ ¦¦ ¦¦ Spatio-Temporal Wind Speed Forecast • The](https://reader036.fdocuments.in/reader036/viewer/2022071101/5fdab73d97c245084a1cc924/html5/thumbnails/42.jpg)
Cumulative variance for bus frequency and voltage magnitude for Texas data.
PCA for Bus Frequency and Voltage Magnitude
42
![Page 43: Integration of Large Data Sets - University of Wisconsin ......sj S P S P rr s j s t j s j s t j s j s j P D D P E T J T ¦¦ ¦¦ ¦¦ Spatio-Temporal Wind Speed Forecast • The](https://reader036.fdocuments.in/reader036/viewer/2022071101/5fdab73d97c245084a1cc924/html5/thumbnails/43.jpg)
43
Early Event Detection
w4 profile. during unit tripping event.4
![Page 44: Integration of Large Data Sets - University of Wisconsin ......sj S P S P rr s j s t j s j s t j s j s j P D D P E T J T ¦¦ ¦¦ ¦¦ Spatio-Temporal Wind Speed Forecast • The](https://reader036.fdocuments.in/reader036/viewer/2022071101/5fdab73d97c245084a1cc924/html5/thumbnails/44.jpg)
• Large-scale PMU data can be reduced to a
space with much lower dimensionality
(surprisingly well).
• Change of dimensionality could be leveraged for
novel early anomaly detection
• Rich physical insights can be obtained from
PMU data
• Possible extensions:
• Event classification, specification and localization
• Online PMU bad data processing
44
Part II: Summary
• Y. Chen, L. Xie, and P. R. Kumar, “Power system event classification via dimensionality reduction of
synchrophasor data,” in Sensor Array and Multichannel Signal Processing Workshop, 2014. SAM 2014
![Page 45: Integration of Large Data Sets - University of Wisconsin ......sj S P S P rr s j s t j s j s t j s j s j P D D P E T J T ¦¦ ¦¦ ¦¦ Spatio-Temporal Wind Speed Forecast • The](https://reader036.fdocuments.in/reader036/viewer/2022071101/5fdab73d97c245084a1cc924/html5/thumbnails/45.jpg)
45
Concluding Remarks
• We investigated the integration of large data
sets for improved grid operations:
– Spatio-temporal analytics for improved renewable
forecast & market operations
– Dimensionality reduction of synchrophasor data for
improved monitoring and anomaly detection
• Many open questions
– Streaming data quality
– Data analytics integration with EMS,DMS, MMS
– How to teach “data sciences” for power systems? A
first attempt in Fall 15 at TAMU
![Page 46: Integration of Large Data Sets - University of Wisconsin ......sj S P S P rr s j s t j s j s t j s j s j P D D P E T J T ¦¦ ¦¦ ¦¦ Spatio-Temporal Wind Speed Forecast • The](https://reader036.fdocuments.in/reader036/viewer/2022071101/5fdab73d97c245084a1cc924/html5/thumbnails/46.jpg)
Acknowledgements
• Support from PSERC and BPA
• T-51 Team (Profs. M. Kezunovic, S. Grijalva, P. Chau)
and Industry Advisors
• Collaborators:
• Prof. Marc G. Genton (KAUST, Spatio-temporal Statistics)
• Prof. P. R. Kumar (Texas A&M, System Theory)
• Students:
• Yingzhong Gary Gu (GE)
• Yang Chen (PJM)
• Anupam Thatte (MISO)
• Chen Nathan Yang (NYISO)
• Meng Wu
46
![Page 47: Integration of Large Data Sets - University of Wisconsin ......sj S P S P rr s j s t j s j s t j s j s j P D D P E T J T ¦¦ ¦¦ ¦¦ Spatio-Temporal Wind Speed Forecast • The](https://reader036.fdocuments.in/reader036/viewer/2022071101/5fdab73d97c245084a1cc924/html5/thumbnails/47.jpg)
47
Key References
[1] L. Xie, Y. Gu, X. Zhu, and M. G. Genton, “Short-Term Spatio-Temporal Wind Power Forecast in Robust
Look-ahead Power System Dispatch,” IEEE Tran. Smart Grid, 2014.
[2] C. Yang, A. Thatte, and L. Xie, “Multi time-Scale Data-Driven Spatio-Temporal Forecast of Photovoltaic
Generation,” IEEE Tran. Sustainable Energy, 2015.
[3] L. Xie, Y. Chen, and P. R. Kumar, “Dimensionality Reduction of Synchrophasor Data for Early Event
Detection: Linearized Analysis,” IEEE Tran. Power Systems, 2014.
[4] M. Kezunovic, L. Xie, and S. Grijalva, “The role of big data in improving power system operation and
protection,” Bulk Power System Dynamics and Control - IX Optimization, Security and Control of the
Emerging Power Grid (IREP), 2013 IREP Symposium.
[5] M. Wang, J.H. Chow, P. Gao, X.T. Jiang, Y. Xia, S.G. Ghiocel, B. Fardanesh, G. Stefopolous, Y.
Kokai, N. Saito, M. Razanousky, "A Low-Rank Matrix Approach for the Analysis of Large Amounts of
Power System Synchrophasor Data," in 2015 48th Hawaii International Conference on System
Sciences.
[6] N. Dahal, R. King, and V. Madani, IEEE, “Online dimension reduction of synchrophasor data,” in Proc.
IEEE PES Transmission and Distribution Conf. Expo. (T&D), 2012, pp. 1–7.
[7] M. He and L. Yang and J. Zhang and V. Vittal, "A Spatio-temporal Analysis Approach for Short-term
Wind Generation Forecast," IEEE Tran. on Power Systems, 2014.
![Page 48: Integration of Large Data Sets - University of Wisconsin ......sj S P S P rr s j s t j s j s t j s j s j P D D P E T J T ¦¦ ¦¦ ¦¦ Spatio-Temporal Wind Speed Forecast • The](https://reader036.fdocuments.in/reader036/viewer/2022071101/5fdab73d97c245084a1cc924/html5/thumbnails/48.jpg)
Thank You!
Le Xie
Associate Professor
Department of Electrical and Computer
Engineering
Texas A&M University
www.ece.tamu.edu/~lxie
48