1. |
LSE for Synchrophasor Data Quality –
Implementation and Performance at
BPA
Tony Faris, Bonneville Power Administration
Lin Zhang, Electric Power Group
March 23, 2016

2. |
 WECC Synchrophasor Data Validation and Conditioning Application (SDVCA)
> Project Overview
> Modeless Validation and Conditioning
> Model-based Conditioning - LSE
 EPG Enhanced Linear State Estimator (eLSE)
> Enhancements
> eLSE Model Builder
> Historical Data Testing
 BPA Field Testing and Results
> BPA Testing Environment
> Live Data Testing with ICCP Integration
 Lesson Learned and Future Work
Outline
1© Electric Power Group. 2016. All rights reserved. 03/23/2016

3. |
 Objective is to develop a validation and conditioning application of PMU data for WECC
utilities
 Specified to include validation & conditioning using:
> Modeless algorithms
> Model Based Linear State Estimator (LSE)
 Modeless approach uses the algorithms developed under a DOE sponsored project
> EPG was the contractor for this project
 Model based approach uses the LSE
> Enhanced VT/Dominion LSE code
 Test and demonstration at BPA
> Test site arranged by WECC
> Historical data testing using archived data
> Real-time data testing using a live data feed
SDVCAOverview
2© Electric Power Group. 2016. All rights reserved. 03/23/2016

4. |
Modeless&LSE-BasedDataValidationandConditioning
3© Electric Power Group. 2016. All rights reserved.
Model-Less/Algorithmic
Validation & Conditioning
LSE-Based
Conditioning
Output Selection
Input:
Raw C37.118
Model-less
conditioned
C37.118
LSE-conditioned
C37.118
Output:
Conditioned C37.118
Data Quality
flags
Model-less
conditioned
C37.118
Data Quality
Reports
Algorithms = Detect
and Flag Bad Data
Linear State Estimator = Replace Bad Data
with Validated Model Based Values
03/23/2016

5. |
 Modeless Error Detection
> Communication problems (format, CRC, etc. errors)
> Problems in measurement (37.118 flags)
> Timing errors and anomalies
> Severe measurement anomalies (out of range values)
> Measurement mismatch (topology comparisons)
 Modeless Conditioning
> Flags bad or suspect values
> Replace with user set value (NaN, last good, set number)
ModelessValidationandConditioning
4© Electric Power Group. 2016. All rights reserved. 03/23/2016

6. |
ModelBasedConditioning–LSE
LSE Application
Utility Network
Model (CIM)
C37.118 PMU data
• Estimated
Synchrophasor Data
• Virtual PMU’s with
Estimated Values
• List of Measurement
Anomalies
Topology
Information
 Network Model (CIM format)
 Converted into LSE format model
 PMU Data
 Real-time or recorded
 Topology Info
 From EMS or recorded
5
© Electric Power Group. 2016. All rights reserved.
03/23/2016

7. |
 EPG started with this open source code and developed a production grade eLSE
that incorporates enhancements to operate on complex systems such as the
WECC/BPA system
 Seven major enhancements
> Bad data detection and identification module
> Series Capacitor
> Shunt capacitor/reactor
> Split bus
> Naming convention
> Bypass breaker modeled in line
> Breaker status interface to accept Inter-Control Center Communications
Protocol (ICCP)
EPG’seLSEandMajorEnhancements
6© Electric Power Group. 2016. All rights reserved. 03/23/2016

8. |
Automatic CIM parsing engine
>Parse the CIM model and convert it to LSE model
Mapping File Creation
>Mapping PMU signals to LSE model
Signal mapping engine
>Read the mapping file and automatically map PMU signal to the LSE model
GUI of network model builder
>Edit network model, eg add or remove lines, breakers
>Update models
eLSENetworkModelBuilder–FourMajorComponents
© Electric Power Group 2016. All rights reserved
703/23/2016

9. |
 Validated for BPA’s entire 500 kV and
portion of 230 kV system
 System reduced to PMU visible area
> 37 Substations with PMU installed
> 220 phasor measurements
> 65 observable substations
 Run properly at 60 frames per second
 Testing with historical and live PMU data
FieldTestingonBPASystem
65 Observable Substations with PMUs at 37 Substations
Elements Number
Substations 65
Lines 96
Line Segments 126
Transformers 129
Nodes 3091
Breakers 849
Switches 2357
Series Capacitors 18
Shunt Capacitors 112
Observable buses 78
8© Electric Power Group. 2016. All rights reserved. 03/23/2016

10. |
 Chief Joseph Brake event
HistoricalEventTestingResults
525
530
535
540
545
550
1
452
903
1354
1805
2256
2707
3158
3609
4060
4511
4962
5413
5864
6315
6766
7217
7668
8119
8570
9021
9472
9923
10374
10825
11276
11727
12178
12629
13080
13531
13982
14433
14884
15335
15786
16237
16688
17139
17590
18041
18492
18943
19394
19845
20296
20747
21198
21649
22100
22551
23002
23453
23904
24355
24806
25257
25708
26159
26610
27061
27512
27963
28414
28865
29316
29767
30218
30669
31120
31571
32022
32473
32924
33375
33826
Chief Joseph 500 kV East Bus Voltage Magnitude
LSE Estimated Raw PMU measurement
2kV
9© Electric Power Group. 2016. All rights reserved. 03/23/2016

11. |
BPATestingEnvironment
10© Electric Power Group. 2016. All rights reserved. 03/23/2016
PMU
PMU Lab PDC
PMU
EPG
SDVCA
Raw PMU
Data Archive
LSE Data
Archive
Long Term
Comparison
• SDVCA replaces measured values with state estimates
• Estimates stored in temporary LSE Data Archive
• Estimate compared to raw signal for reasonability
Breaker Status
via ICCP

12. |
LiveDataTestingResult–17hours
WithReal-TimeICCPUpdate
White: SDVCA Red: Phase A Blue: Phase B Green: Phase C
11© Electric Power Group. 2016. All rights reserved. 03/23/2016

13. |
RecentLiveDataTestingResult–24hours
WithReal-TimeICCPUpdate
White: SDVCA
Red: Raw
12© Electric Power Group. 2016. All rights reserved. 03/23/2016

14. |
RecentLiveDataTestingResult–another24hours
WithReal-TimeICCPUpdate
White: SDVCA
Red: Raw
13© Electric Power Group. 2016. All rights reserved. 03/23/2016

15. |
 Standard CIM Model
> “Common” information model, NOT common, customization required
 Network parameters
> Critical for LSE estimated results
 Good redundancy of measurements
> Help detect bad data and give better estimated results
 PMU digitals for breaker status, instead of using ICCP
> No time skew issue if using PMU digitals
LessonsLearned
© Electric Power Group 2016. All rights reserved 1403/23/2016

16. |
 Continue long term testing
 Automate comparison of estimate vs. raw data
> Flag large differences, trace to issues in field
 Pursue issues with network model parameters
> Update models as necessary
 Test applications using conditioned data
 Investigate use of LSE in operational environment
FutureWork
© Electric Power Group 2016. All rights reserved 1503/23/2016

17. |
ThankYou - Questions?
16
201 S. Lake Ave., Suite 400
Pasadena, CA 91101
626-685-2015
www.electricpowergroup.com
Lin Zhang
zhang@electricpowergroup.com
Tony Faris
ajfaris@bpa.gov
03/23/2016
5411 NE Highway 99
Vancouver, WA 98663
360-418-2005
www.bpa.gov