Upload
others
View
11
Download
0
Embed Size (px)
Citation preview
WECC Composite Load Model (CMPLDW)
Benchmarking Summary WECC Load Modeling Task Force
8/9/2017
Contents Purpose .......................................................................................................................................................... 2
Background ................................................................................................................................................... 2
Acknowledgements ....................................................................................................................................... 2
Test Systems ................................................................................................................................................. 3
Early WECC Benchmarking ..................................................................................................................... 3
Main WECC Benchmarking ..................................................................................................................... 3
EPRI Benchmarking ................................................................................................................................. 4
Benchmark Tests Performed ......................................................................................................................... 4
Early WECC Benchmarking ..................................................................................................................... 4
Main WECC Benchmarking ..................................................................................................................... 5
EPRI Benchmarking ................................................................................................................................. 5
Model Parameters Compared ........................................................................................................................ 6
Early WECC Benchmarking ..................................................................................................................... 6
Main WECC Benchmarking ..................................................................................................................... 6
EPRI Benchmarking ................................................................................................................................. 7
Software Programs and Models Tested ........................................................................................................ 7
Early WECC Benchmarking ..................................................................................................................... 7
Main Benchmarking .................................................................................................................................. 7
EPRI Benchmarking ................................................................................................................................. 7
Summary of Results ...................................................................................................................................... 8
Early WECC Benchmarking ..................................................................................................................... 8
Main WECC Benchmarking ..................................................................................................................... 9
EPRI Benchmarking ............................................................................................................................... 10
Conclusions and Next Steps ........................................................................................................................ 11
Detailed Results .......................................................................................................................................... 11
Early WECC Benchmarking ................................................................................................................... 11
Main WECC Benchmarking ................................................................................................................... 13
EPRI Benchmarking ............................................................................................................................... 21
Purpose
The purpose of this report is to document the various benchmarking efforts between software platforms
for the WECC Composite Load Model (CMPLDW). The benchmarking efforts improve the
implementation of CMPLDW and also ensure more consistent implementation among the different
software platforms. This will lead to more consistent results with the model and ultimately lead to wider
adoption of the Composite Load Model as the industry standard for load modeling.
Background
WECC has undertaken systematic efforts to improve dynamic load modeling over the past two decades.
This effort led to the approval of CMPLDW phase 1 in 2013 for planning and operational studies.
Although the CMPLDW has been used successfully in thousands of simulations for several years, some
differences in implementation for different software platforms have become apparent. A specification
document for the CMPLDW model was developed and approved in 2015 to help foster more consistent
implementation, but the specification was not exhaustive and left many implementation details
unspecified. Therefore, differences in software implementation and corresponding model performance
remained between software platforms.
This document covers three different benchmarking efforts from the recent past
• 2013-2014: WECC LMTF benchmarks GE PSLF, Siemens PTI PSSE, and Power World
Simulator (“Early WECC benchmarking”)
• 2015-17: WECC LMTF benchmarks 4 major software platforms (“Main WECC benchmarking”)
• 2016: NERC LMTF with EPRI benchmarks 4 major software platforms (“EPRI benchmarking”)
The Main WECC benchmarking and EPRI benchmarking included the following 4 major software
platforms
• GE PSLF
• Power World Simulator
• Siemens PTI PSSE
• Power Tech TSAT
Acknowledgements
The primary source material which this report draws from the following:
• The results of the Early WECC benchmarking effort are described in detail in the report
“Composite Load Model Benchmarking PowerWorld Simulator vs. GE PLSF”, October 16,
2014, prepared by Ran Xu, Tom Overbye, and Komal Shetye. Also PSSE vs. PSLF
benchmarking plots were made available in this report directly from Song Wang at Pacificorp,
not as part of a separate report or presentation.
• The Main WECC benchmarking has been documented in power point presentations to WECC
MVWG. The Power World vs. PSLF portion of the Main WECC benchmarking is documented in
a power point presentation “Comparison of CMPLDW on Test Cases”, March 2016, Jamie
Weber. Also The PSSE vs. PSLF portion of the Main WECC benchmarking is described in detail
in the report “Benchmarking Composite Load Model Between PSSE and PSLF”, October 2015,
prepared by Song Wang. Also Power Tech vs. PSLF benchmarking plots were made available by
Fred Howell at Power Tech for use in presentations to WECC MVWG as well as for this report.
• The EPRI benchmarking is documented in a memo from EPRI to NERC LMTF with a subject
heading “Reinforcements to the composite load model implementation across software
platforms”. Also, there is a power point presentation from EPRI, “Benchmarking WECC
Composite Load Model” that was presented at the March 2017 WECC LMTF meeting, as well as
an updated version of that presentation given at the April 2017 NERC LMTF meeting. The EPRI
benchmarking and associated documentation was conducted by Parag Mitra and Anish Gaikwad
at EPRI.
Test Systems
Early WECC Benchmarking
Fig. 1:Early WECC benchmarking test system
Main WECC Benchmarking
Fig. 2: Main WECC benchmarking test system
EPRI Benchmarking
Fig. 3: EPRI benchmarking test system
Benchmark Tests Performed
Early WECC Benchmarking
Two different load compositions were tested simultaneously on a single system (see figure 1):
• Load 1 (bus 102)
o 18% Motor A
o 15% Motor B
o 6% Motor C
o 14% Motor D with stalling enabled, Vstall=0.54 pu, Tstall=0.03 s
o 15% power electronic
• Load 2 (bus 103)
o 20% Motor A
o 25% Motor B
o 30% Motor C
o 0% Motor D
o 20% power electronic
For each scenario, the following disturbances were applied:
• 3 phase bolted fault on bus 101 cleared in 5 cycles.
• Opening a line without a fault (open branch from bus 101 to 102)
Main WECC Benchmarking
The following load composition scenarios were tested in sequence:
• All components are present, phase 1 and phase 2
• 100% Motor A – three-phase compressor motor
• 100% Motor B – fan motor (has to be modeled in place of Motor A in PSLF, but the data is for
Motor B)
• 100% Motor D – 1-phase compressor motor (has to be modeled in place of Motor A in PSLF, but
the data is for Motor D), phase 1and 2
• 100% power electronic load
Phase 1 has stalling disabled, phase 2 has stalling enabled with Vstall=0.6 pu and Tstall=0.02 s.
For each scenario, the following disturbances were applied:
• A fault with 6 cycle clearing
• Voltage sags to 90 and 75% (played in)
• Voltage ramp (played in)
• Under-frequency event (played in)
• Voltage oscillations (played in)
EPRI Benchmarking
The following load composition scenarios were tested:
• 100% motor A
• 100% motor B
• 100% motor D, stalling enabled and disabled
• 100% power electronic
• 100% static
For each scenario, the following disturbances were applied:
• Step changes in voltage at the load bus were created by controlling the generator’s exciter to
simulate disturbances
Model Parameters Compared
The following model parameters were examined during the benchmarking tests to validate the different
software implementations:
Early WECC Benchmarking
Table 1: Early WECC benchmarking parameters compared
Main WECC Benchmarking
Table 2:Main WECC benchmarking parameters compared
EPRI Benchmarking Table 3: EPRI benchmarking comparisons
Software Programs and Models Tested
The goal of this section is to identify exact software versions and model names used in the benchmarking.
Early WECC Benchmarking
• Power World Simulator versions 17 and 18
• PSSE 32.xx
• PSLF 18.xx
Main Benchmarking
• TSAT prior to version 17. Version 17 resolves all issues found
• PSSE 33.9 (play-in added) and 33.10 (CMPLDW rewrite)
• PSLF 19.xx
• PowerWorld 19
EPRI Benchmarking
• PSLF 19.02
• PSLF 21.xx
• PSS/E 33.10
• PowerWorld 19
• DSATools (TSAT) V16.1.27
Summary of Results
These benchmarking efforts have been a multi-year process involving software vendors, utility engineers,
and industry experts. As a result of these three benchmarking efforts, many improvements have been
made to all four software platforms tested. Robust performance can now be achieved with all four
programs. All significant differences between programs that have been identified over the course of these
benchmarking efforts have been resolved. No significant differences between software implementations
remain. The latest versions of the software available have resolved all of these issues. There are minor
differences between software implementations that will continue to be improved upon moving forward.
Early WECC Benchmarking
In 2013, WECC initiated the first benchmarking for composite load model between three software
platforms, General Electric PSLF, PowerWorld and Siemens PTI PSSE. The benchmark tests consisted of
two contingencies studied on a simplified three bus system. The two programs provided similar results in
principle but differences were identified for frequency and voltage response as well as for an under
voltage relay model.
• Difference in frequency response was found to be caused by difference in default value of fault
impedance assumed for a “solid” 3ph bolted fault. A simulated bolted fault cannot have truly 0
fault impedance because that would be infinite admittance. So a software vendor must assume a
very large admittance, and for this small test system that difference in admittance assumed for a
bolted fault between vendors made a noticeable difference in frequency response.
• Difference in how torque is reported was discovered between vendors. This reporting difference
explained what at first appeared to be a difference in torque.
• There were differences found in how motor speed is modeled after 100% of a motor is
disconnected.
• A bug was found in PowerWorld in the double-cage motor speed derivative calculation. The
inertia (H) of the motor was not properly scaled when only a portion of the load was tripped. The
issue has been fixed in PowerWorld Version 17 and 18 released in patch versions after September
19, 2014.
• A bug in the implementation of the LD1PAC contactor model with Tvd >0 was found in Power
World. This was fixed in the Version 17 and 18 patch on September 19, 2014.
• PSLF exhibited discrete changes in motor speed during events, this seems to have little impact on
the overall results.
• The Power World double cage induction motor gave different results than PSLF double cage
induction motor. Preferred resolution is simply to use single case induction motor models because
the composite load model represents an aggregation of a large number of motors on the system
which are a mix of single cage and double cage.
• For PSSE motor A was found to incorrectly reconnect after tripping, even if reconnect time was
set very large.
Main WECC Benchmarking
PSSE vs. PSLF
In 2015, Siemens PTI released a new play-in model that allowed a more complete benchmark of the
composite load model. Several issues were found during subsequent benchmark tests: issues dealing with
placement of compensation capacitor; issues with stalling of 3-phase motors; issues with 1-phase motor
transitioning from run state to stall state and back; other issues with motor D tripping due to contactor
operation.
Following the 2015 benchmarking, Siemens PTI revised their model by rewriting the entire model. In
order to verify the new PSS®E composite load model, a second effort of benchmarking was performed in
2017.
Study results show that Siemens PTI solved many issues with their rewrite of the composite load model:
• Fixed issues with substation transformer so high side, low side and load changing taps are
correctly modeled.
• Fixed code so substation transformer impedance changes correctly with tap changes.
• Fixed tap changing so it is locked if either the transformer reactance value is below threshold or if
the taps are locked.
• Set tap changing to stop at closest to desired value rather than first tap over/under.
• Added restart fraction to electronic load
• Feeder shunt susceptance “Fb” was moved entirely to the downstream end of the feeder, making
initialization more efficient.
• Fixed single-phase air conditioner motor model:
o The voltage time constant on contactor control (to stop oscillations caused by contactors
rapidly opening and closing) was not being used.
o Fixed some problems of P and Q variation with voltage for the regions above and below
Vbrk.
o Fixed issues regarding behavior of motor D after a portion of stalled motors recover.
o Added Vstallbrk and the intersection of stall and run curves, and transition from run to
stall curves.
o Allowed for possibility of Vstall < Vstallbrk
However, some differences between the software packages still remain: the three phase compressor model
(motor A) has a different response for the frequency event; different initialization values of motor speed;
and oscillation issues still remain with different simulation time steps.
• Motor A model:
Frequency event identified that model behaviors are different between PSLF and PSSE. The
power and torque calculation between two programs is different. Power and torque in PSSE are a
first order relationship, while in PSLF power and torque are a quadratic relationship. This could
be caused by a different motor model (motorw in PSLF vs CIM6BL in PSSE). Further
investigation needed for this matter.
• For a three phase motor, simulations with different time steps have different oscillations.
• Initialization issue:
Even the voltage at high side and low side GSU matched very well between two programs, there
is still an initialization mismatch for voltage, speed, reactive power at the load bus. The initial slip
value mismatch may be caused by differences in leakage reactance Xa between programs.
Suggest making Xa an input data for two programs. Further testing is needed after adding Xa in
model specification.
• For small test systems, the difference in “bolted fault” fault impedance between programs can
cause different results. When doing benchmarking, it is important to specify the fault impedance
used rather than using default fault impedance.
• Difference in power flow solution tolerance between the programs may cause slight differences
between programs. Specifically, PSSE has some small chatter.
Power World vs. PSLF:
• Overall models behave mostly the same
• The treatment of the single phase air conditioner model during network boundary solution was
greatly simplified in Power World. Power World had treated the model as a constant power load
during network boundary solution which caused discrepancies, but now all programs do not treat
it as a constant power load during a single network boundary solution.
TSAT vs. PSLF
The following issues were identified with TSAT during the benchmarking:
• After 100% of a 3-phase motor was tripped, its speed would decay and then upon reconnection it
would draw a large amount of reactive power to reaccelerate. The model was changed so that
speed would remain constant during a 100% trip (similar to a 99% or less trip) so that
reacceleration would not occur upon reconnection. This was consistent with the other software
vendor implementations, and ensures that 99% trip would behave similarly to 100% trip.
• Inability to monitor individual components in composite load
• Program stops with error if Vstall < Vstallbrk.
The issues are resolved in TSAT 17. In addition TSAT 17 adds options for ignoring composite load
models based on voltage, load MW, and |P|/Q ratio.
EPRI Benchmarking
• Armature leakage reactance (Xa) causes difference in motor A performance between PSSE and
PSLF. The only way to rectify this is to have Xa as a user input.
• Motor A oscillations were identified in TSAT
• Differences were identified in tripping and reconnection of motors A, B, and C for PSLF vs. the
other three vendors.
• Difference in motor B reconnection speed
• The motor D performance curve needs better documentation to ensure correct implementation,
however the latest versions of all program implementations have resolved the differences for
motor D.
• Power Electronic load and ZIP load performance is identical across all platforms
Conclusions and Next Steps
All major discrepancies between the programs tested have been resolved. Robust performance can be
achieved with all four programs. Minor differences between the programs exist. These differences are not
expected to cause significant differences in simulation results.
Next steps include:
• Consider adding armature leakage reactance (Xa) as a user input to see if this resolves
discrepancies in motor slip initialization. PSLF will change its three phase motor model from
motorw to motor1 which should resolve issues between PSSE and PSLF 3 phase motor
performance.
• Benchmark other load models, including CMPLDWG.
• Once the plug and play version of CMPLDW is fully developed, individual components will be
benchmarked.
• Continue benchmarking the model against historical system events.
Detailed Results
Early WECC Benchmarking
The aim of this benchmarking work is to test the new composite load model in PowerWorld and GE PSLF
software. Some testing is also done in PSSE with the help of Song Wang from PacifiCorp and results are
presented for PSSE when available.
This report shows that GE PSLF and PowerWorld Simulator now give results that are extremely close to
one another – including the response during the fault. Testing the match of program response during the
fault provide better verification that the tools match.
The following figures show the results for the load at Bus 102 for the simulated contingency without stalling
(includes a single-phase air conditioner model as Motor D). The figures on the left show the entire 20
second simulation, while the figures on the right show only from 0.7 seconds to 2 seconds.
Figure D0-1: Final P Results at Bus 102
0 2 4 6 8 10 12 14 16 18 20-20
0
20
40
60
80
100 102 LOAD 1 115.00 --- Pld
GE
PW
0.8 1 1.2 1.4 1.6 1.8 2-20
0
20
40
60
80
100 102 LOAD 1 115.00 --- Pld
GE
PW
Figure D0-2 Final Q Results at Bus 102
The following figures show the results for the load at Bus 103 for the simulated contingency without
stalling. The figures on the left show the entire 20 second simulation, while the figures on the right show
only from 0.7 seconds to 2 seconds.
Figure D0-3 Final P Results at Bus 103
Figure D0-4 Final Q Results at Bus 103
0 2 4 6 8 10 12 14 16 18 20-50
-40
-30
-20
-10
0
10
20
30
40 102 LOAD 1 115.00 --- Qld
GE
PW
0.8 1 1.2 1.4 1.6 1.8 2-50
-40
-30
-20
-10
0
10
20
30
40 102 LOAD 1 115.00 --- Qld
GE
PW
0 2 4 6 8 10 12 14 16 18 20-20
0
20
40
60
80
100
120 103 LOAD 2 115.00 --- Pld
GE
PW
0.8 1 1.2 1.4 1.6 1.8 2-20
0
20
40
60
80
100
120 103 LOAD 2 115.00 --- Pld
GE
PW
0 2 4 6 8 10 12 14 16 18 20-80
-60
-40
-20
0
20
40 103 LOAD 2 115.00 --- Qld
GE
PW
0.8 1 1.2 1.4 1.6 1.8 2-80
-60
-40
-20
0
20
40 103 LOAD 2 115.00 --- Qld
GE
PW
Some modifications were made to PowerWorld Simulator reporting to help BPA perform numerical
comparisons. In addition, a few modifications were made to PowerWorld Simulator to correct the treatment
of tripping a portion of an induction motor load and in the properly using the voltage measurement delay
in the contactor model in the single-phase air conditioner.
With all these changes made, as of October 1, 2014 PowerWorld Simulator and GE PSLF now show a very
good agreement in results. There are a couple open questions with regard to some discrete changes in motor
speed seen in the GE PSLF results and which equations (i.e. block diagram) should be used to simulate a
double-cage motor.
Main WECC Benchmarking
Below are benchmark results of PSLF vs. PSSE both before and after the PSSE re-write
2015 study result (all components are present) 2017 study result (all components are present)
2015 study result (all components are present) 2017 study result (all components are present)
9.6 9.7 9.8 9.9 10 10.1 10.2 10.3 10.4 10.5 10.60
0.5
1
1.5
2
2.5
P-EL
----
ELECTRONIC LOAD REAL PART (PU O
PTI
GE
-20 0 20 40 60 80 100 1200.7
0.75
0.8
0.85
0.9
0.95
1
VOLT-LOAD
----
LOAD BUS VOLTAGE MAGNITUDE
PTI
GE
2015 study result (all components are present) 2017 study result (all components are present)
2015 study result (all components are present) 2017 study result (all components are present)
2015 study result (all components are present) 2017 study result (all components are present)
-20 0 20 40 60 80 100 12022.5
23
23.5
24
24.5
25
25.5
PLOAD
----
REAL POWER
PTI
GE
-20 0 20 40 60 80 100 1205
5.5
6
6.5
7
7.5
P-ST
----
STATIC LOAD REAL PART (PU ON SY
PTI
GE
-20 0 20 40 60 80 100 1207.12
7.13
7.14
7.15
7.16
7.17
7.18
7.19
7.2
7.21
7.22
P-M4
----
MOTOR D P (PU ON SYSTEM MVA BAS
PTI
GE
2015 study result (all components are present) 2017 study result (all components are present)
2015 study result (all components are present) 2017 study result (all components are present)
2015 study result (all components are present) 2017 study result (all components are present)
-20 0 20 40 60 80 100 1200
0.5
1
1.5
2
2.5
3
3.5
4
P-M1
----
MOTOR A P (PU ON SYSTEM MVA BAS
PTI
GE
-20 0 20 40 60 80 100 1200
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
FUV-M1
----
MOTOR A: TRIP-RECLOSE FACTOR
PTI
GE
-20 0 20 40 60 80 100 1200
0.5
1
1.5
2
2.5
P-EL
----
ELECTRONIC LOAD REAL PART (PU O
PTI
GE
2015 study result (100% of motor A) 2017 study result (100% of motor A)
2015 study result (100% of motor A) 2017 study result (100% of motor A)
2015 study result (100% of motor A) 2017 study result (100% of motor A)
-10 0 10 20 30 40 50 60 700.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
1.1
1.2
VOLT-115
----
VLT-LOAD 115
PTI
GE
-20 0 20 40 60 80 100 1200
0.5
1
1.5
TE-M1
----
MOTOR A: TELE (PU ON MOTOR MVA
PTI
GE
-20 0 20 40 60 80 100 1200
5
10
15
20
25
30
PLOAD
----
PLOAD-PLD
PTI
GE
2015 study result (100% of motor A) 2017 study result (100% of motor A)
2015 study result (100% of motor A) 2017 study result (100% of motor A)
2015 study result (100% motor B) 2017 study result (100% motor B)
-20 0 20 40 60 80 100 120
0.65
0.7
0.75
0.8
0.85
0.9
0.95
1
1.05
VOLT-LOAD
----
LOAD BUS VOLTAGE MAGNITUDE
PTI
GE
-20 0 20 40 60 80 100 12019
20
21
22
23
24
25
26
27
PLOAD
----
REAL POWER
PTI
GE
-10 0 10 20 30 40 50 60 700.2
0.4
0.6
0.8
1
1.2
1.4
1.6
VOLT-115
----
VLT-LOAD 115
PTI
GE
2015 study result (100% motor D) 2017 study result (100% motor D)
2015 study result (100% motor D) 2017 study result (100% motor D)
2015 study result (100% motor D) 2017 study result (100% motor D)
-20 0 20 40 60 80 100 1200
0.2
0.4
0.6
0.8
1
1.2
1.4
VOLT-LOAD
----
LOAD BUS VOLTAGE MAGNITUDE
PTI
GE
-20 0 20 40 60 80 100 120-5
0
5
10
15
20
25
30
35
40
45
PLOAD
----
PLOAD-PLD
PTI
GE
-20 0 20 40 60 80 100 1200.88
0.9
0.92
0.94
0.96
0.98
1
FUV-M1
----
MOTOR D: FRACTION OF MOTORS NOT
PTI
GE
2015 study result (100% motor D) 2017 study result (100% motor D)
2015 study result (100% motor D) 2017 study result (100% motor D)
2015 study result (100% motor D) 2017 study result (100% motor D)
However, some differences between the software packages still remain: the three phase compressor model
has a different response for the frequency event; different initialization values of motor speed; and
oscillation issues still remain with different simulation time steps.
-20 0 20 40 60 80 100 1200.8
0.82
0.84
0.86
0.88
0.9
0.92
0.94
0.96
0.98
1
F-AC-TH
----
MOTOR D: KTHA NON-RESTARTABLE C
PTI
GE
-20 0 20 40 60 80 100 1200.5
0.6
0.7
0.8
0.9
1
1.1
1.2
1.3
VOLT-SUB
----
LOW SIDE BUS VOLTAGE MAGNITUDE
PTI
GE
-20 0 20 40 60 80 100 120-5
0
5
10
15
20
25
30
35
40
45
PLOAD
----
PLOAD-PLD
PTI
GE
• Motor A model
Frequency event identified that model behaviors are different between PSLF and PSSE. the
power and torque calculation between two programs is different. Power and torque in PSSE are
first order relationship. While in PSLF, power and torque are in quadratic relationship. This could
cause by a different motor model 9motorw in PSLF vs CIM6BL in PSSE) used by two software.
Further investigation needed for this matter.
• Oscillation issue with different simulation time step
This only applies to three phase motor. Simulation with different time step has different
oscillation. Simulation time step for plot at left is 1/3 cycles; simulation time step for plot at right
is 1/5 cycles.
• Initialization issue
Even the voltage at high side and low side GSU matched very well between two programs. But
there still initialization mismatch for voltage, speed, reactive power at load bus. The initial slip
value mismatch may cause by differences in leakage reactance Xa between programs. Suggest
making Xa an input data for two programs. Further test needed after adding Xa in model
specification.
EPRI Benchmarking