Vous êtes sur la page 1sur 6

1144 IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 19, NO.

2, MAY 2004

Hydro Turbine-Governor Model


Validation in Pacific Northwest
Dmitry Kosterev

Abstract—This paper presents an approach for generator model


validation based on the measurements taken at the point of inter-
connection. The approach is initially applied for The Dalles pow-
erhouse, where the simulated and actual governor responses were
found different. Governor tests and monitoring were performed
and resulted in model revisions. The approach also resulted in im-
provements of John Day turbine-governor models. Grand Coulee
governor response was validated and found adequate. Based on
the project experience, a process for generator model validation
is proposed.
Index Terms—Disturbance monitoring, generator testing, model
validation, power system dynamic modeling.

I. INTRODUCTION

D ECISIONS on operating transfer limits and transmission


network reinforcements are based on system studies. The
studies are relying on models to correctly predict system re- Fig. 1. Comparison of simulated and actual California-Oregon Intertie (COI)
power responses during August 10, 1996 WSCC system disturbance.
sponse to disturbances specified in the reliability criteria. There-
fore, having realistic models is very important for safe, reliable,
and economic system operation.
Under the existing process in Western Electricity Coordi-
nating Council (WECC), formerly WSCC—Western Systems
Coordinating Council, generator owners are responsible for de-
termining models of their generating equipment and providing
data directly to WECC. WECC compiles generator models
in a system database used in the planning and operational
studies. On August 10, 1996 WSCC system experienced major
collapse, while dynamic models predicted system stability
for the same disturbance scenario, Fig. 1 [1]. Following the
outage, WSCC approved a policy requiring that all generators
greater than 10 MW should be tested and their dynamic models
be validated every five years. On August 4, 2000, sustained
oscillations were observed in the system for about one minute
after losing a tie line between Alberta and British Columbia.
Fig. 2. Comparison of simulated and actual COI power responses during
Again, the model predicted a well-damped dynamic response August 4, 2000 WSCC system disturbance.
[2], Fig. 2. Studies of underfrequency events showed that
the governor response is over-represented by more than 50% need to “close the loop” by developing a mechanism for vali-
system wide. Similar discrepancies between recorded and dating and, if necessary, correcting dynamic models. In 1999,
simulated governor responses were observed in the Eastern Bonneville Power Administration (BPA) initiated an effort on
Interconnection [3]. developing techniques for validating generator models based
It is evident that existing process of model development is on the measurements taken at the point of interconnection. The
not working. This is because control of data quality is very lim- project resulted in significant governor model improvements for
ited under the existing “open loop” approach. Basically, it is a The Dalles, John Day, Bonneville and most recently model val-
“good faith” effort by equipment owners. There is an obvious idation of Grand Coulee hydro power plants.

Manuscript received April 23, 2003. II. BPA WIDE-AREA MEASUREMENT SYSTEM
The author is with the Transmission Operations and Planning, Bonneville
Power Administration, Vancouver, WA 93663 USA. Bonneville Power Administration Transmission (BPAT)
Digital Object Identifier 10.1109/TPWRS.2003.821464 recognizes that generators have significant impact on system
0885-8950/04$20.00 © 2004 IEEE
KOSTEREV: HYDRO TURBINE-GOVERNOR MODEL VALIDATION IN PACIFIC NORTHWEST 1145

Fig. 3. Model validation process.

Fig. 5. Comparison of simulated (old model) and recorded responses of The


Dalles generators to system frequency excursion caused by the outage of Four
Corners generators on December 25, 1999.
Fig. 4. One-line diagram of The Dalles powerhouse.
units, and units #15–22 are 104-MVA units. There are four pow-
erhouse lines connecting generators with Big Eddy 230-kV sub-
dynamic performance. BPAT installed several phasor mea- station (point of interconnection). The Bonneville Power Ad-
surement units (PMUs) to monitor dynamic performance of ministration installed a PMU at Big Eddy 230-kV bus.
large generating stations connected to its network. Present The Dalles powerhouse was modeled in General Electric
PMU installations include The Dalles (1776-MW capacity), PSLF program as shown in Fig. 4. The model represents all
John Day (2576-MW capacity), Grand Coulee 3rd Powerhouse four powerhouse lines, stepup transformers and all eighteen
(4598-MW capacity) and Colstrip 3 and 4 (1526-MW capacity) generators. The “infinite bus” machine is attached to the Big
generating stations. PMUs are installed at the point of inter- Eddy 230-kV bus to represent the equivalent of the WSCC
connection, and measure bus frequency, voltage phasors and system.
current phasors in each powerhouse line. Real and reactive
power can be derived from the phasors. The PMU data was B. Model Validation
used for generator dynamic model validation. BPA PMU has been recording powerhouse response to
large generation outages in WSCC system. A large library
III. MODEL VALIDATION SETUP of disturbance recordings allows elimination of single event
abnormalities.
An “infinite-bus” model was developed and implemented in Fig. 5 shows comparison between recorded and simulated re-
General Electric PSLF program. The model injects recorded sponses for the 1975 MW Four Corners plant outage on De-
bus voltage and frequency in dynamic simulations. The “infi- cember 25, 1999. IEEE mechanical-hydraulic governor model
nite bus” is modeled with a large synchronous machine with and nonlinear turbine model were used to represent the tur-
fast-responding exciters and governors. Recorded bus voltage bine-governor [3], Figures 2 and 13 in [4]. Since the recorded
and frequency are input as references for the generator voltage frequency was used to drive the model response, the simulated
regulator and governor respectively. Because of the large size of and recorded frequencies are nearly identical. However, there
the “infinite-bus” machine and fast response controls, simulated is an observable difference between the simulated and recorded
bus voltage and frequency are forced to follow very closely the power pick-up by the generators connected to the powerhouse
recorded voltage and frequency (Fig. 3). This setup was used line #6. The simulated power pickup is much faster in the first
for validating dynamic models of generators and their controls 15 seconds following the generation loss. (The power pick-up
as seen from the point of interconnection. For example, if the was measured at Big Eddy 230-kV bus). Most significantly, the
turbine-governor model is correct, the simulated power should actual governors have no response at the time of minimum fre-
be very similar to the recorded power for events of system fre- quency dip, while the simulated governors complete 50% of
quency excursions. their response by the time of minimum frequency dip.
Fig. 6 shows a comparison for Hoover trip during WSCC gov-
IV. DALLES ernor response test on May 18, 2001. The powerhouse line #6
had three units generating, two were lightly loaded and one was
A. System Setup normally loaded. There is a significant difference in amount and
The Dalles Dam is located on lower Columbia River about 80 speed of the recorded and simulated governor responses. Sim-
miles upstream from Portland, Oregon. Fig. 4 shows a one-line ilar modeling errors were observed for other Dalles and John
diagram of The Dalles powerhouse. Units #5–14 are 94 MVA Day generators.
1146 IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 19, NO. 2, MAY 2004

Fig. 6. Comparison of simulated (old data) and recorded responses of The Fig. 7. Comparison of simulated (new data) and recorded responses of The
Dalles generators to system frequency excursion caused by Hoover trip on May Dalles generators to system frequency excursion caused by the outage of Four
18, 2001. Corners generators on December 25, 1999.

C. Working With Generator Operators on Correcting


Turbine-Governor Representation
Bonneville Power Administration and US Army Corps of
Engineers worked together on resolving modeling differences.
The Dalles Dam (and many other low-head high-flow plants
on Columbia River) have Kaplan turbines, in which power is
controlled by coordinated adjustment of wicket gates and tur-
bine blades. Approximately 14 000 MW of hydro-generating
capacity in Northwest have Kaplan turbines. Existing turbine
models [4], [5] do not recognize blade angle response. Blade
angle control is shown to have significant impact on the tur-
bine power response by affecting water discharge through the
turbine and turbine efficiency. A Kaplan turbine model was de-
veloped and applied to The Dalles project. Model details are
given in the Appendix. Turbine-governor tests were conducted
including online speed reference steps and partial load rejection
tests. Monitoring equipment was left at the plant to record gov-
ernor responses to under-frequency events. Based on the test and Fig. 8. Comparison of simulated (new data) and recorded responses of The
monitoring data, the model parameters were determined. Then, Dalles generators to system frequency excursion caused by the tripping Hoover
generators during WSCC test on May 18, 2001.
the overall response was validated using disturbance recordings
at the point of interconnection, as described above.
Simulations and actual recordings are compared in: tions. A particularly good match is recorded for July 16 event
• Fig. 7 for 1975-MW Four Corners plant outage (response when only tested unit #22 was generating.
using previous model is shown in Fig. 5); Similar work has been performed for John Day and
• Fig. 8 for 750-MW Hoover trip during governor response Bonneville governors.
test (response using previous model is shown in Fig. 6);
• Fig. 9 for 2400-MW RAS trip on July 15, 2002; V. GRAND COULEE
• Fig. 10 for 2590-MW RAS trip on July 16, 2002. Grand Coulee Dam is located in eastern Washington on upper
Columbia River, and operated by the U.S. Bureau of Reclama-
Governor droop and response time were measured on units tions (USBR). The third powerhouse consists of three 615-MVA
17, 19, 20 in the same powerhouse line. We found droop to be (units #19–21) and three 718-MVA (units #22–24) generators.
different among units, ranging from 5.5% to 6.5%. Response The generators were tested by the USBR engineers as required
time was very similar, except unit #17 which was observably by WECC. The governors are double-derivative, and the turbine
faster. No attempt was made to come up with exact numbers. type is Francis.
The new model captures well The Dalles governor responses The model validation was set up in PSLF program similarly to
to system frequency excursions under various operating condi- that at The Dalles. Grand Coulee is one of major load-following
KOSTEREV: HYDRO TURBINE-GOVERNOR MODEL VALIDATION IN PACIFIC NORTHWEST 1147

Fig. 9. Comparison of simulated (new data) and recorded responses of The Fig. 11. Comparison of simulated and actual responses of Grand Coulee
Dalles generators to system frequency excursion caused by the AC RAS event generator #20 to system frequency excursion caused by the AC RAS event on
on July 15, 2002. July 15, 2002.

Fig. 10. Comparison of simulated (new data) and recorded responses of The
Dalles generators to system frequency excursion caused by the AC RAS event
on July 16, 2002 (only tested unit #22 was generating in powerhouse line #6). Fig. 12. Comparison of simulated and actual responses of Grand Coulee
generator #20 to system frequency excursion caused by the AC RAS event on
July 16, 2002.
projects in BPA Control Area and is normally operated on Area
Generating Control (AGC). To exclude AGC effects, we simu-
1) A generating company will provide a dynamic model of
lated only generation trips caused by Northwest Remedial Ac-
its generating plant. The model will characterize plant re-
tion Scheme (RAS), when BPA AGC is suspended. Figs. 11 and
sponses to system disturbances (voltage and frequency
12 show Grand Coulee #20 responses for RAS generation trips
deviations at point of interconnection, oscillations) and
on July 15 and July 16, 2002. It is evident that Grand Coulee
control signals (power and voltage schedule). The model
#20 governor response is modeled correctly.
will be a part of the power system model used in system
studies to determine operating transfer limits and network
VI. CONCLUSION reinforcements. An incorrect model may result in incor-
An approach is developed for generator model validation rect transfer limits, which can either put system at risk of
using measurements taken at the point interconnection. The failure or unnecessarily restrict transmission use.
approach is demonstrated for The Dalles, John Day, and Grand 2) The model data are submitted to a corresponding control
Coulee power plants. Lately, it has been used for thermal area. The control area will review generator model data,
governor response validation in WECC. Based on the project test, and validation report.
experience, the following process for dynamic model validation 3) The control area will monitor performance of a gener-
is proposed. ating plant at the point of interconnection. A control area
1148 IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 19, NO. 2, MAY 2004

Fig. A1. Block-diagram of The Dalles turbine controls.

Fig. A3. Power versus gate steady-state characteristic.

Turbine model inputs:


• gate servomotor stroke (per unit of full gate opening);
• blade servomotor stroke (per unit of full blade
stroke).
Turbine model output is turbine power (per unit of turbine
Fig. A2. Block-diagram of Kaplan turbine model. megawatt capacity). The turbine megawatt capacity is de-
fined at full gate opening at a specified head.
will collect bus voltage and frequency, powerhouse line Turbine model data for The Dalles #22:
currents, and control signals sent to the generating plant. water inertia time constant (s);
The data-acquisition rate will be fast enough to observe h0 steady-state head;
dynamic response of generator controls—voltage regula- no-load turbine flow;
tors, power system stabilizers, and speed governors. a gain representing blade angle impact on the tur-
4) A control area will compare recorded power plant re- bine water flow (per unit flow/per unit of blade);
sponses with those predicted by the model. If a severe dis- gate time constant (s);
crepancy is observed, the transmission utility will require blade time constant (s);
the generation equipment in question be tested, and that steady-state “Gate—Power—Blade” characteristic, Figs. A3
temporary monitoring equipment be installed at the plant and A4.
site. Control area provides operational data to WECC, in- Figs. A3 and A4 show “power versus gate opening” and
cluding which generators are expected to operate base- “blade versus gate opening” steady-state characteristics.
loaded and which generators are on AGC. At no-load, wicket gates are opened about 18% and blades are
This process will provide better control of generator model flat. The blades remain flat until the gate opens about 55–60%.
quality, and should be beneficial to the power industry. At 45% to 55% gate opening, turbine power output is “satu-
rating” as a function of gate. When the gate opens above 60%,
APPENDIX the blade angle starts adjusting, increasing turbine power output.
KAPLAN TURBINE MODEL In the normal operating range (60%–90% of gate opening), the
Fig. A1 shows a block-diagram of The Dalles turbine blade angle adjustment allows nearly linear power-gate regula-
controls. tion. The blade angle can be adjusted up to about 15 degrees
The turbine type is Kaplan with adjustable propeller blades. from flat to steep angle. Values for the flat angle are 18 degrees
The speed governor and blade angle controller are mechanical- and for the steep angle are 33 degrees for The Dalles. Blade
hydraulic. A turbine model was developed to represent effect of angle remain steep with the gate opening above 90%, and the
blade angle on turbine power output, Fig. A2. Solid lines rep- turbine power output “saturates” as a function of gate.
resent existing IEEE turbine model [3], [4], and dashed lines The blade angle controller model is developed based on in-
represent model additions to represent a Kaplan turbine. The in- spection of actual controls and manufacturer’s manual. It is sim-
tent of the model is to capture impact of blade angle adjustment ilar to a model of a mechanical governor.
on turbine power output. Blade angle affects water discharge The blade controller input is gate servomotor stroke (per unit
through the turbine and turbine efficiency in coordination with of full gate opening). The blade controller output is blade ser-
gate opening. Ultimately, both affect turbine power output [4]. vomotor stroke (per unit of full stroke) [Fig. A5].
KOSTEREV: HYDRO TURBINE-GOVERNOR MODEL VALIDATION IN PACIFIC NORTHWEST 1149

“Blade versus gate” cam characteristic, as shown in


Fig. A4.
The speed governor is adequately represented with existing
model for mechanical-hydraulic governors, Fig. 13 in [5].

ACKNOWLEDGMENT
The author is very thankful to B. Mittelstadt and G. Keenan
at Bonneville Power Administration and L. Pereira at Northern
California Power Agency for productive discussions and sug-
gestions of the proposed disturbance monitoring policy. The au-
thor is very thankful to S. Wiese and K. Martin for their effort
in installing PMUs at The Dalles, John Day, and Grand Coulee
generating stations.

REFERENCES
[1] D. N. Kosterev, C. W. Taylor, and W. A. Mittelstadt, “Model validation
for the August 10, 1996 WSCC system outage,” IEEE Trans. Power
Syst., vol. 14, pp. 967–979, Aug. 1999.
Fig. A4. Blade versus gate steady-state characteristic (blade in on cam to get [2] J. F. Hauer, M. J. Beshir, and W. A. Mittelstadt, “Dynamic performance
optimal efficiency). validation in the western power system,” in APEx 2000 in Kananaskis,
Alberta, Oct. 2000.
[3] R. P. Schultz, “Modeling of governor response in the eastern intercon-
nection,” in Proc. IEEE Winter Power Meeting, Jan. 1999, pp. 561–566.
[4] J. M. Undrill and J. L. Woodward, “Non-linear hydro governing model
and improved calculation for determining temporary droop,” IEEE
Trans. Power App. Syst., vol. PAS-86, pp. 443–453, Apr. 1967.
[5] “Hydraulic turbine and turbine control models for system dynamic
studies,” IEEE Trans. Power Syst., vol. 7, pp. 167–179, Feb. 1992.

Dmitry Kosterev received the Ph.D. degree in electrical engineering from


Oregon State University, Corvallis, in 1996.
Fig. A5. Blade controller block diagram. Currently, he is a Network Planning Engineer at the Bonneville Power
Administration, Vancouver, WA. He has been involved in many planning,
operating, and generation integration studies. He was responsible for model
The blade controller parameters: validation studies for August 10, 1996 and August 4, 2000 WSCC system
disturbances. He was actively involved in design, testing, and commissioning
Tc blade cam time constant; of Line Drop Compensation at John Day Powerhouse on the Columbia River.
Tv blade pilot valve servomotor time constant; Recently, he has been leading a project on hydro turbine-governor model
Kb blade servomotor gain. validation in the Pacific Northwest. He is involved in developing BPA generator
interconnection standards.
Unintentional dead-band is used to represent overlap in Dr. Kosterev is a member of WECC Model and Validation Work Group and
distributing valve plunger is Chairman of the WECC Load Modeling Task Force.

Vous aimerez peut-être aussi