You are on page 1of 10

1244 IEEE Transactions on Power Systems, Vol. 3, No.

3, August 1988

VARIABLE, NON-LINEAR TIE-LINE FREQUENCY BIAS


FOR INTERCONNECTED SYSTEMS CONTROL
Thomas Kennedy, Senior Member, IEEE Stephen M . Hoyt, Member, IEEE
Charles E Abell, Member, IEEE
Union Electric Company
St. Louis. Missouri

ABSTRACT tie-line frequency bias. Bvn,but was limited in its application


due t o the nature of the analog control system. The recent
In February, 1984, with the approval of the North Ameri- conversion of the generation and tie-line control from analog
can Electric Reliability Council Operating Committee, Union t o digital equipment has given Union Electric t h e capability t o
Electric began implementation of a test t o identify the im- apply Bvn. This paper describes the application experience of
provement to electric system automatic generation and tie- B, a t Union Electric, including the application algorithm and
line control that could be achieved by the application of an assessment of t h e results obtained.
variable, non-linear tie-line frequency bias. Up t o then, tie-line
frequency bias had generally taken the form of a fixed linear
function. Variable, non-linear tie-line frequency bias is a bet- GLOSSARY OF TERMS
t e r simulation of electric system frequency response than
fixed straight line bias because it recognizes the variable nat- ACE - Area Control Error (MW)
ure of a system's reaction to frequency changes. Important AGC - Automatic Generation Control
components of this response are t h e deadband in turbine- B - Tie-Line Frequency Bias (MW/.l Hz)
generator governors, variations in the number of on-line gen- Bfs - Fixed, Straight-Line Tie-Line
erators, and variations in load. The test showed that when Frequency Bias (MW/.1 Hz)
tie-line frequency bias is better matched t o system response. B,, - Variable, Non-Linear Tie-Line
the result will be improved Area Control Error performance Frequency Bias (MW/. 1 Hz)
and reduced generating unit regulation. In addition, intercon- SRC - System Response t o Frequency
nection reliability is enhanced due to the fact t h a t variable, Deviation Characteristic (MW/.l Hz)
non-linear tie-line frequency bias does not inhibit the full
governor response during large frequency disturbances,
thereby providing better interconnected systems frequency TIE-LIN E FREQUENCY BIAS
stabilization. CONTROL THEORY
For any load-generation imbalance, a frequency deviation
INTRODUCTION occurs. When frequency changes, all systems of the intercon-
nection respond to stabilize frequencythrough a load change
When interconnections between operating utilities were that is proportional to the frequency change and a turbine-
first established, tie-line control was accomplished by compar- generator governor change t h a t is inverse t o the frequency
ing net actual interchange t o net scheduled interchange. A change. The total of these two actions is t h e SRC. SRC action
good deal of excess regulation occurred as systems interacted will stabilize frequency a t a deviated value from t h e scheduled
with each other's load-generation imbalances. It soon became value but it will not return frequency to the scheduled value.
apparent t h a t some new technique was necessary to allow Frequency will remain deviated from schedule until control
each system t o regulate its own load, not interacting with action is taken t o return the frequency t o the scheduled value
other systems' regulation, while contributing to interconnec- by t h e control area with t h e generation surplus or deficiency.
tion frequency regulation. That new technique was tie-line A control area's SRC will usually vary with time.
frequency bias control. See ref. 1 .
Most of a control area's SRC will be reflected on its net tie
The first frequency bias value was a fixed straight-line t o the interconnection. By monitoring the frequency devia-
estimate, Bf,, of the system response to frequency deviation tion from schedule, the interchange deviation from schedule,
characteristic. SRC. This method served the interconnections and by using its SRC. a control area, through its Automatic
well for years but was limited in its effectiveness because a Generation Control, AGC. can determine whether the im-
system's real response to frequency deviation is variable balance in load and generation is internal to its control area, or
throughout the load cycle and non-linear due t o generator mix external to its control area. If internal, the control area's AGC
and turbine-generator governor characteristics. In 1960, Un- should control to correct the imbalance. If external t o the
ion Electric recognized the need t o use variable, non-linear control area, no AGC action should be taken, but t h e control
area's response to the frequency deviation should be allowed
t o continue until the external system corrects its imbalance
and returns the frequency to its scheduled value. Because
a 7 S?t 477-3 A paper recommended and approved
SRC cannot be continuously measured, it must be estimated.
by the IEEF Power System Engineering Committee of
This estimate is the tie-line frequency bias value, B. The closer
the 1F:EE Power Eiigineering Society €or presentation B matches the SRC of the control area. the better t h e control
at the ILEF,/PES 19'37 Summer Meeting, San Francisco, area AGC can separate internal imbalances from external ones
California, . j u l y 1 : ~- 17, 1987. Manuscript submitted and Control to the internal ones. This Will reduce t h e number
.January 2 9 , 1987; made available for printing of unnecessary control actions and thereby provide better
?lay 11, 1987. control. Therefore, the basic requirement of B is t h a t it
matches SRC.
0885-8950/88/0800-1244$01.0001988 IEEE

Authorized licensed use limited to: AMRITA VISHWA VIDYAPEETHAM AMRITA SCHOOL OF ENGINEERING. Downloaded on August 17,2023 at 08:22:20 UTC from IEEE Xplore. Restrictions apply.
1245

Bfs. as used by most control areas. is a fixed, linear value for an external imbalance, the response would be to 1 and
s e t in t h e AGC t h a t remains t h e same throughout remain a t 1 until the external system corrects frequency back
the year. On the other hand, B,, is a variable and non-linear t o 0. A reduction in AGC movement should not only improve
value which is continuously calculated and used in the AGC. the ACE performance but should also reduce generator and
B,, is modeled to match SRC through monitoring the system tie-line regulation.
load and on-line generation.
The SRC varies with t h e amount of synchronized genera-
As Stated above, the tie-line frequency bias curve is a tion and with t h e load level. If B is too large or small, the result
control parameter which simulates the control area's SRC in will be excess regulation under normal conditions. For exam-
the control loop. Any tie-line frequency deviation point on the ple, if B is too small the AGC will tend t o drive frequency
tie-line frequency bias curve will result in no control action, deviations larger and tie-line deviations smaller than t h a t re-
while for a point above the curve the AGC will reduce genera- sulting from the SRC. If B is too large t h e AGC will tend t o
tion and for a point below the curve the AGC will increase drive frequency deviations smaller and tie-line deviations larg-
generation. A properly set B will match the SRC so t h a t a er than that resulting from the SRC.
system will not control t o correct load-generation imbalances
in external systems. It will simply allow its actual net inter- Figure 1 shows t h a t control t o B,, allows full governor
change t o deviate from the scheduled amount until the exter- response to take place for major frequency deviations beyond
nal system corrects its imbalance. points F1 and F2. while control to Bf, can limit governor re-
sponse beyond these points and actually cause additional fre-
The control improvement t h a t is possible by using B,, quency deviations. This aspect of B,, is important t o
instead of Bfs is shown graphically by the analysis in Fig. 1. interconnected systems reliability because it tends t o dampen
The SRC of control area A is a time variable curve comprised of wide frequency excursions.
load response and generator response to frequency deviation,
including a distributed generator deadband. The SRC of the Even though the SRC is composed of both load and gener-
rest of the interconnection is represented by a straight line. ator response to frequency deviation, the governor response
When a load generation imbalance occurs outside of control of the generation is the largest and hardest component t o
area A. resulting in low frequency, it will respond according to estimate. Therefore, estimating governor characteristics. es-
its SRC t o Point 1. If the AGC has a Bfssetting as indicated, the pecially the dead band value, is most important in determin-
AGC will control generation to Point 2 (moving along the exter- ing the bias value. Dead band values vary as t o size and
nal interconnection SRC). By this time, the deficient external distribution around the operating point throughout the load
control area begins correcting and corrects t o Point 3 (moving cycle. Since governor slopes vary the composite system gover-
along A's SRC) after which contwl area A's AGC drops genera- nor effect can vary. Each control area should address these
tion back from 3 t o 0. Control area A's AGC moves from l t o 2 concerns when determining its bias value. ,
and 3 to 0 were in the opposite direction and unnecessary.
These unnecessary moves would have been reduced or elimi- Details of Union Electric frequency bias modes are de-
nated if the SRC were better simulated by Bvn. This is because, scribed in the Appendix.

Figure 1 .-Analysis of Frequency Bias, SRC Mismatch

Authorized licensed use limited to: AMRITA VISHWA VIDYAPEETHAM AMRITA SCHOOL OF ENGINEERING. Downloaded on August 17,2023 at 08:22:20 UTC from IEEE Xplore. Restrictions apply.
1246

THE UNION ELECTRIC SYSTEM Among the data collected were daily averages per hour of
t h e raised and lowered pulses sent t o each controllable unit.
Control Area Description The averages are calculated separately for the pulses sent in
the fixed bias and variable bias mode. The data was combined
The Union Electric Company is an investor-owned utility t o show the difference between average pulses sent in the
with corporate headquarters located in St. Louis. Missouri. It fixed or variable bias mode. Figure 3 shows information plot-
serves over 1,000,000industrial, commercial and residential ted for the Rush Island #2 600MW unit. The solid line on t h e
customers in the states of Missouri, Illinois and Iowa. a service plot represents a cubic regression fit of the data. The results
area of more than 24,000 sq. miles. show that the B,, mode sent less pulses t o t h e regulating
The Union Electric Company has had a gross instantane- unit. Other regulating units had similar results. These pulsing
ous summer peak load of 6810 MW. The present summer reductions for the B,, mode were coincident with the ACE
gross capacity is 8031 MW. Most of the system's load is served excursion reductions discussed above.
a t 60 Hz with a small fraction, some 50 MW. served a t 25 Hz.
System generation is made up of a combination of fossil, nu-
clear, run-of-the-river hydro, stored hydro, pumped storage 50 I
hydro and combustion turbines. Union Electric is intercon-
nected with 21 other utilities. The company is a member of
the Illinois-Missouri Pool, the MAIN Reliability Council and t h e
North American Electric Reliability Council (NERC).
Energy Management Computer System
Union Electric's energy management system is a Control
Data (CDC) digital computer based system consisting of four
CDC System-17 computers and two CDC Cyber 170-7303.
There are approximately 10,000 data points that are scanned
every two seconds. Automatic Generation Control, A X . is
initiated every four seconds.
Some of the major applications performed by the com- ~.
puter system are Supervisory Control and Data Acquisition AW Now MUIS JMS SEPBS JM
DATE
(SCADA). AGC, state estimation, contingency analysis and en-
ergy accounting. The digital computer system was installed in
early 1983. Installation of this system provided Union Electric Figure 2.-ACE Performance
with ability t o make the necessary calculations for determin-
ing B,,.

TEST PROCEDURES AND RESULTS


The test procedure for evaluating the value of B,, con-
sisted of alternating the use of B,, and Bf, every four hours in
the AGC. Each midnight the rotation was shifted four hours t o
avoid inconsistent comparison of the two bias modes. By
keeping track of ACE under B,, and Bf, conditions it was possi-
ble to determine t h e effects of B,, compared to Bf, on Union
Electric's system regulation. The size of the fluctuations in
ACE for each test condition was measured. Each time ACE was
calculated (currently every 4 seconds) it was checked to see if
it exceeded boundaries of & 10, & 20, & 30. and f 40 MW and
whether the value was determined using fixed or variable
bias. The number of times these boundaries were exceeded
were tallied and totaled hourly and daily.
To get a better picture of all the data collected, the daily Figure 3.--Rush Island #2 Control Pulses
totals for ACE in the fixed and variable modes for the f 10 and
& 20 boundaries were plotted showing the fixed mode minus
the variable non-linear mode. Figure 2 shows the number of
times ACE was between & 10and f 20 MW and is given a daily TUNING THE BIAS ALGORITHM
average of the number of times the boundaries were exceeded
per hour. Anytime the difference on the plot is positive, there Documented earlier in reports t o the NERC Performance
were more boundaries exceeded in the fixed mode and there- Subcommittee were changes made to some of t h e B,, model
fore the variable bias mode required less regulation of the parameters. These adjustments to t h e B,, model were de-
system than did the fixed bias mode. Using cubic regression a rived from observations of electric system behavior while op-
line was fitted t o the data plotted. It shows that variable non- erating in the B,, mode. Adjustments were made t o the
linear tie-line frequency bias performed better than fixed bias variables of the B,, model and monitored throughout the test
in reducing ACE. Other boundary curves had similar results. period to make B,, a better fit of Union Electric's SRC.

Authorized licensed use limited to: AMRITA VISHWA VIDYAPEETHAM AMRITA SCHOOL OF ENGINEERING. Downloaded on August 17,2023 at 08:22:20 UTC from IEEE Xplore. Restrictions apply.
1247

While t h e theory indicates t h a t ACE deviations as well as CONCLUSIONS


generator and tie-line regulation should be reduced and t h a t
t h e overall control performance should improve, t h e results All of the data collected up t o this point supports t h e use
are not immediately obvious when t h e mode of opera- of B,, over Bf,. Since implementing B,,, Union Electric has
tion is changed from Bf, t o B,,. To properly evaluate and tune experienced improved performance, reduced unit regulation,
B,,, t h e regulation patterns of ACE, generation, and tie-lines and production cost savings.
should be monitored and corrected with bias modeling in a
manner similar to t h e method described in this paper. More industry study needs t o be directed at a better un-
derstanding of power system control. including the action of
turbine-generator governors and other frequency responsive
OI‘HER IMPROVEMENTS devices on the electric interconnection.
One of t h e benefits of implementing B,, was t o improve
Union Electric’s control performance which has moved from REFERENCES
approximately 8 5 % t o 95% of t h e NERC Control Perform-
ance Criteria over t h e past t w o years. In addition t o imple- I . Nathan Cohn. “Power Flow Control-Basic Concepts for
menting B,, two other efforts were made t o accomplish this .Interconnected Systems:’ Electric Liqht and Power,
improvement . August-September, 1950.
One of these efforts was t o increase t h e operator’s aware- 2. AlEE and ASME Committee, “Recommended Specifica-
ness of how well he is meeting the NERC Operating Commit- tions for Speed-Governing of Steam nrbines,” IEEE 122.
tee Control Performance Criteria. A, and A2 are monitored and 1959.
presented t o t h e operator a t the end of each hour. A Control
Performance Criteria survey is calculated each day. This is 3. Thomas Kennedy, “Possible Effect of Speed Governor
circulated t o on-line personnel with an explanation of excep- Dead Band in Natural System Response:’ Attachment to
L. A. Mollman letter t o Interconnected Systems Group.
tional performance, good or bad, and discussed with t h e oper-
ators who were involved. August 9. 1960.

The third effort was directed a t improving generating 4. Charles Concordia. “Performance of Interconnected Svs-
unit response. This effort, in coordination with the power tems Following Disturbances:’ IEEE-Spectrum. June,
plant personnel, resulted in much improved generator re- 1965. pp. 70-71.
sponse and more generating units available for regulation. 5. Charles Concordia. “Effect of Prime Mover Speed Control
Characteristics on Electric Power System Performance:’
IEEE PAS-88. 1969. pp. 752-754.
SUMMARY
6. Thomas Kennedy, “Discussion of NAPSIC Operating Pol-
Over the past two years, Union Electric has been able t o icy,” letter t o NAPSIC. June 3, 1975.
reduce the number of generating units needed for system
regulation while at the same time improving its control per- 7. Thomas Kennedy, “Cost of Power System Regulation”
formance. One reason for these results is t h e improved model- from “Current Operating Problems Associated with Auto-
ing of t h e tie-line frequency bias. Over this two year period, matic Generation Control,” IEEE PAS 98, January-
Union Electric’s control performance, as measured by t h e February, 1979.
NERC Criteria A1 and A2. has improved by approximately
10%. While other improvements certainly contributed t o im- 8. S. M. Hoyt, B. K. Mukherjee. et al., “The Union Electric
proved control performance. the improved tie-line frequency Company Load Dispatch System,” presentation a t IEEE
bias model was an important contribution. Control of Power Systems Conference. Oklahoma City,
March 17-1 9. 1980.
The data collected also supports very closely t h e simu-
lated results found by Ben Oni in his study on B,, control a t the 9. Ben Oni. Non-Linear Tie-Line Bias Control of Interconnect-
University of Missouri-Columbia. See Ref. 9. ed k w e r Svstems. Doctoral Dissertation, University of
Missouri-Columbia (Dr. Lewis Walker, Faculty Advisor),
Monitoring on unit regulation will continue with regard t o 1980, pp. 175- 176.
ACE and generator pulsing. In addition, efforts will be made t o
determine ways of monitoring the effect of using B,, on tie- 10. Dr. Lewis Walker, “Investigation of Non-Linear Tie Line
line swings. Further tuning of the bias model will continue. Bias Control of Interconnected Fbwer Systems:’ IEEE PAS-
100.May 1981, pp. 2350-2356.
In addition t o matching B t o t h e SRC for better control.
each NERC control area should also provide adequate genera- 11. UE Reports t o NERC-OC Performance Subcommittee,
tion response t o control area requirements and maintain its February, 1984-August, 1986.
turbine-generator governors in an unconstrained state so t h a t 12. North American Electric Reliabilitv Council, ODeratina
they can always be ready t o act t o stabilize interconnection “Guide No. 2-Bias Settings,” NERC Operating ‘Manuac
frequency. The spirit and intent of the ‘NERCOperating Com- December 3. 1986.
mittee Criteria and Guides call for each control area t o meet
these requirements so t h a t t h e reliability and control of
t h e interconnected systems will always be maintained at a
high level.

Authorized licensed use limited to: AMRITA VISHWA VIDYAPEETHAM AMRITA SCHOOL OF ENGINEERING. Downloaded on August 17,2023 at 08:22:20 UTC from IEEE Xplore. Restrictions apply.
1248

APPENDIX FBLD = LRPHZ x SLD MW/.l Hz


100
The two modes of frequency bias control implemented at
Union Electric are Bf, and B,,. The operator can select which Where: LRPHZ = Load Response to
of the two modes is to be in effect. ACE uses whichever bias is frequency deviation
in effect by the following equation: in percent per . 1 Hz
SLD = Current System Load
ACE = (ANI - SNI) - (AF - SF) x BIAS
Currently, Union Electric is using an LRPHZ = 1 %/.l Hz.
Where: SNI = Scheduled Net Interchange
ANI = Actual Net Interchange The calculation for the system generation response,
SF = Scheduled Frequency FBGEN, is not so straightforward. mrbine-generator governor
dead bands and slopes need t o be considered as well as genera-
AF = Actual Frequency tor loading.
In the B,, mode, load and generation response to frequen- Each generator is assumed t o have a 5% droop character-
cy deviation is modeled. FBLD is the frequency bias attributed istic coupled with a k.025 Hz dead band (see Ref. 2). This
to load and FBGEN is the frequency bias attributed to genera- corresponds to a slope of 3.33961.1 Hz of rated output outside
tion. Together they make up the total system response t o of the dead band. Each unit has a status of either off, manual,
frequency, FBIAS, where FBlAS = FBLD + FBGEN. See figure 4. dispatcher, or digital. Anything other than "off' means t h a t
the generator is connected t o the system and should be con-
FRBQUBNCY BIAS TEST DATA sidered for bias contribution. The amount of contribution a
u n i t provides depends on the direction of the frequency devia-
The values for the frequency deviation and tion and the actual output of the unit. For a given case of
the deadbands are in thousandths of hz. frequency deviation outside the dead band, the bias contribu-
tion expected from each turbine generator is calculated by:
3 = ACTUAL FREQUENCY DEVIATION (AFD)
Bias Contribution = SF - AF x UGVRES MW
25 = FREQUENCY DEADBAND 1 (DB1)
20 = FREQUENCY DEADBAND 2 (DB2) Where: SF = Scheduled Frequency
-20 .= FRBQUBNCY DEADBAND 3 (DB3) AF = Actual Frequency
-25 = FREQUENCY DEADBAND 4 (DB4) UGVRES = Unit Governor Response
in MWl.1 Hz
FBLD = (LRPEZ/lOOO) x SLD LIMITS This bias contribution is compared t o t h e unit's actual
= (lO.OO/lOOO) x 5323 ability to move that amount of MW either up or down. If the
* 53.23 FBMAX = 200 unit cannot move the full distance due to a unit limit. then the
FBMIN 15 bias contribution becomes t h e difference between t h e actual
FBGEN * DBV x DSK x SGVRES output and the limit for the machine. For example, a unit has
= 0.200 x 0.500 x 71.99
= 7.20 OTHER VARIABLES
a 20 MWI.1 Hz response and the frequency deviation is - .1
Hz. The 600 MW unit is loaded at 590 MW. The response of
FBIAS = FBLD + FBGEN DBWIN = 0.200 the u n i t will follow its droop characteristic t o move the unit 20
= 53.23+ 7.20 PINC = 0.200 MWI.1 Hz x .1 Hz = 20 MW. But, it can only move 600- 590
= 60.43 = 10 MW. Since 10 MW is all the unit can move t h a t is what is
used for the governor response. All the units responses are
Notes: DBV is a variable that changes according summed together to become the total governor response,
to the following: SGVRES, where SGVRES is the sum of all the governor re-
-
IF T" - sponses in terms of MW/.1 Hz. See Figure 5. A t this point the
total generation response available for a frequency deviation is
AFD > DE1 calculated.
or add PINC to DBV
AFD < DB4

DB2 < AF'D < DB3 sub PINC from DBV

DBV < DBVMIN DBV = DBVMIN

DBV > 1.00 DBV = 1.00

Figure 4.-B,, Calculation


The calculation for FBLD is straightforward. The load
response t o frequency is directly proportional to the system
load. Union Electric has assumed that the load variation is
close t o a linear function and is described as follows:

Authorized licensed use limited to: AMRITA VISHWA VIDYAPEETHAM AMRITA SCHOOL OF ENGINEERING. Downloaded on August 17,2023 at 08:22:20 UTC from IEEE Xplore. Restrictions apply.
1249

FREQUENCY BIAS TEST DATA


Thomas Kennedy was born August 8, 1926, in
(WVR=)
GENERATOR Joplin, Missouri. He received the B.S. degree in
UNIT UNIT UNIT RESPONSE Electrical Engineering from the University of
UNIT
-
NAllE STATUS -
GEN RATING m/ - 1 m Missouri in 1949 and joined Union Electric Com-
pany of St. Louis, Missouri, that same year. He
TS 1 OFF 0 220 < 1.33 was assigned to the System Load Dispatching
rs2 OFF 0 220 7.33 Office in 195 1 and has been in system operations
OSAGE DIS 92 220 1.33 since that time. At present he is General Manager,
RI 1 DIS 461 600 20.00 System Operations, and is responsible for the
RI 2 MAN 320 600 20.00 major electric system operation at Union Electric
LAB1 MAN 594 580 19.33 Company. Mr. Kennedy graduated from the
Power System Engineering Course conducted by General Electric Com-
LAB 2 OFF 0 580 19.33 pany at Schenectady, New York, and is a Registered Professional
Lm3 )IAN 593 580 19.33 Engineer in the state of Missouri. In addition, he is a Senior Member
LAB4 MAN 580 580 19.33 of IEEE Power Engineering Society and a member of Eta Kappa Nu,
SX1 OFF 0 488 16.00 Tau Beta Pi, and Pi Mu Epsilon Professional Societies. Mr.Kennedy
sx 2 )IAN 450 488 16.00 is a member of the IEEE Power System Engineering Committee as well
HER1 )IAN 131 131 4.66 as the System Economics and System Operations Subcommittees.
He has been active in the North American Electric Reliability Coun-
MER 2 MAN 136 131 4.66 cil (NERC) Operating Committee and its predecessor dispatching
HER3 )IAN 293 290 9.66 organizations, the MAIN Region Operating Committee, and the Illinois-
HER 4 MAN 335 325 12.00 Missouri Pool Operating Committee. Mr. Kennedy is past Chairman of
CAC OFF 1 1150 37.50 the NERC Operating Committee.
KEO 25 MAN 80 133 4.00
KBOBO HAN 50 0 0.00

VEN 1 OFF 0 45 1.50


m 2 OFF 0 45 1.50 Stephen M.Hoyt was born October 29, 1931, in
VEN 3 OFF 0 100 3.60 Chicago, Illinois. He received the BSEE from
mu4 OFF 0 100 3.60 Washington University, St. Louis, Missouri, in
VEN 5 OFF 0 105 3.60 195 1. He is a member of the Tau Beta Pi and the
wT6 OFF 0 105 3.60 - * * - . IEEE Power Engineering Society.
He has been with Union Electric Company, St.
\
ASH 25 OFF 0 IO 2.33 Louis, Missouri, from June, 1959, to the present
ASH 60 OFF 0 0 0.00 time. During this period, he has worked eleven
VEN CTG OFF 0 29 1.00 years in power plant operations and sixteen years
m fBm OFF 0 44 1.50 in system operations. Mr. Hoyt is presently
HER CTG OFF 0 49 1.66 4 - z Manager of the Load Dispatching Division at
WPL cn; OFF 0 55 2.00 Union Electric. Mr. Hoyt has been involved in the development of the
digital computer based Union Electric System control center which
W L GEN OFF 0 60 1.00 became operational in 1983. He is now serving a two-year term as Chair-
man of the North American Electric Reliability Council Performance
Subcommittee.
SOVRBS = Sum of all available unit gov. responses
= 11.99

AFD = Actual freq. deviation in thousandths of lus Charles F. AbeU was born November 13,1958 in
= 2
PerryviUe, Missouri. He rewived the WEE degree
from University of Missouri, Rolla in 1983. He
Figure 5. Governor Response is a member of Eta Kappa Nu and the IEEE
Power Engineering Society.
Ref. 2 indicates a dead band of turbine-generator gover- He has been in System Operations at Union
Electric Company, St. Louis, Missouri since
nors of f.03 Hz. However, testing on the UE system has graduation in 1983. During that period Mr. Abell
shown t h a t these values might range from f .02 Hz t o f .04 has been involved in the development and
Hz for t h e most part. UE had used f .025 Hz as t h e dead band enhancement of the Load Dispatch Office energy
value. The UE algorithm accounts for the distributed dead management computer system. His primary
band in the overall composite value used in t h e B,, calculation. responsibilities have tm n in the areas of AGC, SCADA, and computer
Testing is still going on t o determine t h e distribution of the system expansion.
governor dead bands.

Authorized licensed use limited to: AMRITA VISHWA VIDYAPEETHAM AMRITA SCHOOL OF ENGINEERING. Downloaded on August 17,2023 at 08:22:20 UTC from IEEE Xplore. Restrictions apply.
1250

Discussion additional assistance granted to a troubled area for the same percentage of
bias above the area frequency response characteristic.
Nathan Cohn (Network Systems Development Associates, Jenkintown, The paper helped resolve ISC disagreements and confirmed that a bias
above the assumed frequency response is much more advantageous than a
PA): Over the years, this individual has had many discussions with Mr.
Kennedy about bias and bias control. Sometimes we have agreed, sometimes bias below it.
we have not. Always the exchanges have been friendly and educational. In the discussion section of [4], 1953, C.P. Almon, Jr., then with TVA,
Responding to an invitation from Mr. Kennedy, these questions and raised two questions. One, interestingly, asked about the possibility of a
comments are submitted. None are intended to be critical of the technique variable nonlinear bias, which was Mr. Kennedy’s approach. The other
described by Kennedy and his coauthors, who are commended for its appli- asked:
cation in a manner they consider successful. Rather, the questions seek to “Can the system, by bias settings, be made to have any reasonably
clarify certain aspects of variable bias technology, and the comments suggest desired characteristics regardless of the dictates of the natural character-
another, differing, point of view for bias setting. istics?”
This, thirty years later, reflects this discusser’s current approach. His
The questions are as follows.
response at the time was in the affirmative and went on to say:
In the light of the many variables that influence area frequency response,
including: varying makeup, magnitude, and frequency coefficient of “The time well may come when imposing definite predictable regulating
connected load; varying load levels, nonlinear incremental droops, characteristics on each operating area in this manner, instead of accepting
particularly at valve change points, and varying sensitivity and dead or relying on its varying natural characteristics, will be the prevalent
bands of on-line generators; and varying levels of system frequency and operating philosophy and practice.”
magnitudes of frequency changes; can the authors comment on the Such a time, after thirty years, has not yet and may never come. Its potential
accuracy of the on-line computations of area frequency response, and advantages, however, have been evolving in this discusser’s mind. It is
whether it is continuously checked and how? discussed in [ 5 ] .
Would the authors clarify the reasoning that placed point 3 of Fig. 1 at What then would be the advantage of such uniformly computed biases?
the location shown and explain the movement of point 2 to 3? It is not What would it achieve? Perhaps this will be clear if one recalls that a major
apparent why frequency would rise to point 3 while the external system function of bias, in addition to avoiding conflict with area frequency
was correcting frequency. response, is to have participating areas provide bias assistance to areas in
trouble or need. From July 21, 1964, to December 2, 1986, this function
3) Would the authors clarify why there is a relatively small bias for area of bias was explicitly stated in NAPSIC-NERC Operating Guide No. 2,
A SRC at frequency deviations close to 60 Hz and larger ones for greater with the following phrase concerning bias control:
frequency deviations? In asking this question, the discusser is conscious
of his own use for analysis of a curve composed of three linear segments “and thereby assist another control area in an emergency due to loss of
somewhat similar to the SRC curve of Fig. 1 and which appeared in [ I ] generation or load.”
of this discussion. He later concluded that frequency variations do not
Though such a phrase was omitted, perhaps inadvertently, from the
always start at 60 Hz. They may start as small deviations anywhere in
December 1, 1986 revision of the guide, assistance to others remains implicit
the band of f 0.03 or f 0.04 Hz customarily encountered daily. Also,
in the new wording.
+
frequency schedule is periodically shifted 0.02 Hz for time correction.
With such assistance needed, who should supply it? How much should
Thus directional reversals, resultant dead bands, and decreased sensi-
each assisting area furnish? Should it be random, or should the assistance
tivity can occur anywhere in the operating band. Correspondingly, low
’burden for each assisting area have some relation to something and what
and high incremental droops can occur anywhere in the operating range.
should that relationship be?
In [2] of this discussion, the late E.E. George noted that ranges of 6:l
This discusser suggests that bias assistance, a vital operating requirement,
were not uncommon in the incremental slopes of generator governing
would be equitably shared if it is provided by control areas in proportion
characteristics.
to their sizes, i.e., in proportion to their MW peaks. The proportion of area
bias to total system bias is the share of the total required bias assistance
The discusser is pleased that his early 1950 bias paper is [ 11 in this paper. that each area is scheduled to provide. Equitable distribution, if desired,
It summarized the characteristics of prior control techniques and noted the would be achieved if all areas had a bias of the same percentage of its peak,
advantages of frequency-biased net interchange control in all areas of the the percentage being selected to be higher than the percentage-of-peak
interconnection with no central frequency regulating area. It pointed out frequency response of any of the areas.
that a frequency bias was needed to coordinate area acontrol with an area’s The very conditions that create a variable frequency response to which
frequency response for remotely caused frequency changes. It demonstrated an area bias is linked as in the subject paper would create a variable level
through graphical analyses that the best coordination was obtained when an of bias assistance by the area in fulfilling system needs. The NERC 1987
area bias was equal to the area frequency response. That conclusion has listing of area peaks and bias settings in the NERC manual shows a peak
often been cited, as in this paper, as justification of matching bias to for Union Electric of 6835 MW and a total Eastern peak of 401 125 MW.
frequency response, which is fair enough. But from this discusser’s point It shows a varying bias range for Union Electric of 50 to 200 MW/O. 1 Hz
of view, that was only the beginning. Other papers and views later followed and a total system bias of 5573 MW/O.I Hz. Based on that data, Union
resulting in “another point of view.” Electric, with 1.7 percent of the system peak, will be scheduled to provide
Considering the uncertainty of an area’s varying frequency response at the low end of its bias range about 0.9 percent of system bias assistance.
characteristic, this discusser began to examine how bias settings that were At the high end it would provide about 3.5 percent of system bias assistance.
under, equal to, or above the area frequency response would affect system Similar percentages of required total system time correction would be
conditions and bias assistance to areas in need when there were unregulated provided by Union Electric during universal time correction periods.
remote load changes. There was also in those years considerable disagree- It is of course recognized that the extent of bias assistance to be scheduled
ment in operating circles on proper bias settings. This discusser was invited by individual areas is, as are all operating practices, a matter for system
to present his views to the Interconnected Systems Committee’s annual operating personnel to determine. This discusser suggests, however, that
meeting at Des Moines in March 1956. A transcript of the presentation there are potential advantages in bias assistance based on a common
appears in [3]. percentage-of-peak for all areas. It would provide an equitable, cooperative,
Later, expanded and formalized, the presentation became an AIEE paper and democratic systems approach.
[4] which includes curves and formulas derived mathematically for calcu- Regarding bias assistance as a “burden” or “tax” that goes with sharing
lating effects on various system and area conditions for different biases and in the benefits of interconnection, the discusser ventures in this Bicentennial
other operating parameters. Year to quote, for consideration, from Article I, Section 1 of the Consti-
That paper confirmed the fact that the best way to coordinate AGC with tution with parenthetical paraphrasing to transfer its applicability to power
area frequency response was to match bias to area frequency response. systems operations as follows:
However, it showed more, defining the effects of bias settings below or
above the area frequency response. It showed that disadvantageous operating “Taxes (bias assistance) shall be apportioned to the several States (control
results are much greater for a bias that is a given percentage below the areas) which may be included in the Union (interconnected system),
frequency response characteristic, as compared to the magnitude of according to their Numbers (MW peak).”

Authorized licensed use limited to: AMRITA VISHWA VIDYAPEETHAM AMRITA SCHOOL OF ENGINEERING. Downloaded on August 17,2023 at 08:22:20 UTC from IEEE Xplore. Restrictions apply
1251

References depressed frequency condition of the system.


As stated before, the authors have invented a
[l] N.Cohn, “Bias revisited,” presented at the Spring Meeting of the East commendable method for approximating the SRC
Central Systems Group of the North American Power Systems Inter- curve. How is the accuracy of the
connection Committee (NAPSIC), St. Joseph, Mich., Apr. 1970. L & approximation assessed? Choice of bias
N Co., Reprint E7.0014 RP. setting must be conservative to guarantee
[2] E.E. George, “Speed and load control on interconnected systems III,” that error is not on the dangerous side. How
Electrical World, Apr. 28, 1945. is this assured? How conservative should one
[3] N. Cohn, “A step-by-step analysis of load-frequency control showing try to be?
the system regulating responses associated with frequency bias, ”
presented at the 1956 Annual Meeting of the Interconnected Systems A larger bias setting in the subject area
Committee, Des Moines, Iowa, Mar. 1956. L & N Co., Reprint (more negative than SRC) will cause LFC in
E7.3181 Rp. (Presented initially to the Test Committee of the Inter- the area to raise generation which will have
connected System Committee at Cincinatti, Ohio, Feb. 1956.) the effect of more rapidly restoring
[4] N. Cohn, “Some aspects of tie-line bias control on interconnected frequency and reducing system time error
power systems,” Trans. American Institute Electrical Engineers, vol. excursion. Although obtained at the expense
75, pt. III, pp. 1415-1428, 1956. L & N Co., Reprint 461-5(5). of some overcontrol, these are worthwhile
[5] N. Cohn, “Research opportunities in the control of bulk power and benefits, in the discussers opinion. One
energy transfers on interconnected systems,” presented at New England should avoid optimization of control in any
College, Henniker, New Hampshire, Aug. 22-27, 1976. EPRI Special one area that is obtained at the detriment of
Report EPRI-377-SR, Feb. 1977. the interconnected system.
Manuscript received July 24, 1987. System conditions beyond the crossover
points F1 and F2 should occur very rarely
because these points, used to determine
Louis S. VanSlyck (American Electric present bias values, are obtained from the
Power, Columbus, Ohio) The authors are to be largest observed actual disturbances. At the
congratulated in developing an interesting present time in the Eastern Interconnected
method for approximating the non-linear system the loss of about 2500 Mw causes a
system response characteristic (SRC) of a frequency drop of about -0.06 hz. ”his
control area. The subject of this discussion defines the ordinate value of F2.
is whether the SRC is a suitable choice for
the frequency bias coefficient (B) in the
expression for ACE. If, in the interconnected system,
generation fat plants not necessarily close
The paper describes a process which takes together but totaling well over 2500 MW) is
the operating point around a path in Figbre lost within about a minute, the presently
1, the ACE plane of the subject area. Among used bias settings might cause a generation
other things, neglecting the transfer times decrease in several control areas. Could
between points on this path seems to overlook this describe an existing potential
some important considerations. Supposedly catastrophe? Perhaps €or the rare events of
the scenario begins with the system at 60 hz frequency excursion beyond 0.06 hz it would
and ACE in all control areas equal to zero. be prudent to use emergency values of bias in
The interconn&cted system has three parts of all control areas.
interest: a subject area, a disturbed area,
and a composite remainder. Most of the time the frequency deviation
from schedule is less than 0.02 hz. It is
Following loss of generation in the the discussers opinion, from observation,
disturbed area the operating point (in the that under these conditions the frequency
subject area diagram) shifts from 0 to 1 in term of ACE usually plays a minor role and
about a dozen seconds. It does this by going the overcontrol which would be eschewed by
somewhere below the ordinate of point 1 in the authors scheme would also not be very
about two or three seconds, then msvincr to 1 significant. For larger but rarer frequency
due to governor response and b e f o r e LFC excursions, the overcontrol of present
action has been realized anywhere in the practice is beneficial to the system in more
system. The true SRC curve is the locus of rapidly restoring desired conditions.
points 1 for the range of disturbance
magnitudes in the disturbed area. Points 2 Do the authors agree that use of
and 3 are nevdr really reached because insufficient bias in too many control areas
control actions in all the various areas could have a destabilizing effect in the
occur simultaneously. interconnected system? Do they suggest that
bias be set at minimum sufficiency (SRC
The trajectory from point 1 back to the value) in all areas? Can testing, as the
vicinity of the origin will depend on authors have done, in one control area assess
differences between bias settings and SRCs in the implications of universal adoption of
the many control areas and the corrective their scheme?
action of the disturbed area. There are 105
control areas in the Eastern Interconnection The return of inadvertant energy by the
at the present time. disturbed area is outside the scope of the
paper but does have points of interest. This
As stated in the paper, if the true SRC is a subsequent process that takes place,
value is used for bias there will be no LFC perhaps several weeks, after the LFC actions
control action by the subject area after that have been considered by the authors.
arriving at point 1. Further, bias settings
which put point 2 to the left of point 1 are During the idealized scenario (which bars
dangerous because the area would then lower consideration of metering and accounting
generation and aggrevate the already errors, and all control deficiencies) the

Authorized licensed use limited to: AMRITA VISHWA VIDYAPEETHAM AMRITA SCHOOL OF ENGINEERING. Downloaded on August 17,2023 at 08:22:20 UTC from IEEE Xplore. Restrictions apply.
1252

disturbed area receives an inadvertant energy W. Spencer Bloor (Consultant): The authors are to be complimented on
contribution from every other control area in the presentation of this paper and for proposing an alternate way of dealing
the system. In the terminology of N. Cohn’s with bias. A debt of gratitude is also due them for providing an opportunity
decompositlon technique there is primary to disagree with certain aspects of the nonlinear bias control. Points of
receipt f o r the disturbed area and secondary disagreement are enumerated below along with an incidental comment on
delivery for each other area. The payback a problem in Fig. 1.
will be smooth if the disturbed area alone
invokes a unilateral correction. Ei:ateral 1) The similarity between the control characteristic of nonlinear bias and
01- ~ r . i ? a t e r a lcorrect :c:r Izstitutel ir. nther a discarded control mode of years gone by suggests that the widespread
iireas wi:? initiate an osci?:ation of -
use of nonlinear bias has the potential to degrade the quality
~- of control
inadvertant energy accounts and a sequence of in interconnected system operations. The discarded control mode is
required corrections in various areas of the selective frequency [l]. The idea behind selective frequency was to
system. allow only the SRC to respond to remote changes and reserve the AGC
response for local changes. Selective frequency did not work well in
The adaption of Mr. Cohn‘s technique may practice and inspired the invention of tie-line bias 121.
be impacted by anything which makes its use The proportion of ACE accommodated by governor control of area A
more complex. With variable bias in effect in the authors’ Fig. 1 illustrates another limitation of nonlinear bias and
an integration is required to compute dependence on the initial SCR to do the entire control job. The initial
regulating deficiency in one‘s own area and SCR in Fig. 1 is less than one-half of area A’s ACE. Assuming a 2-
communicated values of variable bias from percent bias and a 0.7-percent load response portion of the total SCR
other areas are needed to compute quantities, response (all percentages are per 0.1 Hz), the governor control portion
such as primary inadvertant, that depend on of the SCR is less than 0.3 percent. This indicates that less than 10
total system bias. To keep things simple by percent of area A’s spinning capacity is responding to governor control.
using constant bias (with the possible The effect of this on units under AGC is described in [3] of this discus-
exception of rare and short term emergency sion. It is characterized therein as operating with a low Q.
situations) is very appealing in the opinion Operating with a low Q increases the regulation of AGC units. In a
of the discusser. relatively large area and especially with only one unit on AGC, the AGC
burden can be excessive and requires operation of this unit at a reduced
Notwithstanding some questions and load. One way to relieve the burden is to reduce the bias setting which
disagreements, the paper is considered a has the effect of decreasing the amount of AGC correction required.
valuable and pleasing contribution to this
subject which is far more profound than it While this reduction is advantageous from the viewpoint of the control
appears on casual exposure. area that is decreasing its bias, it is not the best way when viewed from
the needs of the ’total interconnection. From the interconnection’s
Manuscnpt received August 10, 1987 viewpoint the better way is to raise Q or put more units under AGC in
the control area.
This dependence of the nonlinear bias on Q in the control area is a
limitation which is not shared by the linear bias. The latter makes the
AGC independent of Q and the number, size, and geographical diversity
of generation sources within the control area. This Q dependence of the
Robert K. Green, Jr. (Central and Southwest Services, Inc., Dallas, TX): nonlinear bias creates an inequity in the sharing of the regulating burden
There are very few AGC papers presented that can be immediately applied throughout the interconnected system because a larger geographically
to an existing EMS to improve the AGC function. This paper is one of compact control area is able to reduce Q more readily and safely than
those few. I think that the method described is a very practical and a very are smaller areas or geographically large areas.
creative approach to improving the calculation of area control error (ACE) While it may not be a practical problem, the nonlinear bias characteristic
which is the principal feedback variable in most automatic generation control is also similarly dependent on a control area’s load characteristic. Control
(AGC) systems. areas with highly resistive loads are automatically excused from carrying
In actual practice it is very easy to forget that the bias times the frequency a proportionate share of the regulating burden needed to provide smooth
deviation is a mathematical model that simulates how many MW of the and effective interconnected system operation.
measured net interchange has been caused by governor action within the While unrelated to any limitation in the nonlinear bias technique, there
control area. Any improvement to this model whether analytical or heuristic is a problem in Fig. 1 . The frequency cannot go above 60 Hz (or its
will improve system control. scheduled value if offset for a time correction) as a result of the SCR
If the bias value(s) is too small, after a unit trips, the other control area@) and AGC responses to a load increase in the external interconnection.
will predict too small of a governor MW value and supplemental control
will back down generation, further aggravating the low frequency. If the References
bias value is too large, under normal conditions, small frequency deviations
can cause AGC to chase an ACE that never should have been presented to [I] N. Cohn, “The evolution of real time control applications to power
the controller. A problem with our classical ACE equation is that it depends systems,” presented at the IFAC Symposium on Real Time Digital
on near perfect bias values to perform properly in both the normal and Control Applications, Guadalajara, Mexico, Jan. 1983.
disturbance mode. Filtering invalid ACE values work, but improving the [2] A. J. Williams, Jr., and S. B. Morehouse, “Electric generating
model of governor contribution is a better solution. systems,” U.S. Patent 2 124 724, June 26, 1938.
The authors correctly point out that the governor contribution to a given [3] N. Cohn, “Some aspects of tie line bias control on interconnected
frequency deviation is a function of the number of units on-line within the power systems,’’ AIEE Paper 56-670, AIEE Trans. Power App. Sysr.,
control area. If we do not try to model at least that obvious change in the part 111, Feb. 1957.
bias value we risk taking incorrect control action.
I think the adaptive concepts presented can be extended further in the Manuscript received August 10, 1987.
future. For instance, if frequency is within a given dead band and frequency
is low and time is fast, the bias can be set to zero. You are then doing T. Kennedy,~S. M. Hoyt, and C. F. Abell: The authors wish to thank
constant net interchange control while the frequency error is automatically Messrs. Cohn, Green, Bloor, and VanSlyck for their detailed comments to
correcting the time error. In the same manner, if frequency is fast and time the paper. Mr. Kennedy values his many years of association with Mr.
is slow, the bias can again be set to zero. Cohn very much and also appreciates Mr. Cohn’s several contributions to
Finally, in the conclusion to the paper the variable bias is credited with the industry with regard to electric power system control. The authors offer
production cost savings, but no mention is made of the source of the savings the following response.
or the magnitude of the savings. Would the authors please expand on that As Mr. Cohn points out, the authors are well aware of the fact that there
statement? were many divergent views in the 1950’s as to the proper value of tie-line
Manuscript received July 31, 1987. frequency bias. This was probably brought about by the fact that the bias

Authorized licensed use limited to: AMRITA VISHWA VIDYAPEETHAM AMRITA SCHOOL OF ENGINEERING. Downloaded on August 17,2023 at 08:22:20 UTC from IEEE Xplore. Restrictions apply
1253 .

concept was relatively new to most systems and that analog control very areas used BvN, the assistance to control areas in trouble would be greater
much limited the flexibility of the system operator to properly simulate the and better-distributed for larger frequency excursions than with BFs. This
SRC. The operators were usually faced with choosing the lesser of two is due to the fact that BVNrecognizes and maintains full governor action as
evils-too little or too much bias. As Messrs. Cohn and VanSlyck point mentioned above. BFs,which is tied to the control area yearly peak, actually
out, too much bias is probably better than too little bias. However, with constrains and controls against full governor action during larger frequency
the advent of digital computers in present-day operations, bias can be excursions and can actually force frequency lower for large capacity
matched very closely to SRC and at the same time recognize the minimum deficiencies or higher for large capacity surpluses. Union Electric finds that
limit set by the NERC operating committee. The authors believe that this F1 and F2 are smaller than the *0.06 Hz mentioned by Mr. VanSlyck and
improved bias/SRC match gives the system operator the ability to better in fact have had F1 exceeded and the bias increased several times since
assist his neighbor in time of trouble while still maintaining good regulation BvN has been used. The authors tend to agree with Mr. VanSlyck‘s comment
for his control area. The authors believe BVNhas that advantage over BFs. that BFscould be “describing an existing potential catastrophe.”
It is the authors’ opinion that, even though there are many variables that The authors are confident that a SRC/bias mismatch will always result
make up the SRC, the continuous recognition of the individual turbine in excess regulation. In addition, as discussed in the paper, SRChias
generator dead band and the actual governor slope eliminates approximately mismatch can create undesirable frequency or tie-line excursions (such as
three fourths of the inaccuracy of the SRC simulation made by BFs. At the case mentioned by Mr. VanSlyck where 2 is to the left of 1 in Fig. 1).
Union Electric, BVNis calculated every four seconds and recognizes the In response to Mr. Bloor’s question 4, the frequency can certainly be
dead band and governor slope for each on-line unit and the changing magni- driven over 60 Hz if improper bias settings cause control areas other than
tude of the system load. Because SRC is extremely difficult if not impossible the deficient system to over control, which could be the case with BFs.
to measure in practice, a comparison other than an empirical one is not Very few systems have highly reactive loads, but even so, the load
practical at this time. However, the authors believe that a statistical analysis response portion of SRC is small and, as Mr. Bloor points out, is not a
such as that included in the paper goes a long way to demonstrate the practical problem.
effectiveness of BVN. Even though the inadvertent discussion of Mr. VanSlyck is not within
The authors recognize, as Mr. VanSlyck points out, that the initial move the scope of this paper, the authors do recognize that control areas have
from 0 to 1 (Fig. 1) goes through a transient path and does in fact take broken down the components for years and have used it to analyze their
place in seconds. However, it does settle down in six to ten seconds which inadvertent interchange. With regard to Mr. VanSlycrs concern, digital
is well under normal control response time. The move from 0 to 1 is a pure computer application techniques should allow BVNto be applied to Mr.
system response. The move from 1 to 2 is a controller action. To properly Cohn’s decomposition technique for inadvertent energy and it should add
understand interconnection control one must always keep the two types of more accuracy to the process. Union Electric’s efforts in this direction are
moves differentiated. covered in IEEE paper 31 TP 67-136.
The movements from 1 to 2 and 3 to 0 are due to the SRClBias mismatch. Even though Mr. Bloor’s experience with selective frequency control
The external system causing the frequency deviation will, by the nature of was not good, the authors point out that they have had good experience
the event, take some minutes to correct its deficiency. Control area A, with variable nonlinear tie-line f q u e n c y bias. Rather than degrade the
controlling to its BFS value, will have ample time and regulating capability quality of control as the very early method of selective frequency control
to depart from its SRC response and move from 1 to 2. Control am A will apparently did, variable nonlinear bias has improved control at Union
stabilize at 2. When the external system starts to correct its deficiency it Electric.
will move control am A along its SRC to 3. In response to Messrs. Cohn The 1.7 percent of system peak bias calculated by Mr. Cohn for Union
and VanSlyck’s question, if control area A is responding well, movement Electric is about the middle of the .9-percent to 3.5-percent range he calcu-
2 to 3 and 3 to 0 may take place in the same time frame. Movement 3 to lated for the variable bias.
0 is a component that is counter to movement 1 to 2. The graph in the In response to Mr. Cohn’s comment concerning NERC operating guide
paper was simplified showing movement 2 to 3 and 3 to 0 separately to #2, the first sentence of NERC operating guide #2 (revised December 2,
demonstrate the theory. Had control area A been on BVNthere would have 1986) covers the frequency deviation assistances to other control areas. The
been a very small or no 1 to 2 or 3 to 0 movement. NERC operating committee remains steadfast in its determination to allow
The authors agree with Mr. Cohn that the turbine generator dead band control area automatic generation control to assist other control areas in
will not always be centered for each machine. Some may be on the raise emergency situations.
edge and some may be on the lower edge with the rest somewherebetween. The authors look forward to further improvements in the use of tie-line
The Union Electric algorithm assumes that 20 percent of the generators are frequency bias such as Mr. Green suggests. They also appreciate his positive
normally at either edge of their dead band. view of future possibilities in this area. As could be expected the actual
Concerning Mr. Bloor’s remarks that BVNdepends on SRC to do the savings due to B,, are difficult to evaluate, especially due to the fact that
entire control job, the authors point out that SRC is a system response other improvementsare being made at the same time. However, the authors’
characteristic and must not be confused with controller action. These two best estimate of savings is in the range of 1/2 percent of the total production
actions must be considered separately if interconnection control is to be costs.
properly analyzed. In addition, generating unit response is a third important The authors very much agree with Mr. Cohn that each control area should
factor to be considered. consider their bias obligation as a small price to pay for the substantial
In response to Messrs. Bloor, VanSlyck, and Cohn’s concern regarding benefits they m i v e from interconnection operations. It is the authors’
the relatively small bias setting close to 60 Hz,as pointed out in the paper, opinion that the systems of NERC take that obligation very seriously as
the low response for small frequency deviations is the result of governor verified by the statements in the preamble of the NERC Operating
dead band which is in the order of *0.025 Hz. This dictates a h a l l bias Committee Reliability Criteria for Interconnected Systems Operation.
value for small frequency excursions. Likewise, for larger frequency excur- In response to Mr. VanSlyck’s question as to the universal adoption of
sions, full governor response (3.3%/0.1 Hz at 5-percent governor droop) BVNin the tie-line control of all control areas, the authors believe that BVN
plus load response is allowed. is better than BFsbut would welcome more investigation.
Some of the discussers have expressed concern regarding controller action
for larger frequency deviations. The authors are confident that if all control Manuscript received September 4, 1987.

Authorized licensed use limited to: AMRITA VISHWA VIDYAPEETHAM AMRITA SCHOOL OF ENGINEERING. Downloaded on August 17,2023 at 08:22:20 UTC from IEEE Xplore. Restrictions apply.

You might also like