Professional Documents
Culture Documents
NERCs New Control Performance Standards
NERCs New Control Performance Standards
3, August 1999
Authorized licensed use limited to: AMRITA VISHWA VIDYAPEETHAM AMRITA SCHOOL OF ENGINEERING. Downloaded on August 20,2023 at 08:40:52 UTC from IEEE Xplore. Restrictions apply
1093
That is: rapidly changing obligation, as well as episodic mismatch
events (such as unexpected loss of generation.) Thus, if it
ACE Ta-Ts-lOB(Fa-Fs), (3) is not too large for too long, ACE is acceptable, provided
an area can make no accurate forecast of its future ACE
where the subscripts “a” and “s” refer to actual (mea-
trend.
sured) and scheduled (manually set) values, respectively.
When ensuing ACE averages are non-predictable, a
T, is the area’s net scheduled interchange and F, is sched-
control area is using the interconnection’s power exchange
uled frequency. The interconnection frequency error from
only on a random basis. On such basis, interconnection
schedule is A F = Fa - F,. The area’s share of support for
assistance becomes judicious and mutually beneficial for
interconnection frequency is lOBAF, where B is the area’s
all areas. When an area’s ACE is typically off-zero in the
defined frequency bias coefficient in MW per tenth Hz (a
same direction for the same hours of the day, the area has
negative number, not necessarily constant.) Changes of
a predictable ACE structure. The daily cycle of time error
scheduled interchange (except dynamic) are normally fil-
which exists in all NERC interconnections is evidence of
tered (ramped) over ten minutes so that they can be ac-
repetitious structure in ACEs. For interconnected opera-
commodated at practical rates of generation adjustment.
tion to be fair, performance standards should be sensitive
Dynamic schedules are sometimes used to “electroni-
to non-random ACE averages. This point is an important
cally transfer” load from one area to another, the second
consideration in the new performance standards.
area becoming responsible for matching the trend of the
Compared to A1 and A2, the new performance stan-
transferred load. Areas without native generation may
dards will allow larger average ACE for a few minutes,
use a dynamic interchange schedule to serve their entire
provided that sub-hourly ACE averages are maintained
load.
within reasonable bounds.
For areas with no native load, the first term in (1)
Consider average obligation over the next T minutes:
can be zero if the area has no transmission losses. The
a If T is small, less than 5 to 10 minutes, or so, change in
new performance standards are applicable to areas with-
out load or generation. All control areas, however, will average obligation is difficult to accurately predict, and
have some interconnection frequency support responsibil- very difficult, or impossible, to match with generation
ity given by a defined frequency bias coefficient B. changes.
0 As T becomes larger, change in average obligation be-
Throughout an interconnection, ACEs measure what-
ever mismatches exist in the presence of primary control. comes less difficult to predict, and more easily matched
Performance standards, or criteria, assess the overall con- by generation maneuvering.
trol process-the aggregate result of primary, AGC (au- Beyond some value of T , call it T,, average ACE should
tomatic generation control) and manual unit maneuver- be essentially unpredictable (even in sign), otherwise it in-
ing. Standards are needed to specify how closely each dicates anticipated off-schedule operation. Thus, T-min-
area should match generation with the trend of its obliga- Ute average ACE should be random for any averaging in-
tion. Through performance auditing, standards also help terval greater than T,. When this randomness is reason-
with the timely detection of metering errors and sched- ably achieved, AVG,{ACE}, for T > T,, will be reason-
ule setting mistakes. Standards should, of course, allow ably non-coincident with its counterpart AVG,{ACE} in
maximum and fairly shared benefits of interconnected o p other areas of the interconnection. (The authors believe
eration without compromise of reliability. that T, would be less than 30 minutes for most areas, de-
At a control area’s discretion, modifications of the sig- pending on an area’s load and generation characteristics.)
nal given by (3) may be used to develop a signal for control
IV. BASES FOR PERFORMANCE STANDARDS
intended to correct for past error accumulations, e.g., in-
advertent. It should be noted, however, that (3) is the Based on the foregoing observations, strategic design of
only ACE definition on which performance is to be re- performance criteria should consider that:
ported to NERC.5 Thereby, all control areas will be re- * To realize interconnection benefits, neglecting to mat-
porting performance on a common basis. ch rapid obligation changes must be permitted to some
practical extent.
111. STRATEGY FOR PERFORMANCE STANDARDS * The allowed 2’-minute average ACE should decrease
Too tight control of ACE would be impractical and as a function of T .
costly. When tolerated at a reasonable extent, ACE ex- * Area performance over 5 minutes, 20 minutes, or an
cursions allow bulk electricity suppliers to obtain mas- hour, is not as meaningful as performance over a long
sive benefits from interconnected operation. However, it sequence of 5 minute, or 20 minute, etc., intervals.
is necessary to establish operating performance standards Therefore, the new criteria have been designed to con-
for this toleration so that all parties may fairly share the sider long duration statistics of various T-minute in-
benefits of interconnected operation. tervals. NERC’s Performance Subcommittee selected
Spontaneous power exchange is highly beneficial be- 12 months as a practical choice of duration.
cause it allows the interconnection to help serve an area’s These strategies lead to the conclusion that ACE should
Authorized licensed use limited to: AMRITA VISHWA VIDYAPEETHAM AMRITA SCHOOL OF ENGINEERING. Downloaded on August 20,2023 at 08:40:52 UTC from IEEE Xplore. Restrictions apply
1094
satisfy a decreasing function of T : obey (4), then AF will obey a function similar to (4)-
another decreasing function of T . That is:
RMS{AVC:,{ACE}} = a ( T ) , (4)
RMS{AVG,{AF}} = cp(T). (5)
where a ( T ) is the root-mean-square (RMS) of all the T
minute average ACE values over the past 12 months. The similarity between a ( T ) and v ( T )depends highly
It is shown in [2] that if ACE were a random signal, on the coincidence of ACEs in the interconnection. When
a(T) would be proportional to 1/@. Of course ACE it- AW,{ACE} is random for T > T,, cp(T) will decrease
self can not be made to meet this condition because its for T > T,. When this condition is typical for a MW
next data cycle value is far more likely to be close to the majority of areas, frequency error will have small hourly
preyent value than to be random. However, a good con- average (assuming Tc < 60 minutes and ACE magnitudes
trol algorithm can make AVG,{ACE} nearly random for do not greatly exceed their past levels.)
T T,. Moreover, this can be accomplished with far less
a:
Average frequency error (or average ACE) over On-
generation maneuvering than that of many present AGC Peak, Off-peak, any given hour of the day, or even a few
schemes. minutes, indicates a level of operational error over the
corresponding time frame. Nevertheless, frequency error
V. INADVERTENT needs to be tolerated up to some practical level to realize
Operating to zero inadvertent for an hour, or group of benefits of interconnected operation via randomly occur-
hours, is not good area performance unless there happened ring off-zero ACEs. But the longer the *timeframe, the
to be zero average frequency error for the same interval. smaller the average operational error should be. Small
In that event (and only in that event) the area managed hourly average frequency error allows a larger peak-to-
a zero average ACE for the interval. When operating to peak variation about its hourly average. This translates
zero average ACE, an area will, or will not, incur inadver- directly to an allowance of larger peak-tGpeak ACE vari-
tenl; depending on average frequency error. When an area ation about its hourly average, with accompanying re-
inciirs inadvercent while operating at zero average ACE it duction of control costs and greater benefits of intercon-
deserves to ha,ve its inadvertent account corrected while nected operation. That is, for control practice to allow
continuing to operate at zero average ACE. To be mean- the largest practical RMS{ACE}, AVG,{ACE} must be
ingful, coordinating inadvertent balances should always be appropriately bounded for larger values of T .
per formed by giving proper consideration to lAFl hourly
averages. These facts should be axiomatic for anyone as- VII. CPSl
sociated with interconnected power system operation. A limit condition on AF may be written as:
With small hourly average IAFI, inadvertent energy
accounts would become more appropriate (as an opera-
tional performance assessment), and in fact, would be-
come tightly bounded for good performing areas3 As per- where A F , is the one minute average of A F , and el is a
formance indicators, inadvertent accounts are presently target bound for the 12 month RMS of one minute average
rather meaningless because time error corrections do not interconnection frequency error (set by NERC).
separately correct for frequency error from schedule in the Mathematical relationships between frequency and in-
On-Peak and Off-peak accounting intervals. One con- terconnection ACEs show that ACE coincidence with fre-
sequence of this time error correction practice is t o fre- quency error is a measure of the non-randomness of ACE.
quently trap On-Peak and Off-peak inadvertent accounts That is, non-randomness of ACE can be measured by its
so there is little opportunity for payback. The longstand- coincidence with A F . (The “coincidence” of signal z with
ing lack of any criteria incentive to control ACE averages signal y, is AVG{z x y}.) Analyses also show that coin-
over intervals longer than ten minutes could be partly to cidence among ACEs are significant contributors to AF
blame for this situation. This subject is fully explained in magnitudes. These studies lead to basing CPSl on the
[2] by theory <andwith examples. dimensionless compliance factor:
Authorized licensed use limited to: AMRITA VISHWA VIDYAPEETHAM AMRITA SCHOOL OF ENGINEERING. Downloaded on August 20,2023 at 08:40:52 UTC from IEEE Xplore. Restrictions apply
1095
tor, CPSl is a percentage defined as: month condition to be met 90% of the time. The CPS2
standard is based on the dimensionless compliance factor:
cpsl= 100 (2 -AW={CFI}), (8) 1 -
CF2 = -IACEio 1, (10)
L 10
where the average (AVG) is taken over one month for re- -
where, ACE10 is 10-minute average ACE and
porting purposes, but is a twelve month rolling average for
compliance measure. Note that when AVG{CFl} is less
or greater than 1, CPSl is correspondingly greater or less
than 100%. An area fails compliance if CPSl is less than The number Blo is the area’s average B over the ten
100%. minute interval of assessment. The multiplier 1.65 is the
All one minute average data values are to be obtained statistical conversion factor from a 68.3% confidence limit
from topof-minute to topof-minute, and time stamped (1 standard deviation) to a 90% confidence limit. The
with this “clock minute”. Only a few numbers need to be parameter v relates the size of the area to that of the
retained hourly, or daily, monthly, etc., to compute the interconnection.
12 month average. However, NERC has required that one A derivation of v based on fair considerations for elec-
minute data be retained for possible further study and tric interconnection is given in [2]. If all areas use constant
analysis. B , then v = B/B,. Thus, if all B values are constant, (11)
Alternative to defining CF1 with one minute averaged simplifies to:
data, AGC cycle rate data might have been used. How- Llo = 1.65~10 J(10B)(1oB,). (12)
ever, it is practical to perform computations only as of-
ten as necessary to provide acceptable accuracy. One For most areas, Llo values given by ( l l ) , using NERC’s
minute averaging also has other advantages because of recommended €10 target for each of the NERC Intercon-
time stamping facility, and effacing the range of AGC cy- nections, are larger than those used in the A2 criterion.
cle rates among areas. After some study and considera- They are much larger for many areas. Appendix B, of [2],
tion, the NERC Performance Subcommittee chose the one lists these limits as well as all other control area parame-
minute rate.4 Control area processes, however, will still re- ters related to CPSl and CPS2.
quire ACE data acquisition at their typical 2 to 4 second To measure compliance with CPS2, one first computes
rates. the ratio of ten minute interval counts:
NERC Standards include rules for defining an area’s Number of intervals that CF2 > 1
frequency bias coefficient, B, which establishes an area’s R = (13)
Total number of intervals ’
share of secondary frequency support. One may note,
however, that a small IBI makes (7) more sensitive to then CPS2 is the percentage measure:
ACE magnitudes. Alternatively, via (3), a large IBI com-
CPS2 = 100 (1 - R). (14)
mits the area to more interconnection frequency support
responsibility. The interval counts in (13), (6 per hour), are over
In regard to CF1, note that under the hypothetical con- one month for reporting purposes, and over rolling twelve
dition that ACEs in all other areas add to zero, AF = month durations for compliance measure. An area fails
ACE + (-lolls), where B, is the sum of B values for all compliance if CPS2 is less than 90%.
control areas. In this event, CF1 is directly proportional It should be noted that CPS2 is insensitive to ACE non-
to ACE2. Indeed however, sensitivity of CPSl to magni- randomness or its coincidence with other ACEs. Hence,
tudes of ACE exists whether or not all other ACEs add to the chosen Llo values are appropriate only if the coinci-
zero. This fact depreciates the need for CPS2. dence among ACEs does not significantly increase.
Since (7) allows areas to benefit from a large [ACE1 NERC’s function is to set performance standards, not
when ACE x AF is negative, a second performance stan- to specify control practices to meet them. Each control
dard, CPSP, is applied to ten minute average ACE. This area has the opportunity to find methods best suited to
standard is derived from an interconnection objective: its particular situation. Because CPSl considers only one
minute averages of AF and ACE, the benefits of control-
ling so that AVG,{ACE} becomes more tightly bounded
for larger values of T may not be obvious. For this reason
where AFIO is the ten minute average of AF, and €10 is a the NERC Performance Subcommittee gave serious con-
target bound for the 12 month RMS of ten minute average sideration to adapting a performance standard based on
interconnection frequency error. a compliance factor:
This standard is similar to the A2 criterion, but with
a technically defensible “ L d ” . Like A2, it is a rolling 12
CFh = -
1 ACE
E:
{ -}
-lOB h
xAFh,
Authorized licensed use limited to: AMRITA VISHWA VIDYAPEETHAM AMRITA SCHOOL OF ENGINEERING. Downloaded on August 20,2023 at 08:40:52 UTC from IEEE Xplore. Restrictions apply
1096
Authorized licensed use limited to: AMRITA VISHWA VIDYAPEETHAM AMRITA SCHOOL OF ENGINEERING. Downloaded on August 20,2023 at 08:40:52 UTC from IEEE Xplore. Restrictions apply
1097
Authorized licensed use limited to: AMRITA VISHWA VIDYAPEETHAM AMRITA SCHOOL OF ENGINEERING. Downloaded on August 20,2023 at 08:40:52 UTC from IEEE Xplore. Restrictions apply.
1098
s. p. Hoffman, H. F. Illian (ComEd, Lombard, IL): The of Cps2. Clearly, any issue of performance should be
authors should be recognized and commended for their thoroughly answerable by examination of acquired and
contributions to the development of the New Control retained data. Needed data should be on hand whenever
Performance Standards. We generally agree with their its review seems worthwhile. Thus, we again call out the
analysis and discussion. We have only one substantive importance of establishing an interchange data clearing
comment. house for ongoing reconciliation of schedules and inter-
The authors have suggested that CPS2 be replaced with changes to aid in the timely identification of operating
an hourly limit that could be represented by a frequency practice deficiencies or metering malfunctions when they
profile value called Eh, epsilon sub h. By setting E,,, at a value occur. This function would also provide the benchmark
less than E,, they would be able to control the shape of the data set mentioned in Mr. McReynolds last paragraph.
profile of frequency variation versus measured period length. NERC’s new transaction tagging procedure addresses the
This would force the reduction in the RMS of the frequency schedule side of the clearing house function. Equal dili-
error as the measurement period increases from one minute to gence should be given to the metered interchange side.
one liour, thus assuring that undesirable coincidence between Messrs. Hoffman and Illian provide an excellent
ACES is eliminated. We agree with this goal of having description of points we have made in the paper and in the
smaller average frequency variations as the measurement EPRI report [2],and what they say is technically accurate.
period increases. However, for the operator who is not statistically astute,
However, we do not agree that an additional standard for CPSl (with or without CPS2) might suggest coordination
frequency that nieasures the hourly variation is necessary at of control with interconnection frequency, and might not
this time. The reason for this can be seen if one considers the suggest controlling sub-hourly ACE averages. The first
nature of the measurements. The one-minute measures used could be counterproductive while the second would be a
for CPS1 include the RMS values of the coincident variation missed opportunity.
for all periods greater than one-minute, including the RMS An hourly CPS, (call it CPSh), based on an Eh objective,
values for the hourly measure. Since the hourly measure is sends a clear message that avoids both these subtleties.
included within the one-minute measure, anytime control All this being true, there is a still more significant benefit
alternatives are considered for meeting the one-minute limit, in establishing CPSh. It can be defined in terms of energy
those alternatives will include both the reduction of the and time error data (as described in [2]), thereby providing
variation for periods between one-minute and one-hour and an auditing check on power and frequency data. All these
the reduction of the variation for periods greater than one- benefits would be secured by establishing CPSh and the
hour. Our expectation is that the economic cost of reducing clearing house function. In regard to the historical eh data
the variations for periods greater than one-hour will be needed for a CPSh target-it should be directly obtainable
significantly less than the economic cost of reducing from records of hourly time error.
variations for periods between one-minute and one-hour. Mr. McReynolds notes that the WSCC abandoned
Therefore, the wedge shape that the authors have described as automatic time error control in October 1997 for prac-
being desirable will occur naturally as the result of the tical reasons. To continue would have necessitated new
economic decisions that will be made with respect to the infrastructure development. The WSCC has needed to
control actions implemented. If the economic drivers invoke several manual time error corrections in the past
assoc iated with control cause the appropriate characteristic, few months. Time error is a direct indication of non-
there will be no need for the additional measure. random ACE averages and the coincidence of ACES.
If one takes the view that less is usually better, then the CPSl is designed to be sensitive to such control defi-
prudent approach would be to wait and see if the economics ciency and reasonably meeting CPSl by most areas will
associated with control will achieve the goal without reasonably achieve €1, .the targeted value of R M S { A F } .
implementing an additional measure within the Control This achievement, however, does not limit the integral of
Performance Standard: AF, and hence, long-term accumulations of time error
Manuscript received June 29, 1998. are unbounded by CPSl. Indeed no combination of
CPS1, CPS2 and/or CPSh can force the integral of A F
to converge. A practical method t o establish such an
Nasser Jaleeli and Louis S. VanSlyck: We would first like objective is described in Section 4 of [2].
to thank all the discussers for their comments. During the Inadvertent accounts are only meaningful if accumula-
years we worked with the NERC committees on this devel- tions of long-term On-Peak and Off-peak time errors are
opment they were all active participants in the research individually limited. Moreover, as accurate and timely
effor::. We particularly thank Mr. Badley who adminis interchange accounting is crucial to the business process,
tered the work so effectively as chairman of the CCTF. reconciled energy and time error data should be considered
Mr. Badley mentions likely benefits of the new stan- as the basis for judging not only the accuracy of power and
dards in the marketing function. Indeed, Messrs. Illian frequency data but the quality of areas’ performance as
and Hoffman have published a method for this in [7] and well. The EPRI report [2] suggests that only those areas
[8]. Mr. Badley also mentioned deterioration of WSCC causing imbalance should be responsible for its correction.
frequency performance as a possible call for re-examination These business objectives would require little new infras-
Authorized licensed use limited to: AMRITA VISHWA VIDYAPEETHAM AMRITA SCHOOL OF ENGINEERING. Downloaded on August 20,2023 at 08:40:52 UTC from IEEE Xplore. Restrictions apply
1099
tructure, only a strengthening of what exists-namely the and the coincidence of ACES. Our recommendation is to
clearing house function and a suitable control criterion like implement the suggested CPSh.
that presented on page 7-3 of 121. Mr. McReynolds worries about a scenario where there
Mr. McReynolds mentions a stopgap practice of is a constrained flow path between a region with good
suspending AGC action when the correction of ACE would control taking advantage of the “liberal” limits of CPSl
worsen frequency (i.e., when ACE x AF is negative). This and CPS2 and a region with inadequate contingency
will be effective only if the AGC action is realized before reserves. Where a flow path is so weak as to be endangered
AF changes sign. Although not recommended, implemen- by normal condition ACE excursions the path must be
tation of this practice may be acceptable as a temporary severely endangered by disturbance possibilities. It seems
scheme. As future improvements are made, however, the to the authors that such a condition should be addressed
objective should be to control with the largest RMS{ACE} by shoring up the network rather than tightening bounds
that meets NERC CPS Standards (while maintaining on performance standards.
transmission flows within reliability constraints, whenever
they are more stringent). This concept is described in [2] REFERENCES
as priority-based control. [7]Illian, Howard F., Hoffman, Stephen P., “Real Energy
Mr. McReynolds notes that p(T) is not decreasing Interconnected Operations Services”, Proceedings of
satisfactorily as T increases. As noted in the paper, the American Power Conference, April 1998, Vol. 60-11,
RMS{AVGT{X)}only decreases as 1/@ if x is a random Pages 1070-1085.
signal. Clearly, the result that p(T) is not decreasing [S] Illian, Howard F., Hoffman, Stephen P., “Enabling
enough is directly related to non-random ACE averages Market Managed Reliability”, ibid, Pages 1089-1097.
Authorized licensed use limited to: AMRITA VISHWA VIDYAPEETHAM AMRITA SCHOOL OF ENGINEERING. Downloaded on August 20,2023 at 08:40:52 UTC from IEEE Xplore. Restrictions apply.