You are on page 1of 6

852 IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 15, NO.

2, MAY 2000

AGC Logic Based on NERC’s New Control


Performance Standard and Disturbance Control
Standard
Maojun “Mike” Yao, Student Member, IEEE, Raymond R. Shoults, Senior Member, IEEE, and
Randy Kelm, Member, IEEE

Abstract—Presented in this paper are the design and simulation during the clock hour is within specific limits, referred to as
results of a new AGC logic. The new logic is specifically designed . B1 and B2 criteria set the system performance requirements
to work under NERC’s new Control Performance Standard and under disturbance conditions. B1 requires that ACE return to
Disturbance Control Standard with improved effectiveness and ef-
ficiency. The old Control Performance Criteria and the new Con- zero within ten minutes following the start of a disturbance. B2
trol Performance Standard are briefly described and compared. requires that ACE begin to trend toward zero within one minute
The considerations in constructing a simulation environment are following the start of a disturbance.
included. Texas Utilities Electric Company (TU), in a joint effort
with the University of Texas at Arlington, has begun implementa- B. Control Performance Standard (CPS)
tion of this new AGC logic.
1) CPS1: The CPS1 standard replaces A1 criterion. Instead
Index Terms—AGC, CPS, DCS, CPC.
of requiring ACE to cross zero at least once every ten minutes,
CPS1 takes a more reasonable approach based upon statistical
I. INTRODUCTION theory. First, an expression is identified which represents, quan-
titatively, a control area’s contribution to the reliability objective
F OR decades members of the North American Electric Reli-
ability Council (NERC) used the Control Performance Cri-
teria (CPC) to measure their control performance. However, it
of the interconnected system to which it belongs. This expres-
sion, called the Compliance Factor (CF), is comprised of two
components, frequency deviation ( ) and ACE.
has been pointed out recently that these criteria lack technical
Whenever a control area has a nonzero ACE and there is
justification. As a result, NERC replaced the CPC with the new
a frequency deviation ( ) at the same time, a nonzero CF
Control Performance Standard (CPS) and Disturbance Control
is formed that could be either positive or negative depending
Standard (DCS) [1], [2] on February 1, 1997. After January
upon the signs of ACE and at the moment. A positive CF
1998, control areas must be no less than 100% compliant with
means the control area is acting as a burden to the Intercon-
CPS1 and no less than 90% compliant with CPS2. Since the cri-
nection’s regulation requirement for that particular time. On the
teria for measuring AGC control performance have changed, an
other hand, a negative CF indicates that the area is helping the
analysis of the prevailing AGC logic was conducted with respect
Interconnection’s regulation requirement. For each individual
to the new CPS and DCS to determine: (i) if the logic is still the
clock-minute, a CF value is calculated using the clock-minute
best choice, (ii) if not, what is the better substitute. The result
average of frequency error and clock-minute average of ACE di-
of this analysis leads to this New CPS and DCS Oriented AGC
vided by its bias. At the end of each month, the overall average
Logic. A joint effort by the University of Texas at Arlington and
CF for the past twelve months is used to derive the control area’s
TU is currently underway to implement this new AGC logic in
percentage CPS1 compliance as in (1) and (2). The CPS1 stan-
TU’s Energy Management System
dard requires that each control area must have no less than 100%
compliance.
II. OLD CRITERIA AND NEW STANDARDS
A. Control Performance Criteria (CPC) (1)
CPC consists of A1, A2, B1, and B2 criteria [4]. A1 and A2
criteria set the system performance requirements under normal
operating conditions. A1 criterion requires that ACE return to
zero within ten minutes of previously reaching zero. A2 requires (2)
that the average ACE for each of the six ten-minute periods
Where is the CPS1 control target of the Interconnection (in
Manuscript received January 18, 1999. Hz), is the clock-minute average of frequency deviation (in
M. Yao and R. R. Shoults are with the Energy Systems Research Center, The Hz), is the frequency bias of the control area (in -MW/0.1Hz),
University of Texas at Arlington, Arlington, TX 76019-0048.
R. Kelm is with the Transmission Grid Management, Texas Utilities Electric and is the clock-minute average of ACE divided by
Company, Dallas, TX 75211-6716.
Publisher Item Identifier S 0885-8950(00)03837-2. the control area’s frequency bias (in Hz).
0885–8950/00$10.00 © 2000 IEEE
YAO et al.: AGC LOGIC BASED ON NERC’S NEW CONTROL PERFORMANCE STANDARD AND DISTURBANCE CONTROL STANDARD 853

2) CPS2: The CPS2 standard requires that the average of


ACE for each of the six ten-minute periods during the hour must
be within a specific limit referred to as , and no less than 90%
of compliance is required.
This standard is very similar to the A2 criterion. The dif-
ference is that CPS2 uses a statistically derived ten-minute av-
erage ACE limit, referred to as , in lieu of the A2’s empir-
ical . Note that is based only upon a control area’s load
characteristic [4]. The depends upon the Interconnection’s
(targeted RMS of ten-minute average frequency error from
schedule), a control area’s frequency bias, and the Interconnec-
tion’s total frequency bias [1], [2]. Hence, a control area con-
nected to a larger Interconnection Control tends to have a larger
due to the larger total frequency bias of the Interconnection
Fig. 1. Overview of the CPS and DCS oriented AGC logic.
[5].

C. Disturbance Control Standard (DCS)

The Disturbance Control Standard requires that ACE return


either to zero or to its pre-disturbance level within ten DCS min-
utes following the start of a disturbance.
This standard is very similar to the old B1 criterion. However,
DCS is more easily satisfied because it requires Supplemental
achieving either zero ACE or pre-disturbance ACE within ten
minutes following the start of a disturbance, while B1 requires
that ACE return to zero within the same time period. The B2
criterion is no longer required in the new DCS.

D. Advantages of CPS

One of the most significant advantages of the new CPS is Fig. 2. Two-layer CPS1 control scheme.
the elimination of the ACE zero-crossing requirement. The fol-
lowing statement is found in [3]. “One control area reported that III. DESIGN OF CPS AND DCS ORIENTED AGC LOGIC
over 75% of its control actions were required by the Al criteria
to cross zero, and almost one-half of those actions were iden- A. Overview
tified to be in a direction to adversely impact the system fre- The new AGC logic, as illustrated in Fig. 1, has four control
quency.” Therefore, by controlling units according to CPS re- modules: the CPS1 control, the CPS2 control, the DCS control,
quirements in which the ACE zero crossing is eliminated, unit and the Supplemental Control. Each control module is designed
fuel efficiency improvements and unit wear-and-tear reductions to pursue its own objective and produce its own control output.
are expected. Additionally, the ten-minute average ACE limit, From time to time, the control outputs from these different con-
, required by the CPS2 standard is significantly larger than trol modules may contradict each other, therefore, a coordina-
as required by the A2 criterion [6]. The value for control tion scheme was designed.
areas in the Eastern and Western interconnections were all dou- The new AGC also employs a performance feedback mech-
bled or tripled with respect to . A larger average ACE limit anism that helps to adjust the control tightness based upon his-
means looser control and implies that further savings on control torical CPS performance data.
costs can be expected.
B. CPS1 Control
E. Advantages of the New AGC This control module is designed to accomplish a control
area’s CPS1 compliance target. A two-layer control scheme
The new AGC logic is designed specifically to work under was designed for this control, as illustrated in Fig. 2. The
the CPS and DCS standards in order to realize all of the benefits bottom layer is the fundamental control called the Real-time
that come with the new standards, i.e., less control actions, im- CF Control, and the top layer is the high-level Long-Term
proved unit fuel efficiency, and reduced unit wear-and-tear. It is Performance Feedback Control.
designed to help a control area meet its CPS1, CPS2, and DCS 1) Bottom Layer Control: The Real-Time CF Control takes
obligations with minimum control efforts and costs. It provides the ACE and frequency error as input and determines a proper
a straightforward and easy-to-adjust means to manage an area’s control output with respect to CPS1 requirements. First, two
control performance. input variables are defined for this control, the normalized ACE
854 IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 15, NO. 2, MAY 2000

Fig. 4. CPS1 control output.


Fig. 3. XY ) vs. control threshold.
Instantaneous, smooth CF (
regulate and (2) the parameters of the Real-Time CF Control.
The former is the most effective means for control adjustment,
and the normalized frequency error, referred to as and , re-
and the latter functions as a fine-tuning method. One of the most
spectively. By substituting (1) into (2), (3) is derived as follows,
important parameters of the Real-Time CF Control is the con-
forming the basis for defining and .
trol threshold.

C. CPS2 Control

(3) This control module is responsible for a control area’s CPS2


compliance. A two-layer control scheme is employed similar to
Note that clock-minute block averages are used in (3), yet for the CPS1 Control. The bottom layer is the Real-Time Average
the sake of convenience for control and input filtering, one- ACE Control, and the top layer is the high-level Long-Term Per-
minute moving averages are used to formulate the dimension- formance Feedback Control.
less quantities (normalized ACE) and (normalized fre- 1) Bottom Layer Control: The Real-Time Average ACE
quency error).an example of this control output. Control closely monitors the development of progressive
average ACE during each ten-minute period, and promptly
issues control actions when necessary.
(4) A progressive average ACE is defined as the average of raw
ACE from the beginning of the current ten-minute interval to
the present time. It is used to represent the control area’s CPS2
condition of the moment. Note that at the end of a ten-minute
(5) interval, the value of this progressive average ACE is equal to the
average ACE for the entire ten-minute period. The ten-minute
When a control action is needed, the current value of the average is compared against for CPS2 compliance.
one-minute moving average of raw ACE is used as the control The progressive average ACE is checked against its upper and
output. Otherwise, the control output is zero if no control action lower control thresholds. Whenever a control threshold is vio-
is needed. lated, a control action must be taken. The direction of the con-
In comparing (3) with (4) and (5), it is noted that the product trol action is such that it will help lower the progressive average
of and , referred to as , produces a good indicator of ACE. Therefore, it is possible that the control action may tend
how well the control area is currently performing with respect to to increase the control area’s current raw ACE in order to lower
the CPS1 standard. For example, if is greater than 1.0, it the progressive average ACE.
means that the so-called “instantaneous” CPS1 percentage com- The magnitude of the control action is set as follows. If the
pliance for the moment is less than 100%. Based on this observa- one-minute moving average ACE is both in the right direction as
tion, the control module is designed such that the control output discussed above and larger than a predefined minimum control
is set to the current value of the one-minute moving average of value, the control output is set equal to the one-minute moving
raw ACE whenever is greater than a predefined threshold. average ACE value. Otherwise, the control output is set equal to
Otherwise, the control output is set to zero, which means no con- the predetermined minimum control value.
trol action is to be taken by this control module. Figs. 3 and 4 The progressive average ACE is computed as follows. It is
illustrate an example of this control output. set equal to the raw ACE value at the beginning of each ten-
2) Top Layer Control: The Long-Term Performance Feed- minute interval. As each new value of raw ACE is acquired in the
back Control monitors the historical CPS1 compliance data, and interval, a progressive moving average is recalculated until the
determines the amount of control-tightness adjustments needed end of the interval is reached. As a result, the control threshold
in order to comply with the CPS1 requirement by the end of at the beginning of each ten-minute interval is necessarily set
every month. Two types of control-tightness adjustments are larger than at the end of the interval. Figs. 5 and 6 illustrate an
identified. These are (1) the total capacity of units allowed to example of this control.
YAO et al.: AGC LOGIC BASED ON NERC’S NEW CONTROL PERFORMANCE STANDARD AND DISTURBANCE CONTROL STANDARD 855

Fig. 7. Benchmark power system overview.

Fig. 5. Progressive average ACE vs. control threshold. Big Frequency Error Control. It starts controlling units to help
system frequency if the frequency error is too large.

F. Prioritized Control Coordination Strategy


There are four control modules in this new AGC. Each con-
trol module is designed to achieve its own objective and produce
a control output according to its specific logic. Simultaneous
control outputs from the four modules can not be honored at the
same time since they may be contradictory to each other. The
new AGC processes only one overall control output, therefore
the control outputs from the individual modules must be coor-
dinated to yield a single overall beneficial control action.
A prioritized control coordination strategy was adopted. This
strategy develops a single output from these four modules ac-
cording to their order of priority. The four control modules are
Fig. 6. CPS2 control output.
listed from highest to lowest priority as follows: DCS Control,
CPS2 Control, CPS1 Control, and the Supplemental Control.
2) Top Layer Control: It is similar to the one for CPS1 Con- The coordination rules are, a) an activated DCS Control sus-
trol. pends all other controls, b) an activated CPS2 Control super-
sedes CPS l Control, c) Supplemental Control is considered
D. DCS Control when none of the previous three controls is activated.
This control is used when a disturbance occurs. Its responsi-
bility is to try to recover from the disturbance within ten min- IV. SIMULATION STUDY
utes using all available resources under its control. At the very
beginning of a disturbance, it evaluates its capability against the A. Simulation Environment
magnitude of the disturbance. If its capability is not sufficient to 1) Overview: In order to test this new AGC logic, a realistic
recover from the disturbance within ten minutes, it will report simulation environment was developed. This environment in-
the amount of shortage to dispatchers who in turn are respon- cludes a comprehensive AGC simulator and data for a bench-
sible to obtain extra resources. mark power system. The AGC simulator includes individual
During the ten-minute recovery time, the DCS Control will power plant dynamics and the system frequency dynamic using
update dispatchers with the following progress information: an integration step of 0.25 sec. The AGC control function is sim-
minutes left, MW’s to be recovered, MW’s that can be recov- ulated every 4 seconds.
ered by the DCS Control during the remaining time (which The benchmark power system has three control areas inter-
is system ramp rate times minutes left), system ramp rate connected with each other. The first area is the detailed study
(MW/minute, AGC/DCS controllable), and the MW’s short if area, which includes ten units of different types and MW ca-
any (which is the MW’s to be recovered minus the MW’s that pacities. The other two control areas are simplified areas repre-
can be recovered). senting equivalent external areas. Each of the two external areas
has two units. The sizes of the three control areas are illustrated
E. Supplemental Control in Fig. 7. Note that External Area II can be specified with two
Supplemental Control is a strategy to enhance the above men- different capacities.
tioned three individual controls and improve the overall control 2) Load Data and Non-Conforming Load Data: Since load
performance. One supplemental control is the Big ACE Control. change is the forcing function of the simulator, it is important
Control action is issued whenever the magnitude of ACE ex- to provide realistic and representative load data. One month of
ceeds a threshold. Another supplemental control is the so-called continuous load data recorded by two large US utilities was used
856 IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 15, NO. 2, MAY 2000

to test the new AGC logic. One set of load data includes coin-
cident nonconforming load (NCL) data. This makes it possible
to adjust the magnitude of NCL and study its effect on control
performance. The set of load data with NCL was used for the
detailed study area. The other set of load data was used for the
two external areas and made different for each by scaling and
time shifting.
3) Unit Model: The simulator is equipped with the fol-
lowing unit models [7]. The fossil unit model simulates super-
and sub-critical once-through units, boiler-following drum
units, and coordinated-control drum units. The other unit
models include combustion turbine, single-lag, hydro, and
nuclear.
4) Operating Mode: The following three unit operating Fig. 8. Overall comparison of three AGC’s.
modes were implemented. In base loaded mode a unit receives
a fixed setpoint, and does not participate in regulation or
economic dispatch. In manual dispatch mode, a unit receives a
setpoint given by economic dispatch every 5 minutes but is not
allowed to participate in regulation. In automatic mode a unit
participates in both regulation and economic dispatch.
5) Unit Commitment: A simplified Unit Commitment (UC)
scheme was incorporated to enable the simulator to commit
units in the study area as load conditions changed. Each unit is
assigned a commitment order number. Whenever the study-area
generation reserve falls below a minimum value, UC selects the
next off-line unit with the highest available order number. Simi-
larly, when the reserve exceeds a maximum value, UC turns off
the on-line unit with the lowest order number, but only if this
action will not cause the reserve to fall below the minimum.
6) Economic Dispatch: The Economic Dispatch (ED) Fig. 9. Regulation capacity effects on CF.
module is also a simplified one. Each unit is assigned an
economic participation factor. When ED is called every five CF by 41%, improved the percentage CPS2 compliance by 17
minutes, it takes the load plus net interchange schedule less the points, reduced the number of pulses by 19%, and lowered the
total generation of units in base loaded mode and distributes number of pulse reversals by 30%. Compared to the improved
this amount to those Automatic units and Manual Dispatched conventional AGC, the improvements are, 21% for CF, 4.4
units according to their participation factors. points for CPS2%, 10.4% for number of pulses, and 25% for
7) Conventional and Improved Conventional AGC Logic: A number of pulse reversals.
conventional AGC logic and an improved conventional AGC The number of pulses is defined as the average number of
logic [8] were implemented and used for comparison with the pulses that are sent to each regulating unit per hour. Similarly,
new AGC logic. the number of pulse reversals is defined as the average number
of direction changes in pulses sent to each regulating unit per
B. Simulation Result hour.
Simulation showed that the DCS Control and the Supple- Note that the CF values illustrated in Fig. 8 might be smaller
mental Control both performed as expected. Intensive studies than what most control areas actually experience. One of the
have been carried out concerning the bottom layers of the CPS1 reasons is that the units modeled in the simulator respond ideally
Control and the CPS2 Control since they are the very kernels to control in accordance with their model specifications. This is
of the new AGC. The long-term CPS1 and CPS2 performance not always the case for actual units.
feedback controls (i.e., top layers) are yet to be evaluated. Tests 2) Regulation Capacity Effects: The total capacity of regu-
will be performed following field tests of the bottom layer CPS1 lating units was found to have a major influence on the control
and CPS2 controls. Simulation highlights of the bottom layer area’s overall control performance. This is illustrated in Fig. 9
controls are presented as follows. and Fig. 10. Note that performance saturation occurs when reg-
1) Overall Comparison of the Conventional AGC, the ulating capacity reaches approximately 50%.
Improved Conventional AGC, and the New AGC: Fig. 8 3) CPS1 Control Threshold Effects: The CPS1 control
threshold (refer to Fig. 3) provides a means to adjust the shows threshold (refer to Fig. 3) provides a means to adjust the control
that the new AGC achieves the best CPS1 control tightness tightness to achieve a desired level of CPS1 performance.
to achieve a desired level of CPS1 performance and the best Fig. 11 illustrates the correlation between the control threshold
CPS2 performance with the least amount of control. With and the CF. It is noted that as the control threshold decreases
respect to the conventional AGC, the new AGC reduced the in value from 4.0, CF likewise decreases (improves). However,
YAO et al.: AGC LOGIC BASED ON NERC’S NEW CONTROL PERFORMANCE STANDARD AND DISTURBANCE CONTROL STANDARD 857

only) while the improved conventional AGC yields a CPS2 per-


formance of 89.2%.

V. CONCLUSIONS
A new AGC logic based on NERC’s new CPS and DCS stan-
dards was developed. Simulation results indicated significant
reductions in the number of pulses and pulse reversals sent to
controlling units. The new logic provides a convenient means
for controlling the degree of compliance while minimizing unit
control. These encouraging results provided justification for TU
Electric and the University of Texas at Arlington to begin a joint
effort to implement the new AGC logic in TU’s Energy Manage-
ment System. Field tests will be conducted and data compiled,
Fig. 10. Regulation capacity effects on CPS2%. analyzed, and reported at a later date.

REFERENCES
[1] North American Electric Reliability Council (NERC), “Policy 1—Gen-
eration Control and Performance,” in Operating Manual, Dec. 1996, pp.
pl.l–14.
[2] North American Electric Reliability Council (NERC), “Performance
Standard Training Document,” in Operating Manual, Nov. 1996, pp.
ps.l–20.
[3] Control Criteria Task Force, Performance Subcommittee, NERC,
Control Performance Standard and Disturbance Control Standard
Frequently Asked Questions, Nov. 1996.
[4] North American Electric Reliability Council (NERC), “Control Perfor-
mance Criteria Training Document,” in Operating Manual, Aug. 1996,
pp. cpc.1–20.
[5] North American Electric Reliability Council (NERC), Control Perfor-
mance Standard Limits (1997), 1998.
[6] N. Jaleeli and L S. VanSlyck, “Control Performance Standards and Pro-
cedures for Interconnected Operation,” EPRI, Final Report RP3555-10,
Fig. 11. CF vs. CPS1 control threshold. Aug. 1996.
[7] V. Kola, A. Bose, and P. M. Anderson, “Power Plant Models for Oper-
ating Training Simulators,” IEEE T-PWRS, vol. 4, no. 2, pp. 559–565,
May 1989.
[8] R. R. Shoults, M. Yao, R. Kelm, and D. Maratukulam, “Improved
System AGC Performance with Arc Furnace Steel Mill Loads,” IEEE
T-PWRS, vol. 13, no. 2, pp. 630–635, May 1998.
[9] D. C. H. Prowse, “Improvements to a Standard Automatic Generation
Control Filter Algorithm,” IEEE T-PWRS, vol. 8, no. 3, pp. 1204–1210,
August 1993.

Maojun “Mike” Yao received his BSEE and MSEE degrees from Shanghai
Jiao Tong University (Shanghai, China) in 1987 and 1990 respectively. He is
currently a Ph.D. candidate at The University of Texas at Arlington and works
full-time for Central and South West Services, Inc. as an EMS engineer. His
industrial experience included 5 years with Jiangsu Electric Power Company
(Nanjin, China), a half-year EMS on-the-job training at Empros Power Systems
Control of Siemens, and a summer internship at Bailey Network Management
Fig. 12. CPS2% vs. CPS2 control threshold. Technology, Inc.

when the control threshold decreases below 0.75, CF begins to


increase (made worse). What is not shown in Fig. 11 is that the
number of pulses and pulse reversals always increase when the Raymond R. Shoults received his BSEE and MSEE degrees from New Mexico
State University in 1963 and 1969, respectively; and the Ph.D. degree from the
control threshold is decreased. University of Texas at Arlington (UTA) in 1974. He is Professor of EE, and Di-
4) CPS2 Control Threshold Effects: Fig. 12 illustrates that rector of the Power System Simulation Laboratory (PSSL) at the Energy Sys-
the CPS2 compliance reaches maximum when the control tems Research Center.
threshold is 0.5. A good choice for the threshold is 0.5 or
greater. When a looser CPS2 control is desired, one may choose
a bigger threshold or even turn the CPS2 Control off.
Randy Kelm received his BSEE degree from the University of Texas at Austin.
As a note of comparison the new AGC yields a CPS2 perfor- He is a Sr. Engineer in the Transmission Grid Management Unit of Texas Util-
mance of 88% with the CPS2 control turned off (CPS1 control ities Electric Company, Dallas, Texas.

You might also like