A Study on AGC Scheme Based on Real Time

Frequency Characteristics
Hyun-Shin Park Kern-Joong Kim
Chungnam National University, Korea Chungnam National University, Korea
mrhspark@[Link] kjkim@[Link]

system electrical frequency to nominal frequency because

Abstract- Automatic generation control (AGC) has been one of there is no tie line flow among control areas [2].
the important issues in the operation of local area and AGC uses some field data measured through the
interconnected power systems. AGC provides power demand supervisory control and data acquisition (SCADA) system
signals for AGC power generators to control frequency. In
AGC, some variables of power system such as frequency and which is usually implemented in energy management system.
power generation are measured to calculate the area control The measured data include power output of units, tie line
error (ACE). The ACE value is an important factor of AGC power flow and electrical frequency of control area, etc. The
performance. Generally, ACE value is determined by a constant power generation demand signals from the AGC function are
frequency bias factor of power system or 1% of load method. provided to the related power generation units through
The ACE value based on the constant frequency bias factor or
1% of load method, however, may not represent the real time SCADA system.
status of power system correctly. The incorrect ACE value is a In AGC program, the frequency of power system and
cause of over regulation of power system. This paper presents interchanged power flow among control areas are used to
the method to decide actual frequency bias factor on real time determine area control error (ACE) value. ACE is critical for
frequency characteristics. In this paper, the actual frequency operation of AGC. In a general rule, ACE calculation
bias factor is determined by the measured variables such as
generator power output change and frequency deviation due to includes frequency deviation, power flow deviation on tie-
disturbance of power system. The actual frequency bias factor line, and prime mover control of units. AGC controller is
calculated is used to determine the ACE of power system. The designed to regulate ACE to zero. The frequency bias factor
AGC divides the calculated ACE among AGC power generators is an important factor for ACE calculation of control areas.
periodically. The period is determined by the time of recovered The frequency bias factor is often identical to the frequency
frequency. This paper also presents the simulation results for
the power system model about the proposed AGC scheme. The response characteristic of the control area.
simulation results show that the performance of AGC is This paper describes the merits and demerits of the existing
improved when the proposed AGC scheme is applied to AGC. ACE calculation methods using constant frequency bias
factor and the one percent-load method. And then this paper
presents the ACE calculation method based on real time
I. INTRODUCTION frequency response characteristic. This paper also provides
the reasonable execution period of AGC algorithm. The
The scheduling and control of power generation are vital simulation results are also provided to demonstrate the
aspects in the daily operation of power systems. Generally, usefulness of the proposed ACE calculation method and AGC
energy management system is provided to control the execution period in this paper.
electrical power output so as to supply ever changing system
load for power as economically as possible. Automatic
generation control (AGC) program operating within energy II. EXISTING ACE CALCULATION METHODS
management system can provide the necessary means to the
system operators. The AGC has been one of major issues in A. Constant frequency bias factor
the operation of local control area systems and interconnected ACE using frequency bias factor is calculated as follows;
power systems for several years.
Four basic objectives of power system operation during ACE = ( Pa − Ps ) + B ( f a − f s )
(1)
normal operating conditions are associated with AGC: 1) = ΔPtie + BΔf
matching total system generation to total system load; 2)
regulating system electrical frequency error to zero; 3) where
distributing system generation among control areas so that net Pa : Actual net power interchange
area tie flows match net area tie flow schedules; 4) Ps : Scheduled power interchange
distributing area generation among area generation sources so fa : Measured frequency
that area operating costs are minimized [1]. For the isolated fs : Nominal frequency
control area, the main function of AGC is to restore the B : Frequency bias factor
ΔPtie : Power interchange deviation
Δf :Frequency deviation
 method.” This method is based on the frequency deviations
Frequency bias factor, B, is determined as follows data which has been observed in many events for several
years [5].
1 Because the frequency bias in the “1% of load method” is
B sys = +D
R eq (2) determined by measured data such as frequency, power and
1
voltage of the past and it is fixed through the year for all load
R eq = levels, it cannot represent the current status of power system
1 / R1 + 1 / R 2 + 1 / R 3 + ⋅ ⋅ ⋅ + 1 / R n
[4].
Figure 1 shows the observed frequency deviations against
where events such as generation loss. The straight line represents
R : Speed regulation parameter or Droop the average frequency response of “1% of load method.” The
D : Load damping constant dots below the line imply a greater frequency response for the
Bsys : System frequency bias factor given events. The dots above the line represent a smaller
Req : System droop frequency response [4]. As seen in the Fig. 1, the frequency
bias determined by “1% of load method” cannot represent the
Since the objective of AGC is to regulate generation of system status correctly.
units, the input of AGC should be the actual ACE. When the
frequency bias of the ACE matches with the actual frequency
bias of the area, the calculated ACE and the actual ACE are
identical [3]. However, it is not practical to consider the
frequency bias factor of (2) as the actual frequency bias,
because the actual frequency bias is non-linearly related to
the operating state of control area, i.e., frequency, load,
generation and voltage [2].
For example, the speed-droop characteristic ideally ranges
from 3 to 5 percent. However, the actual speed-droop
characteristic may thus exhibit incremental regulation ranging
from 2 to 12 percent, depending on the unit output [2]. The
speed-droop characteristic also depends on the type of units,
such as steam unit and hydraulic unit. For example, steam
turbines have a number of control valves, each having
nonlinear flow area versus position characteristic [2]. The Fig. 1. Frequency Excursion vs. Generation Loss [4].
generator response is not always predictable and has time
delay. Load damping characteristic, D, is the relationship
between the changes in actual load due to a change in III. PROPOSED AGC SCHEME
frequency. The relationship cannot be measured and varies
continuously with the system status. Typical values of D are A. ACE Calculation Method
1 to 2 percent [2]. Therefore, the fixed estimated frequency Actual ACE is the error between generation and load of a
bias value such as (2) is not the same as the actual frequency control area. It is different from the estimated ACE which is
bias. obtained through the constant frequency bias factor. Under
If the estimated frequency bias is much higher than the ideal conditions, when the frequency bias factor for estimated
actual frequency bias, then variations in the calculated ACE ACE matches the actual frequency bias of the control area,
can be much larger than variations in the actual ACE. It the estimated ACE and the actual ACE are identical.
causes often over-regulation. The AGC depends on the However, an estimated ACE based on a constant frequency
calculated ACE using constant frequency bias using system bias factor is nonlinearly related to the required change in
frequency, generator regulation and load damping as inputs. generation because the actual frequency bias factor is
Therefore, it is critical to get actual frequency bias value for nonlinearly related to the operating state of the control area,
better AGC performance. i.e., frequency, power and load. Using an estimated ACE
In conclusion, ACE calculation method using constant based on the constant frequency bias factor as the system
frequency bias cannot represent the actual power system error in AGC introduces an undesirable nonlinearity into the
status correctly. control system. Therefore, it is necessary to determine the
B. One-percent of load method actual frequency bias factor based on real time system data
North America Electric Reliability Council (NERC) offers such as electrical outputs of generators and frequency.
several possible ways to calculate frequency bias. The Frequency response of power system is the characteristic
general practice is to establish frequency bias in each control determined by load and generation within the control area. It
area once a year based on the area’s natural regulation varies in response to a significant change in load and power
characteristic (1/R+D) corresponding to the forecasted peak generation balance. In normal operation, events such as the
load of the coming year [4]. It is called as “1% of load loss of a large generator or the load increase may bring about
the imbalance between power source and load of control area. Δ fT (3)
Actual Frequency Bias factor =
The imbalance necessarily causes frequency variation. The n

frequency variation is primarily compensated by inertia of ∑ ΔP

i =1
Gi , T

units. The primary frequency control is accomplished by

turbine governor’s droop characteristic. The turbine where
governors respond proportionally to frequency deviation and ΔfT : Frequency deviation at time T
normally bring the rate of change of frequency to zero. It ΔPGi,T : Power output deviation of unit i at time T
takes about several seconds. The system would be settled in N : The number of generator unit
the new equilibrium state. It is called primary control. T : Time of recovered frequency
Additional control, supplementary control, is required to
restore the frequency to nominal value [2].
The actual frequency bias factor of (3) can be updated
Fig. 2 shows a frequency excursion resulting from a
generator trip. Point A is a pre-disturbance frequency, whenever any events such as generator loss or large load
typically 60Hz. Point C is the maximum excursion point, and disturbance are occurred. Therefore, it is possible to
point B is the new settling frequency after a generator trip. represent the current status of control area.
The frequency decline from point A to point C is held back B. Execution Period of AGC Control
by generators and load of control area. Most of load in An important factor from the AGC point of view is the
control area consists of motor. Motor consumes less energy execution period of AGC control. If the execution period is
when frequency drops. Therefore, the rate of frequency drop short, the AGC control signal for generators may disturb
from point A to C is slowed. The rate of frequency change is generator’s control action. For example, AGC control signal
dependent on the portion of motor load in control area. may require generation power increase when the generation
Generally, it is known that 1% change in frequency causes a
power is decreasing. It may cause the fluctuation of
2% change in load [4]. All generators have some type of
generator output. On the contrary, the long execution period
governor control. The governor action halts the decline and
may degrade the performance of AGC.
brings frequency back to point B. The frequency does not
recover point B to nominal frequency, 60Hz, until the Therefore, it is important to determine a reasonable
supplementary control provides the amount of lost generation execution period of AGC control. However, the execution
[2]. period has been usually determined by a kind of operation
The frequency back to point B from point A after generator experience of power system. For example, the execution
trip is accomplished by droop and load damping period is 8 seconds or 30 seconds according to specific power
characteristics of control area. The frequency of point B is systems [6]. In the case of the Hellenic Interconnected
resulted from the effect of governor speed droop and the System, ACE calculation period is 5 seconds and AGC
frequency sensitivity of load after generator trip in control control period is 30 seconds, because it is considered
area during primary control condition [2]. Therefore, the sufficient to achieve non-interaction with generator dynamics
actual frequency bias factor can be determined by the [6]. For another case, ACE calculation period and AGC
frequency deviation from nominal value and the power output execution period are 2 seconds and 8 seconds, respectively in
change during primary control condition. And the frequency Korea. Those time periods were determined by a kind of
and power output of generator units can be measured operation experience or assumptions such as generator
periodically. dynamics.
This paper provides the basis to determine AGC execution
period. As explained in previous section of Fig. 2, frequency
excursion curve shows a response characteristic of power
system resulting from power system transient event. As
mentioned in previous section, Point T in Fig. 2 represents
the effect of governor speed droop and the frequency
sensitivity of load for power system transient event.
Therefore, the ACG execution period shall be determined as
the time of recovered frequency to improve the frequency
settling time and AGC control performance.

IV. SIMULATIONS
Fig. 2. Frequency Excursion after Power System Transient
A. Simulation Model
Equation (3) represents the actual frequency bias factor In order to test this new AGC scheme based on dynamic
which is determined by the frequency and power output characteristic of frequency response, a simulation model is
measured at the time of recovered frequency during primary used. The simulation model consists of three thermal power
control condition.
units. Figure 3 shows three thermal power units simulation Time(Second)
model. 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20

The simulation has been done with following assumptions; 0 60.1

60
1) Control area is consisted of three power units. 2) Load -0.2
59.9

change has been modeled as a step change. 3) Economic load -0.4 59.8

Actual FB(pu MW/Hz)

Frequency(Hz)
-0.6 59.7
distribution is not considered for supplementary control. That 59.6
-0.8
means three power units share the load change equally. 59.5
-1 59.4

-1.2 59.3
1 59.2
-1.4
R1 59.1
-1.6 59
1 1 1 + sTr 2
Σ 1 + sT g1 1 + sTt1 1 + sTr1 FB Factor Frequency

ΔPC1 1 Fig. 4. Frequency and Frequency Bias Factor Excursion. The actual
ΔPC2 R2 frequency bias factor is determined by (3). And the AGC execution period
is 4 second in this simulation model.
1 1 1 + sTr 4 kp AGC
Σ Σ Calculatio
n As shown in Figures 5, 6, 7 and Table I, when the actual
1 + sTg 2 1 + sTt 2 1 + sTr 3 1 + sTp Block frequency bias factor and the optimal execution period which
ΔPC3 are determined by (3) are applied to AGC scheme, the
1 1 1 + sTr 6 frequency settling time and oscillation characteristics and the
Σ 1 + sTg 3 1 + sTr 5 power output response of units are improved compared with
1 + sTt 3
1 other inaccurate frequency bias factor and execution period
R3 are used in AGC scheme.
Fig. 3. Simulation Model 60.1

In this model, following parameters are used; 60

Δf : Frequency deviation 59.9

Ri : Speed regulation parameter or Droop

59.8
(R1 =R2=R3= 5Hz/pu MW)
Frequency(Hz)

Tg : Governor time constant [sec] 59.7

(Tg1 = 0.59, Tg2 = 0.60, Tg3=0.65) 59.6

Tt : Thermal turbine time constant [sec] 59.5

(Tt1 = 0.4, Tt2=0.5, Tt3=0.3)

Tr : Thermal unit reheat time constant [sec]
59.4

(Tr1 = 8.0, Tr2 = 3.2, Tr3=9.0, Tr4=3.5, Tr5=10.0, Tr6=4.0 ) 59.3

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20

Kp, Tp : Load frequency characteristic constant FB-Cas e 1 FB-Case 2

Tim e(S ec ond)

FB- Cas e 3 FB-Cas e 4 FB-Ca se 5

(Kp = 120Hz/pu MW, Tp = 20sec)

ΔPL : Load change (ΔPL = 0.1 pu MW) Fig. 5. Simulation Result – Frequency Bias Factors. The FB-Case 3 means
actual frequency bias factor. Other simulation cases are for actual frequency
B. Simulation Results bias factor -10%, -20%, +10% and +20%, respectively.
Fig. 4 shows the frequency and the actual frequency bias
factor excursion during primary control condition for the 60.2

simulation model of Fig. 3. As mentioned before, the actual 60.1

frequency bias factor can be determined at the time of 60

recovered frequency. The time of recovered frequency is 59.9

determined by the deviation of frequency with time. In this

Frequency(Hz)

paper, the criterion of the frequency deviation is 0.01Hz. The

59.8

time which meets the criterion is 4.0 second after a transient, 59.7

0.1pu MW load change. And the actual frequency bias factor 59.6

by (3) is -1.25 pu MW/Hz. The AGC execution period is 59.5

determined by the time of recovered frequency, 4.0 second. 59.4

This paper provides various simulation cases to demonstrate

the effectiveness of the proposed AGC scheme. First of all, 59.3
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20

simulation results for several frequency bias factors such as S-Case 1 S-Case 2
Time(Second)

S- Case 3 S-Case 4 S-Case 5

10% and 20% deviation from actual frequency bias factor are
provided in Fig. 5. And then, simulation results for several Fig. 6. Simulation Result – AGC Execution Periods. The optimal execution
AGC execution periods of 2, 3, 4, 5 and 6 second are period of AGC is 4 second, S-Case 3, after a transient in this simulation.
Other simulation cases are for 2, 3, 5 and 6 seconds, respectively.
provides in Fig. 6. Fig. 7 shows the power output of unit no.
1 of the simulation model.
 0.07 value based on the frequency bias factor or 1% of load
method, however, may not represent the real time status of
0.06 power system correctly. The incorrect ACE value is a cause
of over regulation of power system.
0.05
This paper presents the method to decide frequency bias
factor on real time frequency characteristics. In this paper,
Unit #1 Power Output(pu MW)

0.04
the frequency bias factor is determined by the measured
variables such as generator power output change and
0.03
frequency deviation due to disturbance of power system. The
frequency bias factor calculated is used to calculate the ACE
0.02
of power generators. The AGC divides the calculated ACE
among AGC power generators periodically. The period is
0.01
also determined by real time frequency characteristics.
0
This paper also presents the simulation results for the
0 1 2 3 4 5 6 7 8 9 10 11
Time(Second)
12 13 14 15 16 17 18 19 power system model about the proposed AGC scheme. As
S-Case 1 FB-Case 2 S-Case 3 S-Case 4 FB-Case 5
described in the simulation results, when the actual frequency
Fig. 7. Simulation Result – Unit #1 Power Output. The unit #1 power bias factor and the optimal execution period which are
output of S-Case 3, determined by (3), is stabilized with rapidity determined by (3) are applied to AGC scheme, the frequency
compared with other simulation cases. settling time and oscillation characteristics and power output
Table I shows the simulation results of the frequency response of units are improved compared with other
settling time, the accumulated AGC control demand and the inaccurate frequency bias factor and execution period are
power output settling time for several simulation cases. used in AGC scheme.
TABLE I
SIMULATION RESULTS
Simulation Cases
REFERENCES
S- S- S- S- S-
Case 1 Case 2 Case 3 Case 4 Case 5 [1] T. M. Athay, “Generation Scheduling and Control,” Proceedings of the
FB- FB- FB- FB- IEEE, Vol. 75, No. 12, pp. 1592~1606, December 1987.
FB-
Case 1 Case 2 Case 4 Case 5 [2] P. Kundur, “Power System Stability and Control,” McGraw-hill.
Case 3
[3] A.J. Wood and B.F. Wollenberg, “Power Generation, Operation &
Frequency 1.25 1.25 1.25 1.25 1.25 Control,” John Wiley & Sons, 1984.
Bias factor [4] NERC Training Resources Working Group, “Understand and Calculate
(pu MW/Hz) 1.0 1.125 1.25 1.375 1.5
Frequency Response,” unpublished.
AGC [5] Standard BAL-003-0, Frequency Response and Bias, NERC, April,
2.0 3.0 4.0 5.0 6.0
Execution 2005.
Time [6] C. D. Vournas, E. N. Dialynas, N. Hatziargyriou, A. V. Machias, J. L.
4.0 4.0 4.0 4.0 4.0
(sec) Souflis, B. C. Papadias, “A flexible AGC Algorithm for Hellenic
Frequency 18.5 14.9 9.3 20.7 19.7 interconnected system,” IEEE Transactions on Power Systems, vol. 4,
Settling Time pp. 61-68, February 1989.
(sec) 14.5 9.9 9.3 17.4 18.1
Accumulated
1.53 1.16 0.54 1.70 1.61
AGC Control
Demand
0.92 0.56 0.54 1.45 1.60
(pu MW)
Power Output 19.9 16.8 15.9 18.3 23.9
Settling Time
(sec) 15.3 15.4 15.9 16.6 16.7

V. CONCLUSIONS
The scheduling and control of power generation are vital
aspects in the daily operation of power systems. AGC has
been one of the important issues in the operation of local area
and interconnected power systems. AGC provides power
demand signals for AGC power generators to control
frequency and it is usually installed as a part of EMS. In
AGC, some variables of power system such as frequency and
power generation are measured to calculate the ACE. The
ACE value is an important factor of AGC performance.
Generally, ACE value is determined by a fixed frequency bias
factor of power system or 1% of load method. The ACE