Você está na página 1de 11

Electric Power Components and Systems

ISSN: 1532-5008 (Print) 1532-5016 (Online) Journal homepage: http://www.tandfonline.com/loi/uemp20

Load Frequency Control of Power Systems with


Governor Deadband (GDB) Non-linearity

Wen Tan, Shuaibing Chang & Rong Zhou

To cite this article: Wen Tan, Shuaibing Chang & Rong Zhou (2017): Load Frequency Control of
Power Systems with Governor Deadband (GDB) Non-linearity, Electric Power Components and
Systems, DOI: 10.1080/15325008.2017.1348404

To link to this article: http://dx.doi.org/10.1080/15325008.2017.1348404

Published online: 15 Nov 2017.

Submit your article to this journal

View related articles

View Crossmark data

Full Terms & Conditions of access and use can be found at


http://www.tandfonline.com/action/journalInformation?journalCode=uemp20

Download by: [Gothenburg University Library] Date: 16 November 2017, At: 02:51
Electric Power Components and Systems, 0(0):110, 2017
Copyright C Taylor & Francis Group, LLC
ISSN: 1532-5008 print / 1532-5016 online
DOI: 10.1080/15325008.2017.1348404

Load Frequency Control of Power Systems with


Governor Deadband (GDB) Non-linearity
Wen Tan, Shuaibing Chang, and Rong Zhou
School of Control and Computer Engineering, North China Electric Power University, Beijing, China

CONTENTS
Downloaded by [Gothenburg University Library] at 02:51 16 November 2017

AbstractLoad frequency control of power systems with a gover-


1. Introduction nor deadband (GDB) is studied in the paper. The effects of GDB on
2. Preliminaries the linear active disturbance rejection controller (LADRC) are ana-
3. Effect of GDB lyzed first. It is shown that the limit cycle can be avoided by reducing
the controller and/or observer bandwidth. However, the disturbance
4. Anti-GDB Schemes rejection performance will be degraded. To maintain the perfor-
5. Multi-Area Power Systems mance of the original LADRC controller, two anti-GDB schemes
6. Conclusion are proposed. One is based on the observer structure of active distur-
bance rejection control, which utilizes the output of the governor to
Funding reconstruct the input to the system in order to correctly estimate the
References controller states. Another is based on the error between the realistic
output of the governor and its ideal value which is fed back and added
to the output of the original LADRC. Simulation results on a single-
area and a two-area power system with the GDB show that the two
schemes are effective in reducing the limit cycle caused by the GDB.
The observer-based anti-GDB has some limitations in practical
implementation and performance, while the error-compensation-
based scheme is simple to implement and can approximate the
disturbance rejection performance of the original LADRC.

1. INTRODUCTION
In power systems, frequency stability is an important index
of power quality. Any sudden load perturbation can cause the
deviation of tie-line exchanges and the frequency fluctuations.
Therefore, to ensure the power quality, a load frequency con-
trol (LFC) system is required. The goal of LFC is to return the
frequency to its nominal value and minimize the unscheduled
tie-line power flows between interconnected control areas [1].
With the increase in size and complexity of modern power
systems, the system oscillation might propagate into wide area
resulting in a wide-area blackout. So, advanced control meth-
ods were applied in LFC; see [2, 3] for a review of control
methods in LFC. Improved performance might be expected
Keywords: load frequency control, electric power system, governor from the advanced control methods; however, these methods
deadband (GDB), active disturbance rejection control, anti-GDB scheme
Received 20 October 2016; accepted 14 June 2017 require either information on the system states or an efficient
Address correspondence to Wen Tan, School of Control and Computer on-line identifier, and thus, it may be difficult to apply them in
Engineering, North China Electric Power University, Zhuxinzhuang, Dewai, practice.
Beijing 102206, China. E-mail: wtan@ieee.org
Color versions of one or more of the figures in the article can be found online
Thus, attention has been paid to the tuning of proportional-
at www.tandfonline.com/uemp. integral-derivative (PID)-type LFC. Reference [4] proposed a

1
2 Electric Power Components and Systems, Vol. 0 (2017), No. 0

robust PID load frequency controller design method based on Kp Electric system gain (Hz/puMW)
maximum peak resonance. References [5, 6] proposed to tune Tp Electric system time constant (sec)
a PID load frequency controller via an internal model control Tt Turbine time constant (sec)
technique. It was shown that with two tuning parameters, the Tg Governor time constant (sec)
R Speed regulation due to governor action (Hz/puMW)
method can achieve good performance for power systems with
f Incremental frequency deviation (Hz)
non-reheat, reheat, and hydro turbines. Pd Load disturbance (puMW)
Recently, an active disturbance rejection control (ADRC) PG Incremental change in generator output (puMW)
method was applied to the LFC problem [7, 8]. The method XGV Incremental change in governor valve position
aims to reject the disturbance by providing its estimation Gg (s) Transfer function of speed governor
through an extended observer; thus, external disturbance can Gt (s) Transfer function of turbine
be rejected more quickly. Furthermore, just like the PID con- Gp (s) Transfer function of electric system
troller, the scheme is only needed to tune two parameters, so Gc (s) Transfer function of load frequency controller
it is easy to be understood by practical control engineers. The TABLE 1. Nomenclature
Downloaded by [Gothenburg University Library] at 02:51 16 November 2017

idea is also proved effective in deregulated environment [9].


In practice, there are non-linearities in the LFC system.
Typical non-linearities in LFC include governor deadband performance will not degrade too much from the performance
(GDB) and generation rate constraint (GRC). Deadband usu- in the linear region. Anti-GDB schemes are proposed and the
ally occurs with loose gears. There is no response until the compensation effects will be demonstrated with some simula-
deadband engages, and once engaged, there is a hysteresis in tion results.
the response. The rest of the paper is organized as follows: Section 2
GDB will degrade the performance of an LFC system, even gives some preliminaries that are used in this paper. Section 3
cause instability. However, only a few control methods con- discusses the effect of GDB on LFC systems, and Section 4
sidered these non-linearities in LFC. References [1012] ana- presents anti-GDB schemes. Finally, some conclusions are
lyze the effects of GDB on LFC performance in the time and given in Section 5.
frequency domain, and show that there will probably exist All the symbols used in the paper are standard in power
limit cycles if the integral gain of load frequency controller system control and are explained in Table 1.
is high. To overcome the effects of GDB, Ref. [13] adopts the
Lyapunov method to optimize the integral gain and the tie-
2. PRELIMINARIES
line coefficient of a two-area power system with GDB; Refs.
[14, 15] adopt the multi-objective evolution algorithm and the 2.1. Deadband and Describing Function Analysis
PSO algorithm to optimize the PID-type load frequency con- Typical characteristics of deadband are shown in Figure 1,
troller; Ref. [16] optimizes a two-degree-freedom PID param- where k is the slope and D is the deadband width. It is noted
eters via a differential evolution algorithm for power systems that the deadband depends not only on the input x, but also on
with GDB. Reference [17] designs a neural network load its direction x.
frequency controller for the power system with GDB; Ref. Assume that the input x is sufficiently close to a sinusoidal
[18] analyzes the stability of two-area power systems with oscillation, that is,
GDB and proposes a design method to avoid limit cycles.
Reference [19] applies the fuzzy-logic technique to design x A sin 0t (1)
the load frequency controllers for power systems with both
GDB and GRC. Reference [20] designs a fuzzy-logic con-
troller for power systems with both GDB and GRC; Ref. [21]
considers optimizing the PID-type load frequency controllers
for power systems with both GDB and GRC using a genetic
algorithm.
In this paper, LFC for power systems with GDB will be
investigated. Unlike the existing methods that handle GDB
directly in the control design, this paper adopts a compensa-
tion method, i.e., ignores GDB in the control design and then
implements a scheme to compensate it. The advantage of such FIGURE 1. Governor deadband and its inverse: (a) deadband
a method is that if the system is in the non-linearity region, the and (b) deadband inverse.
Tan et al.: Load Frequency Control of Power Systems with Governor Deadband (GDB) Non-linearity 3

where A is the amplitude and 0 is the frequency. The non-


linear function F (x, x) of the deadband can be developed in a
Fourier series as follows:
N2
F (x, x) = F 0 + N1 x + x + (2)
0
Since the backlash non-linearity is symmetric about the
origin, F0 is zero and F (x, x) can be approximated by the fol- FIGURE 2. Structure for the describing function analysis.
lowing linear function:
N2 2.2. ADRC
F (x, x) N1 x + x (3)
0
ADRC was first proposed by Han [22] for rejecting distur-
The typical value for GDB is 0.0006 per unit or 0.06% bance of a non-linear system. The idea is to use an extended
[10], which is the value allowed by the joint AIEE-ASME state observer (ESO) to estimate the disturbance of the sys-
Downloaded by [Gothenburg University Library] at 02:51 16 November 2017

Standards for large steam-turbine speed governors. It is known tem and then try to reject it effectively using a simple con-
that the backlash non-linearity tends to give a continuous sinu- trol law. ADRC assumes that the control engineer has very
soidal oscillation with a natural period of about 2 sec. There- little knowledge of the controlled plant and the external dis-
fore, a backlash of approximately 0.05% is chosen for all the turbance, except the relative order p and the high-frequency
simulations [13]. In this case, the Fourier coefficients for the gain b0 of the plant, i.e., it is assumed that the controlled plant
non-linear function of deadband are as follows: is of the following model:
N1 N2
= 0.8 and = 0.2 (4) y(p) (t ) = b0 u(t ) + f (y(t ), u(t ), d(t )) (7)
k k
So, a linear approximation of the backlash non-linearity of where f(y, u, d) is the combination of the unknown internal
width D = 0.05% is as follows: dynamics of the system and the external disturbance, and is
0.2 assumed to be unknown and is denoted as the generalized
F (x, x) = 0.8x x (5) disturbance.

Let
This linear approximation is used by most of the existing 
literature in the load frequency controller design [1321]. z = Aoz + Bou + Eoh
(8)
For a more accurate stability analysis, a describing func- y = Coz
tion analysis should be used instead of using just the linear denote the state-space model of (7), where
approximation (5). The describing function of deadband with
width D is as follows [12, 13]: 0 1 0 0
0 0 1 0

N (A) = Nr + jNi (6)
Ao = ... ... ... . . . ...

where 0 0 0 1
  
k 2D 0 0 0 0
Nr = + arcsin 1 (p+1)(p+1)
2 A  T
     Bo = 0 0 b0 0 (p+1)1
2D D D 1/2  T
+ 2 1 1 Eo = 0 0 0 1 (p+1)1
A A A
   
4k D D Co = 1 0 0 0 1(p+1) (9)
Ni = 1 , AD
A A
z = [z1 z2  zp z p + 1 ]T is the state vector of the system,
To analyze the stability of LFC systems with GDB, we sep-
z1 := y, z2 = y, . . . , z p = y(p1) , z p+1 = f (y, u, d )
arate the non-linearity from the rest of the system and put the
(10)
system into the structure, as shown in Figure 2.
and h(t ) := f (t ).
The Nyquist stability criterion can then be used to deter-
An ESO can be used to estimate the states of the plant and
mine the stability of the system with non-linearity. If Gn (j)
the generalized disturbance:
does not encircle 1/N(A), the system will be stable; if Gn (j)

encircles 1/N(A), the system will be unstable; if Gn (j) z = Aoz + Bou + Lo (y y)
(11)
intersects with 1/N(A), the system will exhibit a limit cycle. y = Coz
4 Electric Power Components and Systems, Vol. 0 (2017), No. 0

FIGURE 4. Single-area power system with governor


deadband.
FIGURE 3. Structure of LADRC.

r Gg (s) is the governor dynamics: Gg (s) = 1


where Lo is the observer gain Tg s+1
r Gt (s) is the turbine dynamics: Gt (s) = 1
Tt s+1
Lo = [1 2 p p+1 ]T (12) r Gp (s) is the power system dynamics: G p (s) = Kp
Tp s+1
Downloaded by [Gothenburg University Library] at 02:51 16 November 2017

If Lo is chosen such that Ao Lo Co is stable, then the states When there is no GDB, the LFC system can be controlled
z1 (t ), . . . , z p (t ) will approximate the output y(t) and its deriva- by a third-order LADRC and the performance can be tuned
tives y(t ), y(t ), . . . , y(p1) (t ), and z p+1 (t ) will approximate the via the controller bandwidth c and the observer bandwidth
generalized disturbance f(y, u, d). o [7, 8]. However, with the GDB, the performance of
With the states of the plant and the generalized disturbance LFC might be limited, as demonstrated in the following
estimated, the state-feedback control law can be chosen as fol- example.
lows:
k1 (r(t ) z1 (t )) + + k p (r(p1) (t ) z p (t )) z p+1 (t ) Example 1. Consider a single-area power system with the
u(t ) = following parameters [6]:
b b
=: Ko (r(t ) z(t )) (13)
Kp = 120, Tp = 20, Tt = 0.3, Tg = 0.08, R = 2.4
where r(t ) is a generalized reference signal composed of the (17)
reference signal r(t) and its derivatives

r(t ) = [r(t ) r(t ) r(p1) (t ) 0]T (14) According to [8], a third-order LADRC is used for load fre-
quency control and the parameters of the third-order LADRC
and can be chosen as follows:
Ko = [k1 k2 kp 1]/b0 (15) b0 = 250, c = 3, o = 30 (18)
is the state-feedback gain.
When there is no GDB, the response of the LFC system is
In summary, a pth-order linear active disturbance rejection
shown in Figure 5 for a step load Pd = 0.01 at t = 10. It is
controller (LADRC) has the following state-space form:

z = Aoz + Bou + Lo (y Coz)
(16)
u = Ko (r z)
Its structure is shown in Figure 3. It is obvious that LADRC
has an observer-based state-feedback control structure, and
has two sets of gains to tune: Lo , the observer gain for ESO,
and Ko , the controller gain. For simplicity, the tuning of these
two gains is reduced to two tuning parameters: c , the con-
troller bandwidth, and o , the observer bandwidth. Details can
be found in [23].

3. EFFECT OF GDB
LFC for a single-area power system with GDB is shown in
Figure 4, where Gc (s) is the load frequency controller and 1/R FIGURE 5. Responses of the single-area power system under
represents the droop characteristics. LADRC.
Tan et al.: Load Frequency Control of Power Systems with Governor Deadband (GDB) Non-linearity 5
Downloaded by [Gothenburg University Library] at 02:51 16 November 2017

FIGURE 6. Describing function analysis for the single-area


power system with deadband.

shown that the frequency deviation returned to zero quickly as


desired.
Now suppose there is a deadband in the governor (k = 1,
D = 0.05%), then there is a limit cycle in the responses of
LFC system as shown in the solid line in Figure 5.
To analyze the stability of LFC systems with GDB, the
closed-loop system can be transferred to the structure in
Figure 2. For the single-area power system with GDB shown
in Figure 4, the linear dynamics in Figure 2 can be computed
as follows:
 
1 FIGURE 7. Analysis and simulation of LADRC with differ-
Gn (s) = G p (s)Gt (s)Gg (s) Gc (s) + (19) ent controller bandwidths. (a) Describing function analysis.
R
(b) Respones of the single-area power system.
Figure 6 shows the Nyquist plot of linear dynamics (19) for
the LADRC with parameters (18). Also shown are the trajec-
tories of 1/N(A) for GDB as A increases. It is clear that GDB region, either the controller bandwidth is small or the observer
will lead to a limit cycle as Gn (j) intersects with 1/N(A). bandwidth is small, and the disturbance rejection performance
To overcome GDB, the parameters of LADRC can be of the LFC system controlled by LADRC is not acceptable.
re-tuned to avoid entering into the non-linear region. From So, we consider anti-GDB schemes to overcome the adverse
the viewpoint of the describing function analysis, the re- effect of GDB while retaining the performance of LADRC in
tuning should not make the Nyquist plot of Gn (s) to encircle the next section.
1/N(A). Figure 7(a) shows the Nyquist plot of Gn (s) when
the controller bandwidth c = 2 and c = 1 with the observer
4. ANTI-GDB SCHEMES
bandwidth o = 30 fixed. It is shown that when c = 1, the
system will not have a limit cycle since it does not encircle Most of the existing methods consider GDB directly in the
1/N(A). This can be verified by the responses of the LFC control design by approximating it as a linear block expressed
system as shown in Figure 7(b). It is observed that the limit in (5) and trying to find the optimal controller settings under
cycle due to GDB can be avoided by decreasing the controller GDB. While this kind of methods may handle GDB success-
bandwidth of LADRC. However, the performance of LFC is fully, the performance may be conservative as illustrated in
also degraded and the load disturbance cannot be be rejected the previous section for LADRC. In this section, a compen-
quickly for c = 1. sation method will be proposed. The non-linearity is ignored
In fact, for the system in Example 1, the stability region of first in the control design and an anti-windup scheme is then
the parameters of LADRC is shown in Figure 8. In the stability implemented to compensate the non-linearity. The advantage
6 Electric Power Components and Systems, Vol. 0 (2017), No. 0

1. The deadband is inside the governor, and its output


cannot be measured directly as for the actuator output.
2. The output of the primary load frequency controller
1
R
goes through the deadband. The output of the gov-
ernor is the superposition of the outputs of the sec-
ondary load frequency controller and the primary load
frequency controller.
3. Since the deadband is non-linear, the superposition
principle does not hold, so it is difficult to reconstruct
the input from the output of a non-linear unit.
In this paper, an anti-GDB scheme shown in Figure 10 is
adopted, where the realistic input to the LFC system is recon-
structed as follows:
Downloaded by [Gothenburg University Library] at 02:51 16 November 2017

FIGURE 8. Stability region of LADRC.


1
u = Gg1 (s)XGV + f (21)
of such a method is that if the system is in the non-linear R
region, the performance will not degrade too much from the
where Gg1 (s) is the inverse of the governor dynamics. In
performance in the linear region.
order to reconstruct Pc accurately, the signal is used in ESO
To compensate the effects of GDB, two anti-GDB schemes
after passing through the inverse of the deadband as shown in
are proposed for LADRC.
Figure 1(b). Since Gg (s) is a simple dynamics with inertia, its
4.1. Observer-based Anti-GDB Scheme inverse amounts to a differentiator

It is noted that LADRC is an observer-based state-feedback Tgs + 1


Gg1 (s) = (22)
controller. Reference [24] proposed an anti-windup scheme Tg /s + 1
for LADRC with actuator non-linearity, which takes the fol-
lowing form: where is the differential coefficient, chosen between 5
and 10.
z = Aoz + Bou + Lo (y Coz)
u = Ko (r z) (20)
4.2. Error-Compensation-based Anti-GDB Scheme
u = N (u)
The observer-based anti-GDB scheme requires the parame-
where N() is the actuator non-linearity. The anti-windup ters of the deadband and its inverse. If the governor dynamics
scheme is shown in Figure 9. Compared with Figure 3, the is complex or the deadband is not exactly known, the scheme
realistic output u is used in ESO instead of the ideal output might lead to a steady-state error. To overcome this problem,
u of LADRC, which makes use of the correct information in another anti-GDB scheme is proposed as shown in Figure 11.
estimating the states; thus, it can overcome the effect due to This scheme makes use of the error between the realistic out-
the non-linearity of the actuator [24]. put of the governor and its ideal value and feeds it back to the
For the LFC system controlled by LADRC, to diminish the output of the original LADRC. It takes effect until it is zero.
effect of GDB, a similar idea can be adopted. However, the From Figure 11, it is found that
structure in Figure 9 cannot be directly utilized in the LFC
system due to the following reasons: [1 + kc (GgN Gg )]G p
f = Pd
[1 + kc (GgN Gg )] + G p Gt Gg N (Gc + 1/R)
(23)
where Gc denotes the transfer function of the load frequency
controller and N denotes the deadband non-linearity. If kc is
chosen as the inverse of Gg (s), then

NG p
f = Pd
N + G p Gt GgN (Gc + 1/R)
G p
FIGURE 9. Observer-based anti-windup scheme for Pd (24)
LADRC. 1 + G p Gt Gg (Gc + 1/R)
Tan et al.: Load Frequency Control of Power Systems with Governor Deadband (GDB) Non-linearity 7
Downloaded by [Gothenburg University Library] at 02:51 16 November 2017

FIGURE 10. Observer-based anti-GDB scheme for LADRC.

FIGURE 11. Error-compensation-based anti-GDB scheme for LFC.

if the deadband N is close to linear. It is exactly the transfer


function from Pd to f when there is no deadband. So ide-
ally, the anti-GDB scheme can restore the performance of the
original load frequency controller Gc (s) if N is linear. In prac-
tice, the inverse of Gg (s) can be chosen as (22). Due to the fact
that the deadband is non-linear, the scheme cannot completely
remove the effect of GDB; however, it can indeed diminish the
limit cycle. The aforementioned discussion can be illustrated
in the following example.

Example 2. Consider the single-area power system in


Example 1 and LADRC with parameter (18). Suppose there
is GDB of magnitude D = 0.05%. The responses of the
LFC system for a step load Pd = 0.01 at t = 10 with anti-
FIGURE 12. Responses of LFC system under LADRC with
GDB schemes are shown in Figure 12. Here, the differential the anti-GDB scheme.
coefficient is chosen as 10. It is observed that both schemes
8 Electric Power Components and Systems, Vol. 0 (2017), No. 0
Downloaded by [Gothenburg University Library] at 02:51 16 November 2017

FIGURE 13. A two-area power system with GDB.

can diminish the amplitude of the limit cycle. The response as in the single-area power systems. The local load frequency
of the system with the error-compensation-based anti-GDB controller takes the form:
scheme is almost the same as that without GDB, while the
response of the system with observer-based anti-GDB scheme ui = Gci (s)Bi  fi (27)
is slightly sluggish due to the reconstruction.
Example 3. Consider a two-area power system with GDB
shown in Figure 13. The two areas are assumed to be the
5. MULTI-AREA POWER SYSTEMS same with reheated turbines, and the parameters are as fol-
lows [19]:
For multi-area power systems, load frequency controllers
need to maintain the scheduled tie-line exchange power flow 1 120
Ggi = , G pi = ,
besides the frequency, so a composite variable, area control 0.2s + 1 20s + 1
error (ACE), is used as the feedback variable in multi-area 1 20/3s + 1
Gti =
power systems instead of the frequency in single-area power 0.3s + 1 20s + 1
systems. For Area #i, ACE is defined as follows: Ri = 2.4, Bi = 0.425, T12 = 0.0707 (i = 1, 2) (28)

ACEi = Ptie i + Bi  fi (25) with typical GDB 0.05%.


where Bi is the frequency bias coefficient. The feedback con- Decentralized LADRC can be tuned for the two-area
trol law in Area #i is then system:

ui = Gci (s)ACEi (26) b0 = 14.17; c = 2; o = 20 (29)

Reference [6] proposed a decentralized controller design The responses of the two-area system under LADRC with
method for multi-area power systems by ignoring the tie-line GDB for a step load Pd = 0.01 at t = 1 are shown in
power flow, i.e., assuming that Ptie i = 0(i = 1, . . . , n), then Figure 14. It is observed that the system has an acceptable
the load frequency controller for each area can be designed dynamic performance but there is a limit cycle after 70 sec. To
Tan et al.: Load Frequency Control of Power Systems with Governor Deadband (GDB) Non-linearity 9

errors in the reconstructed signal and the errors are ampli-


fied in the tie-line. The error-compensation-based anti-GDB
scheme can achieve better performance.

6. CONCLUSION
In this paper, the effect of GDB on the LFC system was
first analyzed. It was shown that the LFC system controlled
by LADRC can avoid entering the limit cycle by decreas-
ing the controller bandwidth and/or observer bandwidth of
LADRC, which will degrade the disturbance rejection perfor-
mance. To maintain the disturbance rejection performance of
the original LADRC, two anti-GDB schemes were proposed,
Downloaded by [Gothenburg University Library] at 02:51 16 November 2017

i.e., observer-based and error-compensation-based. Simula-


tion results for a single-area and a two-area power system with
GDB showed that the two anti-GDB schemes are effective in
overcoming the adverse effects of GDB in that the perfor-
mances are close to that of the original LADRC. There are
some limitations in the observer-based anti-GDB scheme and
the error-compensation-based anti-GDB scheme is better in
performance and implementation.

FUNDING
This work was supported by the National Natural Science
Foundation of China under Grant 61573138.

REFERENCES
[1] Tan, W., Load frequency control: problems and solu-
tions, Proceedings of the 30th Chinese Control Conference,
pp. 62816286, Yantai, China, 2011.
[2] Ibrabeem, P. K., and Kothari, D. P., Recent philosophies
of automatic generation control strategies in power systems,
IEEE Trans. Power Syst., Vol. 20, No. 1, pp. 346357, 2005.
[3] Shayeghi, H., Shayanfar, H. A., and Jalili, A., Load frequency
control strategies: A state-of-the-art survey for the researcher,
Energy Convers. Manage., Vol. 50, No. 2, pp. 344353, 2009.
[4] Khodabakhshian, A., and Edrisi, M., A new robust PID load
frequency controller, Control Eng. Pract., Vol. 16, pp. 1069
1080, 2008.
[5] Tan, W., Tuning of PID load frequency controller for power
FIGURE 14. (a) Frequency deviation of Area 1. (b) Fre- systems, Energy Convers. Manage., Vol. 50, No. 6, pp. 1465
quency deviation of Area 2. (c) Tie-line power flow. 1472, 2009.
[6] Tan, W., Unified tuning of PID load frequency controller for
power systems via IMC, IEEE Trans. Power Syst., Vol. 25,
overcome the limit cycle, the two proposed anti-GDB schemes No. 1, pp. 341350, 2010.
are adopted, where = 10. It is clear that both anti-GDB [7] Dong, L., Zhang, Y., and Gao, Z., A robust decentral-
ized load frequency controller for interconnected power sys-
schemes have diminished the limit cycle; however, the tie-
tems, ISA Trans., Vol. 51, No. 3, pp. 410419, 2012.
line exchange power flow exhibits a minor oscillation for the doi:10.1016/j.isatra.2012.02.004
observer-based anti-GDB scheme, which is caused by the fact [8] Tan, W., Zhou, H., and Fu, C., Linear active disturbance rejec-
that the deadband inverse is not ideal; thus, there are some tion control for load frequency control of power systems (in
10 Electric Power Components and Systems, Vol. 0 (2017), No. 0

Chinese), Control Theory and Applications, Vol. 30, No. 12, interconnected power system, Energy Convers. Manage., Vol.
pp. 16071615, 2013. 52, pp. 22472255, 2011.
[9] Tan, W., Hao, Y., and Li, D., Load frequency control in [22] Han, J., From PID to active disturbance rejection control,
deregulated environments via active disturbance rejection, Int. IEEE Trans. Ind. Electron., Vol. 56, No. 3, pp. 900906, 2009.
J. Elect. Power Energy Syst., Vol. 66, No. 3, pp. 166177, doi:10.1109/TIE.2008.2011621
2015. [23] Gao, Z., Active disturbance rejection control: A paradigm
[10] Concordia, C., Kirchmayer, L. K., and Szymanski, E. A., shift in feedback control system design, Proceedings of Amer-
Effect of speed-governor dead band on tie-line power and fre- ican Control Conference, pp. 23992405, Minneapolis, MN,
quency control performance, IEEE Trans. Power Apparatus 2006.
Syst., Vol. PAS-76, No. 3, pp. 429434, 1957. [24] Zhou, H., and Tan, W., Aanti-windup schemes for linear active
[11] Taylor, C. W., Lee, K. Y., and Dave, D. P., Automatic gener- disturbance rejection control (in Chinese), Control Theory
ation contro analysis with governor deadband effects, IEEE Appl., Vol. 31, No. 11, pp. 14571463, 2014.
Trans. Power Apparatus Syst., Vol. PAS-98, No. 6, pp. 2030
2036, 1979.
[12] Wu, F. F., and Dea, V. S., Describing-function analysis of BIOGRAPHIES
Downloaded by [Gothenburg University Library] at 02:51 16 November 2017

automatic generation control system with governor deadband,


Electr. Power Syst. Res., Vol. 1, No. 1, pp. 113116, 1977/1978. Wen Tan received the B.Sc. degree in applied mathematics
[13] Tripathy, S. C., Hope, G. S., and Malik, O. P., Optimisation and the M.Sc. degree in systems science from the Xiamen
of load-frequency control parameters for power systems with University, Xiamen, China, and the Ph.D. degree in automa-
reheat steam turbines and governor deadband nonlinearities,
tion from the South China University of Technology, China,
IEE Proc Gen. Trans. Distrib., Vol. 129, No. 1, pp. 1016,
1982. in 1990, 1993, and 1996, respectively. He joined the faculty
[14] Ganapathy, S., and Velusami, S., Decentralized load fre- of the Power Engineering Department, North China Electric
quency control of interconnected power systems with SMES Power University, Beijing, China, in 1996. He was a Postdoc-
units and governor dead band using multi-objective evolution toral Fellow in the Department of Electrical and Computer
algorithm, J. Elect. Eng. Technol., Vol. 4, No. 4, pp. 443450,
Engineering, University of Alberta, Edmonton, Canada, from
2009.
[15] Gozde, H., and Taplamacioglu, M. C., Automatic genera- 2000 to 2001. He is currently a Professor with the School
tion control application with craziness based particle swarm of Control and Computer Engineering, North China Electric
optimization in a thermal power system, Int. J. Elect. Power Power University. His research interests include modeling,
Energy Syst., Vol. 33, No. 1, pp. 816, 2011. analysis, and control of complex industrial processes.
[16] Sahu, R. K., Panda, S., and Rout, U. K., DE optimized par-
allel 2-DOF PID control for load frequency control of power
Shuaibing Chang received the B.E. and M.E. degrees from
system with governor dead-band nonlinearity, Int. J. Elect.
Power Energy Syst., Vol. 49, No. 1, pp. 1933, 2013. the Hebei University of Science and Technology, China, in
[17] Demiroren, A., Sengor, N. S., and Zeynelgil, H. L., Automatic 2009 and 2012, respectively. He is currently working toward
generation control by using ANN technique, Electr. Power the Ph.D. degree at the School of Control and Computer Engi-
Compon. Syst., Vol. 29, No. 10, pp. 883896, 2001. neering, North China Electric Power University. His research
[18] Tsay, T.-S., Load-frequency control of interconnected power
interest includes load frequency control of power systems.
system with governor backlash nonlinearities, Int. J. Elect.
Power Energy Syst., Vol. 33, No. 9, pp. 15421549, 2011.
[19] El-Sherbiny, M. K., El-Saady, G., and Yousef, A. M., Efficient Rong Zhou received the B.E. and M.E. degrees from the
fuzzy logic load-frequency controller, Energy Convers. Man- North China Electric Power University, China, and the Ph.D.
age., Vol. 43, pp. 18531863, 2002. degree from the Chinese Academy of Science, China, in
[20] Anand, B., and Jeyakumar, A. E., Fuzzy logic based load fre- 1999, 2003, and 2015, respectively. She is currently a Lec-
quency control of hydro-thermal system with non-linearities,
turer with the School of Control and Computer Engineering,
Int. J. Elect. Power Eng., Vol. 3, No. 2, pp. 112118, 2009.
[21] Golpira, H., Bevrani, H., and Golpira, H., Application of GA North China Electric Power University. Her research interests
optimization for automatic generation control design in an include computer applications in power systems.

Você também pode gostar