Control Engineering Practice 61 (2017) 81–92

Contents lists available at ScienceDirect

Control Engineering Practice

journal homepage: www.elsevier.com/locate/conengprac

Non-linear sliding mode load frequency control in multi-area power system MARK
⁎
Sheetla Prasad , Shubhi Purwar, Nand Kishor
Department of Electrical Engineering Motilal Nehru National Institute of Technology, Allahabad, India

A R T I C L E I N F O A BS T RAC T

Keywords: This paper addresses non-linear sliding mode controller (SMC) with matched and unmatched uncertainties for
Load frequency control load frequency control (LFC) application in three-area interconnected power system. In conventional LFC
Three-area power system scheme, as the nominal operating point varies due to system uncertainties, frequency deviations cannot be
Non-linear sliding mode minimized. These lead to degradation in the dynamic performance or even system instability. In this paper, an
Matched and unmatched uncertainties
effective control law is proposed against matched and unmatched uncertainties.. The proposed controller has
Performance index
Governor dead band and generation rate
ability to vary closed-loop system damping characteristics according to uncertainties and load disturbances
constraints present in the system. The frequency deviation converges to zero with minimum undershoot/overshoot, fast
settling time, significantly reduced chattering and ensures asymptotic stability. In addition, the controller is
robust in the presence of parameter uncertainties and different disturbance patterns. It also guarantees high
dynamic performance in the presence of governor dead band (GDB) and generation rate constraint (GRC).
Simulations are performed to compare the proposed controller with linear SMC. Using proposed control
strategy, undershoot/overshoot and settling time gets reduced by approximately 30% with respect to linear
SMC. The computed performance indices and qualitative results establish the superiority as well as applicability
of the proposed design for the LFC problem. Further, the proposed controller scheme is validated on IEEE 39
bus large power system.

1. Introduction
Load frequency control (LFC) is one of the important issues in
multi-area power systems. The basic objective of the LFC is balanced
generation and load demand, such that frequency deviation and tie-line
power deviation converges to zero in different control areas defined in a
multi-area power system (Bevrani, 2014; Kundur, 1994). Frequency
control during load and generation variation in any area is an
important operational aspect in a large interconnected power system
(Camblong, Vechiu, Etxeberria, & Martínez, 2014; Yinsong, Shizhe,
Jingyu, & Zheng, 2016). Conventionally, LFC uses an integral
controller. It is well known that a high integral gain may deteriorate
the system performance, causing large oscillations and instability.
Thus, the integral gain must be set at a level so as to provide a
compromise between a desirable transient recovery and low overshoot
in the dynamic response of the overall system. The methods to tune the
gain of the integral controller have been reported in (Ibraheem,
Kumar, & Kothari, 2005). In general, the design approach for load
frequency controller is employed on the basis of linearized model with
fixed PI parameters (Bevrani & Hiyama, 2005). However, classically
tuned PI control strategy results have longer settling time and relatively
large overshoots in transient response. Besides, such PI control
algorithm provides desired response of the system only in the vicinity
of the designed operating point. In other words, these cannot perform
over wide range of operating conditions of load changes/disturbances
and parameter uncertainties in the multi-area power system.
Recently, several control design approaches have been reported.
The application of advanced control methods in both single area and
multi-area systems has been found widely during the literature survey
such as in (Bevrani, Mitani, & Tsuji, 2004; Ersdal, Imsland, Uhlen,
Fabozzi, & Thornhill, 2016; Rerkpreedapong, Hasanovic, & Feliachi,
2003; Zribi, Al-Rashed, & Alrifai, 2005). An optimal fractional order
PID controller was designed and tuned using genetic algorithm for LFC
(Ismayil, Ramdas Sreerama, & Thiruthimana Krishnan, 2015). The
LFC for single area power system has been reported in (Saxena, Hote
& Yogesh, 2013) using internal model control and model-order
reduction. A direct-indirect adaptive fuzzy controller was developed
in (Yousef, AL-Kharusi, Albadi, & Hosseinzadeh, 2014) for multi-area
LFC scheme. A H∞ performance criterion was used to minimize the
effect of disturbances. A cooperative control technique was formulated
in (Chen, Ye, Wang, & Lu, 2015) to allocate the regulation burden
among the control areas.
Due to increase in size and complexity of modern power systems,
oscillations are observed when PI control strategy is implemented.

⁎
Corresponding author.
E-mail address: sheet296@gmail.com (S. Prasad).

http://dx.doi.org/10.1016/j.conengprac.2017.02.001
Received 16 August 2016; Received in revised form 25 November 2016; Accepted 1 February 2017
0967-0661/ © 2017 Elsevier Ltd. All rights reserved.
S. Prasad et al. Control Engineering Practice 61 (2017) 81–92

These system oscillations might propagate into wide area resulting into
blackout. So, advanced control methods; optimal control (Fosha,
Elgerd); variable structure control (Zribi et al., 2005), (Goshaidas,
Sitansu, & Bhattacharyya, 2004); robust control (Bevrani, 2014),
(Bevrani et al., 2004), (Lim, Wang, & Zhou, 1996); and adaptive
control (Khooban & Niknam, 2015; Zribi et al., 2005) have been
applied. The authors (Barisal, 2015) proposed a teaching learning
based optimization to tune the parameters of Integral and PID
controller in hybrid system following a step change in load. The
application of fuzzy PI controllers for LFC is suggested in (Hassan,
2015; Tarkeshwar & Mukherjee, 2015). An adaptive neuro-fuzzy
inference system approach for automatic generation control in three
area hydro-thermal power system is given in (Prakasha & Sinhab,
2015).
An efficient approach for LFC was formulated and investigated in
deregulated environments (Chandra Sekhara, Sahua, Baliarsinghb, &
Pandaa, 2016). The decentralized LFC scheme is more practical than
the centralized one because it only uses the local area state information
to attenuate the frequency deviation (Goshaidas et al., 2004), (Lim
et al., 1996), (Tarek Hassan Mohamed et al., 2012). A variable
structure control based LFC was formulated in (Goshaidas et al.,
2004) in the presence of parameter matched uncertainty. A decentra-
lized LFC scheme was designed using the model predictive control
technique for an interconnected power system concerning wind
turbines in (Tarek Hassan Mohamed et al., 2012).
The operating characteristic of generating unit changes with time,
therefore, parameter uncertainties is an important issue in the
controller design (Siaramakrishana, Hariharm, & Srisailam, 1984).
Therefore, the designed controller may be suitable for a specific
operating point but may not be effective under parameter uncertain-
ties. This necessitates for verification in robustness of designed LFC
against the parameter changes. The objective of load frequency control
(LFC) action in an area is to compensate against these variations.
Several authors (Kundur, 1994), (Goshaidas et al., 2004),
(Khodabakhshian & Edrisi, 2008), (Hsu, 1998), (Mi, Fu, Wang, &
Wang, 2013) have applied variables structure theory for the design of
LFC.
Sliding mode control (SMC) is a form of variable structure control.
Mi et al. Mi et al. (2013) has presented decentralized LFC design using
SMC strategy for solving matching and unmatched parameter uncer-
tainties. Vrdoljak, Peric, and Petrovic (2010) has proposed a discrete-
time sliding mode controller for LFC. The application of SMC for LFC
problem has been addressed in references (Mi et al., 2013; Vrdoljak
et al., 2010). The non-linear sliding mode control is applied on ship-roll
stabilization problem (Fulwani, Bandyopadhyay, & Fridman, 2011)
ing to present uncertainties and load disturbances in the system. The
effectiveness of the proposed controller is evaluated with MATLAB® in
Section 5. The proposed; non-linear SMC for LFC (NLSMLFC) is
compared with linear SMC design strategy for the same system
parameters and various scenarios of three area interconnected power
system as reported in (Mi et al., 2013). Mi et al. (2013) has not
considered linear SMC performances against random load disturbances
and GDB. Further, in this study, the robustness of NLSMLFC is
compared with linear SMC (Mi et al., 2013) against random load
disturbance. The results have been compared and found better for GRC
and GDB conditions of the generating unit. The performance indices
are also calculated and compared with linear SMC. The designed
control signal provides significantly reduced chattering effect signifying
reduced wear-out of actuators (valve) in steam turbine. The overshoot/
undershoot and settling time gets reduced by approximately 30% with
respect to linear SMC (Mi et al., 2013). Finally, to demonstrate the
validation and effectiveness of the proposed NLSMLFC, IEEE 39 bus
large power system (Bevrani, 2014) is considered and successfully
tested in Scenario 6 of Section 5. The frequency oscillations are reduced
to zero and a significant reduction in chattering in the control signal is
found with NLSMLFC that enhance the system stability. The results
illustrate that the proposed scheme can effectively enhance the
frequency stability by improving the damping of the system.
2. LFC scheme for multi-area interconnected power system
2.1. LFC scheme for i-area
For an interconnected system, in addition to primary speed control
loop, supplementary control action in secondary loop is
required to schedule the generation. As such, the power network
although being a complex non-linear system, can be considered
linearized for the study of LFC problem. The linearized model around
the operating point of three area power system is referred from (Lim
et al., 1996), (Mi et al., 2013). The study model of the ith area of power
system is shown in Fig. 1. The generator in each area is assumed to be
equipped with non-reheat turbine. The equations that govern the
dynamics of ith area are given as (Mi et al., 2013):
1 Kp Kp
Δfi ̇ (t ) = − Δf (t ) + i
ΔPp (t ) − i
ΔPdi(t )
Tp i Tp i Tp
i i i
Kp N
− i
∑ Ksij{Δδi(t ) − Δδj (t )}
2πTp (1)
i j =1, j ≠ i

and frequency regulation in power system (Prasad, Purwar, & Kishor,

2016) to improve the system dynamic performance in terms of low 1 1
ΔPġ (t ) = − ΔPg (t ) + ΔX g (t )
overshoot and reduced settling time. i Tti i Tti i
(2)
The conventional LFC controllers (Lim et al., 1996; Zribi et al.,
2005) are designed with fixed parameters and operated only around −1 1 1 1
the nominal point. Mi Yang (Mi et al., 2013) claimed that SMLFC has ΔX ̇g (t ) = Δf (t ) − ΔX g (t ) − ΔEi(t ) + ui(t )
i RiTGi i TGi i TGi TGi (3)
improved steady response only in the presence of system matched and
unmatched uncertainties, step load disturbance and non-linearities
⎡ N ⎤
such as generation rate constraint (GRC). This paper addresses the 1
ΔEi̇ (t ) = K Ei⎢K BiΔfi (t ) + ∑ Ksij{Δδi(t ) − Δδj (t )} ⎥
effect of the matched and unmatched uncertainties, different load ⎢⎣ 2π ⎥⎦
j =1, j ≠ i (4)
disturbance patterns and non-linearities such as GRC, governor dead
band (GDB) on LFC model for multi-area power system as described in
Δδi̇ (t ) = 2πΔfi (t ) (5)
Section 2. The paper objective of this paper is to achieve both minimum
overshoot/undershoot and reduced settling time simultaneously with
where, i = 1, .... , N is the number of areas. Eqs. (1)–(5) can be
significant reduction in chattering.
represented in state space form as:
The non-linear switching surface is described in Section 3, followed
by selection of non-linear function and stability of NLSMLFC under N

both matched and unmatched uncertainties in multi-area power xi̇ (t ) = Ai xi(t ) + Biui(t ) + ∑ Eijxj(t ) + FΔP
i d i (t )
j =1
system. Here, an effective control law is designed for matched as well j≠i (6)
as unmatched uncertainties in Section 4. Thus, the proposed controller
has ability to vary closed-loop system damping characteristics accord- where,

82
S. Prasad et al. Control Engineering Practice 61 (2017) 81–92

Fig. 1. Block diagram of the ith area of a multi-area power system (Lim et al., 1996).

⎡ 1 Kp Kp ⎤ ⎡ z1̇ (t ) ⎤ ⎡ A11 A12 ⎤⎡ z1 (t ) ⎤ ⎡ 0 ⎤ N ⎡ E11 E12 ⎤⎡ z1 (t ) ⎤

⎢ − Tp
i 0 0 − 2πTi ∑ Nj =1 Ksij ⎥ ⎢ i ⎥=⎢ i i
⎥⎢ i ⎥ + ⎢ ⎥u ( t ) + ∑ ⎢ ij ij
⎥⎢ j ⎥
T
⎢ ⎥ ⎢⎣ E21ij E22ij ⎥⎦⎢⎣ z2j (t )⎥⎦
p
⎢⎣ z2̇ i(t )⎥⎦ ⎢⎣ A21i A22i ⎥⎦⎢⎣ z2i(t )⎥⎦ ⎣ B2i ⎦ i
i pi i
j≠i j =1, i ≠ j
⎢ 1 1
⎥
⎢ 0 −T
Tti
0 0 ⎥ ⎡ F1 ⎤
⎢ ti
⎥ + ⎢ i ⎥ΔPdi(t )
Ai = ⎢− 1
0 −T −T
1 1
0 ⎥, ⎢⎣ F2i ⎥⎦ (7a)
⎢ i Gi
R T Gi Gi ⎥
⎢ K Ei ⎥
⎢ K EiK Bi 0 0 0 2π
∑ j =1 sij ⎥
N K
where,
⎢ j≠i ⎥
⎢⎣ ⎥⎦ T
2π 0 0 0 0 z1j = [0 0 −Δδj 0 ] , z2j = [0]T (7b)
⎡ Kp ⎤
⎢0 0 0 0 2πTi Ksij ⎥
⎢ ⎥
pi
Matrix sub-blocks in Eq. (7) are computed as:
⎢0 0 0 0 0 ⎥
Eij = ⎢ 0 0 ⎥
0 0 0 ⎡ A11 A12 ⎤ ⎡ F1 ⎤
⎢ KE ⎥ TcrAi Tcr −1 = Areg = ⎢ i i
⎥ , TcrFi = Freg = ⎢ i ⎥ , TcrEij
⎢0 0 0 0 − 2πi Ksij ⎥ i ⎢⎣ A21i A22i ⎥⎦ i
⎣⎢ F2i ⎥⎦
⎢⎣ ⎥
0 0 0 0 0 ⎦ ⎡ E11 E12 ⎤ ⎡ ⎤
Tcr −1 = Ereg = ⎢
ij ij
⎥ , T B = Breg = ⎢ 0 ⎥
ij ⎢⎣ E21ij E22ij ⎥⎦ cr i i ⎣ B2i ⎦
⎡ 1 ⎤T ⎡ K ⎤T (8)
Bi = ⎢ 0 0 0 0 ⎥ ; Fi = ⎢− pi 0 0 0 0 ⎥ ; xi(t )
⎣ TGi
⎦ ⎣ Tpi ⎦
⎡ ⎤T
= ⎢ Δfi (t ) ΔPgi(t ) ΔX gi(t ) ΔEi(t ) Δδi(t )⎥
⎣ ⎦
Remark 1:. The matrix pair (Ai , Bi ) for ith area interconnected system
where, xi(t ) ∈ R ni is the state vector, xj (t ) ∈ R nj is the adjacent area should be controllable. Then it implies that the pair
state vector of xi(t ); ui(t ) ∈ R mi is the control vector; ΔPdi(t ) ∈ R ki is ⎛ N N ⎞
⎜A11i + ∑ j =1, i ≠ j E11ij , A12i + ∑ j =1, i ≠ j E12ij ⎟ is also controllable in regular
the vector of load disturbance. State variables ⎝ ⎠
Δfi (t ), ΔPg (t ), ΔX g (t ), ΔEi(t ) and Δδi(t ) are the incremental deviation in form (Utkin, 1992).
The state equations of ith area in regular form are given as:
i i
frequency (Hz), generator output (p.u. MW), governor valve position
(p.u. MW), integral control and rotor angle, respectively; K pi, Ri , K Ei
N N
and K Bi is the power system and machines gain, speed regulation z1̇ i(t ) = A11i z1i(t ) + ∑ j =1, j ≠ i E11ijz1j (t ) + A12i z2i(t ) + ∑ j =1, j ≠ i E12ijz2j (t )
coefficient (p.u. MW−1), integral control gain and frequency bias factor, + F1iΔPdi(t )
respectively; Tpi, Tti and TGi are the time constants in (sec) of power
N N
system and machines, turbine and governor, respectively; Ksij is the z2̇ i(t ) = A21i z1i(t ) + ∑ j =1, j ≠ i E21ijz1j (t ) + A22i z2i(t ) + ∑ j =1, j ≠ i E22ijz2j (t )
interconnection tie line gain between area i and j ( j ≠ i ). The dimen- + B2iui(t ) + F2iΔPdi(t )
sion of ith area matrices in (6) are Αi ∈ R ni × ni, (9)
Eij ∈ R ni × nj , Bi ∈ R ni × mi, Fi ∈ R ni ∈ ki .
Where,
The aforementioned ith area can be transferred in the regular form
z1i(t ) ∈ R(ni − pi ), z2i(t ) ∈ R pi, A11i ∈ R(ni − pi )×(ni − pi ), A12i ∈ R(ni − pi )× pi, A21i -
(Fulwani et al., 2011) using transform matrix Tcri .
∈ R pi ×(ni − pi ), A22i ∈ R pi × pi,
⎡ z1 (t ) ⎤
zi(t ) = ⎢ i ⎥ = Tcrixi(t ) E11i ∈ R(ni − pi )×(ni − pi ), A12i ∈ R(ni − pi )× pi, A21i ∈ R pi ×(ni − pi ), A22i ∈ R pi × pi, E11i
⎢⎣ z2i(t )⎥⎦
∈ R(ni − pi )×(ni − pi ), E12i ∈ R(ni − pi )× pi
⎡ 0(n − p )× p ⎤ and F2i ∈ R pi × pi are the transformed matrix forms at nominal value.
where, the matrix Tcri satisfies TcriTcrTi = I and, TcriBi = ⎢ i⎥
i i
, where I The matrix form of ith area dynamic model with parameter uncertain-
⎢⎣ [B2i ] pi × pi ⎥⎦
ties is given as:
is identity matrix and B2i is a nonsingular matrix. The transformed
⎡ ⎤T ⎡ ⎤T
states for ith area; z1i = ⎢ Δfi ΔPgi −Δδi ΔEi ⎥ ,z2i = ⎢ΔX gi ⎥
⎣ ⎦ ⎣ ⎦
Eq. (6) in the regular form is given by:

83
S. Prasad et al. Control Engineering Practice 61 (2017) 81–92

N
z1̇ i(t ) = (A11i + ΔA11i )z1i(t ) + ∑ j =1, j ≠ i E11ijz1j (t ) + (A12i + ΔA12i )z2i(t ) ⎡ ⎤⎡ z1 (t ) ⎤
T
σi(t ) = ⎢ Ki − ψi (yi )A12 Pi 1⎥⎢ i ⎥
N ⎣ i ⎦⎢⎣ z2i(t )⎥⎦ (13c)
+ ∑ j =1, j ≠ i E12ijz2i(t ) + (F1i + ΔF1i )ΔPdi(t )
N In Eq. (13), matrix Ki is designed through pole assignment
z2̇ i(t ) = (A21i + ΔA21i )z1i(t ) + ∑ j =1, j ≠ i E21ijz1j (t ) + (A22i + ΔA22i )z2i(t )
N
(Siaramakrishana et al., 1984)/ linear inequality matrices (LMIs)
+ ∑ j =1, j ≠ i E22ijz2j (t ) + (B2i + ΔB2i )ui(t ) optimization (Prasad et al., 2016) to ensure that (A11i − A12i Ki ) results
+ (F2i + ΔF2i )ΔPdi(t ) in stable eigenvalues and the dominant poles are characterized to have
a low-damping ratio. ψi (yi ) is a non-positive differentiable function in
(10) output yi = Δfi , which is used to change the damping ratio (Prasad et al.,
where, ΔAi , ΔBi and ΔFi represent parameter uncertainties. In order to 2016). Pi is a positive-definite matrix which is considered on the basis
simplify the proof, the uncertainty is not considered in the intercon- of desired final damping ratio (ζ2i ), satisfying the Lyapunov's equation
nection term Eij . for some positive definite matrix Wi .
Defining g1 (t ) = ΔA11i z1i(t ) + ΔA12i z2i(t ) + (F1i + ΔF1i )ΔPdi(t ) and
i (A11i − A12i Ki )T Pi + Pi (A11i − A12i Ki ) = − Wi (14)
g2 (t ) = ΔA21i z1i(t ) + ΔA22i z2i(t ) + ΔB2iui(t ) + (F2i + ΔF2i )ΔPdi(t ) as the ag-
i
gregate uncertainties, Eq. (10) becomes The damping ratio of ith area is increased from its initial value (ζ1i )
N N to the final value (ζ2i ) using non-linear switching surface.
z1̇ i(t ) = A11i z1i(t ) + ∑ j =1, j ≠ i E11ijz1j (t ) + A12i z2i(t ) + ∑ j =1, j ≠ i E12ijz2j (t )

+ g1 (t )
i 3.1. Selection of non-linear function ψi (yi )
N N
z2̇ i(t ) = A21i z1i(t ) + ∑ j =1, j ≠ i E21ijz1j (t ) + A22i z2i(t ) + ∑ j =1, j ≠ i E22ijz2j (t ) The damping ratio is changed from its initial low value to final
desired value on frequency deviation in the system, with selection of
+ B2iui(t ) + g2 (t ) non-linear function ψi (yi ), for area i. The non-linear function is
i

(11) characterized by following properties:

where,
• Changes from 0 (or a very low negative value) to - βi as frequency
deviation occurs in the systems, where βi > 0 .
g1 ≤ ΔA11i
i
z1i + ΔA12i z2i + ( F1i + ΔF1i ) ΔPdi = di,
• Differentiable with respect to time to ensure the existence of the
and g2 ≤ ΔA21i z1i + ΔA22i z2i + ΔB2i ui(t ) sliding mode.
i

+ ( F2i + ΔF2i ) ΔPdi = ηi (12) The non-linear function ψi (yi ) is defined as (Prasad et al., 2016),
(Chen et al., 2003):
where, . is the matrix norm. The matched and unmatched uncer-
tainties in Eq. (12) have been discussed in (Goshaidas et al., 2004), (Mi ⎛ −(1−((y − y )/(r − y ))2)
βi ⎞
ψi (yi ) = − ⎜e i i0
−1 ⎝
i i0 − e−1⎟
et al., 2013). 1−e ⎠ (15)

Where, yi 0 = yi (0) is initial value of the corresponding state,ri is positive

2.2. Objective of the paper
small scalar value and βi is a tuning parameter that determines the
selection of final damping ratio along with the matrix Pi for ith area. It
The objective of study is to achieve overall minimum overshoot/
is important to note that selection of ψi (yi ) is not unique and any
undershoot and settling time, on the controlled output of the closed- function with the above properties can be considered in the design.
loop system in the presence of system matched and unmatched As the dynamic trajectory reaches the sliding mode, the switching
uncertainties, different load disturbance patterns and non-linearities function satisfies the following condition:
such as GRC, GDB. An effective control law is designed for matched as
well as unmatched uncertainties. Thus, the proposed controller has σi(t ) = 0 and σi̇ (t ) = 0 (16)
ability to vary closed-loop system damping characteristics according to
σi(t ) = 0 implies:
present uncertainties and load disturbances in the system. The
designed control signal provides significantly reduced chattering effect z2i = − c1iz1i (17)
resulting to reduce wear-out of actuators (valve) in steam turbine.
Differentiating Eq. (13b) and substituting σi̇ (t ) = 0 gives

3. Design of non-linear sliding surface σi̇ (t ) = c1̇ iz1i + c1iz1̇ i + z2̇ i = 0 (18)

Substituting Eq. (11) into Eq. (18) gives

This section presents the design of non-linear sliding surface for
uncertain power system. Traditionally, the SMC design (Goshaidas ⎡ N
et al., 2004) includes two parts; switching control law and equivalent uieq(t ) = − B2−1⎢( c ̇ + A c + A )z + ∑ (E11ijc1i + E21ij )z1j
i ⎢ 1i 12i 1i 21i 1i
control law. The switching sliding surface σi(t ) is meant to drive ith area ⎣ j =1, j ≠ i

states move towards a predefined sliding surface. On the other hand, N ⎤

control law ui(t ) guarantees ith area states trajectory to the sliding + (A12i c1i + A22i )z2i + ∑ (E12ijc1i + E22ij )z2j + g2 ⎥
i⎥
surface and maintain convergence to the surface. j =1, j ≠ i ⎦ (19)
The dynamic performance and robustness during the reaching
Substituting Eq. (19) in Eq. (11), the equivalent dynamic equation
phase with the matched and unmatched uncertainties is guaranteed
in sliding mode is given as:
with selection of non-linear switching surface (Chen, Lee, & Peng,
2003; Fulwani et al., 2011; Prasad et al., 2016; Utkin, 1992) given as: N N
z1̇ i(t ) = A11i z1i + ∑ j =1, j ≠ i E11ijz1j + A12i z2i + ∑ j =1, j ≠ i E12ijz2j + g1 (t )
i
⎡ z1 (t ) ⎤ N
z2̇ i(t ) = − (c1̇ i + A12i c1i )z1i − c1i ∑ j =1, j ≠ i E11ijz1j − A12i c1iz2i
σi(t ) = ciz1i(t ) = [ c1i 1]⎢ i ⎥
⎢⎣ z2i(t )⎥⎦ (13a) N
− c1i ∑ j =1, j ≠ i E12ijz2j − g1 c1i (20)
i
T
c1i = Ki − ψi (yi )A12 P
i i (13b) Substituting Eq. (17) in Eq. (20) gives

84
S. Prasad et al. Control Engineering Practice 61 (2017) 81–92

N T
Vi̇ (z ) = z1Ti [(A11i − A12i Ki ) + ψi (yi )A12i A12
T
P ] Pz
z1̇ i(t ) = (A11i − A12i c1i )z1i(t ) + ∑ (E11ij − E12ijc1i )z1j (t ) + g1 (t )
i
i i i 1i
j =1, j ≠ i (21) + z1Ti Pi [(A11i − A12i Ki ) + ψi (yi )A12i A12
T
P ]z
i i 1i
N N ⎛ ⎞
T
z1̇ i(t ) = ((A11i − A12i Ki ) + ψi (yi )A12i A12 P )z + ∑ ((E11ij − E12ijKi ) +z1Ti Pi ∑ j =1, j ≠ i ⎜(E11ij − E12ijKi ) + ψi (yi )E12ijA12
T
P ⎟z
i ⎠ 1j
i i 1i ⎝
j =1, j ≠ i
T N ⎛ ⎞T
+ψi (yi )E12ijA12 P )z
i i 1j
+ g1 (t ) (22) + z1Tj ∑ j =1, j ≠ i ⎜(E11ij − E12ijKi ) + ψi (yi )E12ijA12
T
P ⎟ Pz
i ⎠ i 1i
i ⎝
The next subsection proves that the system Eq. (22) using designed Vi̇ (z ) = z1Ti [−Wi + + 2ψi (yi )PA T
i 12i A12i Pi ]z1i
switching surface of the non-linear SMC assures asymptotically stable. ⎛ ⎞
N
+ z1Ti Pi ∑ j =1, j ≠ i ⎜(E11ij − E12ijKi ) + ψi (yi )E12ijA12
T
P ⎟z
i ⎠ 1j
Theorem 1. If (A11i − A12i Ki ) is stable and ψi (yi ) is defined by Eq. (15), ⎝
then subsystem in Eq. (22) is asymptotically stable under both matched N ⎛ ⎞T
and unmatched uncertainties. +z1Tj ∑ j =1, j ≠ i ⎜(E11ij − E12ijKi ) + ψi (yi )E12ijA12
T
P⎟ Pz
⎝ i ⎠ i 1i

Remark 2. During the sliding mode, the stable dynamics of the system Thus,
in presence of the uncertainties are decided by the matrix Pi even for
any non-positive intermediate value of ψi (yi ) from 0 to – βi . Whenever a Vi̇ (z ) = z1Ti [−Wi + 2ψi (y )PA T T T
i 12i A12i Pi ]z1i + z1j Mi Pz
T
i 1i + z1i MiPz
i 1j (28a)
frequency deviation occurs in the system, the non-linear function is
small (zero) which implies that the effect of the non-linear part on the As discussed above, the non-linear function varies from 0 (a very
overall system is very limited and damping ratio is contributed by Ki low negative value) to - βi . Then Eq. (28a) becomes:
which is designed for a low damping ratio ζ1i of the system. When the
frequency deviation approaches to zero and non-linear function Vi̇ (z ) = z1Ti [−Wi − 2βiPA T T T
i 12i A12i Pi ]z1i + z1j Mi Pz
T
i 1i + z1i MiPz
i 1j (29a)
becomes - βi , the proposed control law provides desired damping ratio
ζ2i . Thus, the non-linear function varies from 0 (a very low negative N ⎛ T
⎞ T
where, Mi = ∑ j =1, j ≠ i ⎜(E11ij − E12ijKi ) − βiE12ijA12 P⎟ is a positive definite
value) to - βi , whenever a frequency deviation occurs in the system. Thus ⎝ i ⎠

in order to achieve our objectives and limit the effect of non-linear part matrix. Therefore, Vi̇ (z ) < 0, this proves that Eq. (25a) is asymptotically
on the overall system, we have chosen a non-linear SMC and its stable.
variables.

3.3. Sliding surface for unmatched uncertainties

3.2. Sliding surface for matched uncertainties

This subsection describes a theorem to assure that system dynamic

In this section, the stability of system dynamic trajectory in sliding
trajectory in sliding mode is stable with unmatched uncertainties,
mode is proved based on Lyapunov's stability theorem considering that
based on Lyapunov's stability theorem. It is considered that uncertain-
uncertainties satisfy the following assumptions:
ties satisfy the following assumption.
Assumption 1. The following matching condition is satisfied.
Assumption 2. rank (Bi , gi(t ))≠ rank (Bi ),
rank (Bi , gi(t )) = rank (Bi ), then g1 (t ) ≤ di = 0 (23) Under Assumption 1, the equivalent controller including un-
i

matched uncertainty is given in Eq. (19)

Substituting Eq. (23) into Eq. (19) gives
⎡ (c ̇ + A c + A )z + ∑ N (E11ijc1i + E21ij )z1j + g1 c1i ⎤⎥
⎡ N ⎢ 1i 12i 1i 21i 1i j =1, j ≠ i i
uieq(t ) = − B2−1⎢( c ̇ + A12i c1i + A21i )z1i + ∑ (E11ijc1i + E21ij )z1j ⎢ ⎥
i ⎢⎣1i uieq(t ) = − B2−1
i ⎢
+ (A12i c1i + A22i )z2i ⎥
j =1, j ≠ i
⎢ N ⎥
N ⎤ ⎢⎣ + ∑ j =1, j ≠ i
( E c
12ij 1i + E ) z
22ij 2j + g2i ⎥⎦
+ (A12i c1i + A22i )z2i + ∑ (E12ijc1i + E22ij )z2j ⎥
⎥⎦ (24b)
j =1, j ≠ i (24a)
From Eq. (11) & Eq. (17), Eq. (22) becomes the equivalent
Substituting Eq. (23) in Eq. (22), the equivalent dynamic equation
dynamic equation in sliding mode:
in sliding mode is obtained as:
T
T z1̇ i(t ) = ((A11i − A12i Ki ) + ψi (yi )A12i A12 P )z
i i 1i
z1̇ i(t ) = ((A11i − A12i Ki ) + ψi (yi )A12i A12 P )z
i i 1i N
⎛ ⎞
N
⎛ T
⎞ + ∑ T
⎜(E11ij − E12ijKi ) + ψi (yi )E12ijA12 P⎟z1j + g1 (t )
+ ∑ ⎜(E11ij − E12ijKi ) + ψi (yi )E12ijA12 P ⎟z
i ⎠ 1j ⎝ i ⎠ i
j =1, j ≠ i
⎝ (25a)
j =1, j ≠ i (25b)

For the case of matched uncertainty, the stability is proved with Similarly, the stability of unmatched uncertainty is also analyzed
Theorem 1. with Theorem 1.

Proof. Consider the Lyapunov's function for the ith area of the power Proof. Choosing a Lyapunov's function.
system in Eq. (25a) to be defined as:
Vi (z ) = z1Ti (t )Pz
i 1i (t ) (26b)
Vi (z ) = z1Ti (t )Pz
i 1i (t ) (26a)

Then, it follows that

Vi̇ (z ) = z1̇ Ti (t )Pz T
i 1i (t ) + z1i (t )Pz
i 1̇ i (t ) (27b)
Vi̇ (z ) = z1̇ Ti (t )Pz T
i 1i (t ) + z1i (t )Pz
i 1̇ i (t ) (27b)

Substituting Eq. (25a) gives Substituting Eq. (25b) gives

85
S. Prasad et al. Control Engineering Practice 61 (2017) 81–92

Table 1
Parameters of three area power system.

Area Tp Kp TTi TGi Ri K Ei K Bi Ksij

i i

1 20 120 0.3 0.08 2.4 0.15 0.41 0.55

2 25 112.5 0.33 0.072 2.7 0.15 0.37 0.65
3 20 115 0.35 0.07 2.5 0.12 0.40 0.545

4. Formulation of control law

The sliding surface discussed in above section is nonlinear and

surface parameters change continuously with time. The control law
should be chosen so that from any initial condition, the system
trajectory is attracted towards the sliding surface and then slides along
the surface. The existence of such control law is described by the
following theorem.
Theorem 2. The control law in Eq. (30).
⎛
ui(t ) = − B2Ti ⎜ciT Areg zi(t ) + ciT Ereg jzj (t ) + diσi(t ) + qisign(σi(t ))
⎝ i i

dψ (y ) T ⎞
− i i A12 Pz (t )⎟
i i 1i
dt ⎠ (30)
enforces the trajectory of Eq. (11) to move from any initial condition to
the sliding surface in finite time and remain on it thereafter. The scalar
qi is chosen from the maximum bound of the uncertainty.
qi > (ηi )max (31)
Where, di is given in Eq. (12).
Proof. For a quadratic function,
V1i = 0.5σiT (t )σi(t ) (32)

V1̇i(t ) = σiT (t )σi̇ (t ) (33)

Substituting Eqs. (17), (13) and Eq. (11) into (33), gives
⎡ dψ (y ) T ⎤
V1̇i(t ) = σiT (t )⎢Kiz1̇ i(t ) − i i A12 T
Pz (t ) − ψi (y )A12
i i 1i
Pz ̇ (t ) + z2̇ i(t )⎥
i i 1i
⎣ dt ⎦
Fig. 2. Flow chart illustrating the design steps and its parameters for the proposed Using Eqs. (11) and (13), it leads to
controller.
T
⎡ T T
̇
V1i(t ) = σi (t )⎢ci Areg zi(t ) + B2iui(t ) + ci Eregijzj (t ) + c1idi + ηi
⎣ i

Vi̇ (z ) = z1Ti [−Wi + 2ψi (yi )PA T

i 12i A12i Pi ]z1i ⎤
dψi(yi )
T
− A12 Pz (t )⎥
N ⎛ ⎞T dti i 1i ⎦
+ z1Tj ∑ j =1, j ≠ i ⎜(E11ij − E12ijKi ) + ψi (yi )E12ijA12 T
P ⎟ Pz
i ⎠ i 1i Using control law Eq. (30), it can be seen that
⎝
N ⎛ ⎞ V1̇i(t ) = σiT (t )[−diσi(t ) − qisign(σi(t )) + c1idi + ηi ]
+z1Tj g1 (t )Pz T T
i 1i + z1i Pi ∑ j =1, j ≠ i ⎜(E11ij − E12ijKi ) + ψi (yi )E12ijA12i P ⎟z1j
i ⎝ ⎠
Considering Eq. (32), it is straightforward to verify V1̇i(t ) < − ρi σi(t )
+ z1Ti Pg1 (t )z1j for some ρi > 0 . This ensures finite time convergence of σi and thus,
i
completes the proof.
From the Eq. (12), above equation becomes:
4.1. Summary on the proposed control design steps
Vi̇ (z ) = z1Ti [−Wi + 2ψi (yi )PA T T T
i 12i A12i Pi ]z1i + z1j Mi Pz
T
i 1i + z1i MiPz
i 1j

− di(z1Tj Pz T
i 1i + z1i Pz
i 1j ) (28b)
Detailed design of the proposed approach is summarized in the
following steps:
As discussed above, the non-linear function varies from 0 (a very
low negative value) to - βi . Then Eq. (28b) becomes: Step 1) Obtain the state-space model in regular form using
transformation from the linear model of the LFC and test the
Vi̇ (z ) = z1Ti [−Wi − 2βiPA T T T
i 12i A12i Pi ]z1i + z1j Mi Pz
T
i 1i + z1i MiPz
i 1j controllability condition as given in Remark 1.
− di(z1Tj Pz T Step 2) Ki is chosen such that (A11i − A12i Ki ) has stable eigenvalues
i 1i + z1i Pz
i 1j ) (29b)
and the dominant poles have low damping ratio.
N ⎛ T
⎞T Step 3) The real symmetric positive-definite matrix Pi is designed by
where, Mi = ∑ j =1, j ≠ i ⎜(E11ij − E12ijKi ) − βiE12ijA12 P⎟ is a positive-definite testing and solving the Eqs. (23), (12) and the Lyapunov stability Eq.
⎝ i ⎠
matrix. Therefore, Vi̇ (z ) < 0, this proves that the Eq. (25b) is asympto- (14).
tically stable even in presence of the unmatched uncertainties. Step 4) Design non-linear sliding surface Eq. (13) with the non-

86
S. Prasad et al. Control Engineering Practice 61 (2017) 81–92

0.3
0.2
Area-1
Area-1
Area-2 0.2
Area-2
Area-3

fi (Hz)
0.1

fi (Hz)
Area-3
0.1

0
0

-0.1 -0.1
0 5 10 15 0 5 10 15
Time (sec) Time (sec)
a) Scenario 1 (b) Scenario 2
Fig. 3. Variation in Δfi without the NLSMLFC. a) Scenario 1. (b) Scenario 2.

Fig. 4. Power system response for Scenario 2 with the NLSMLFC.

Table 2 The parameters of three area power system are given in Table 1. The
Poles location for control AREA-1. primary objective of designing a controller is to reduce undershoot/
overshoot and settling time in presence of parameter uncertainties.
Methods Pole locations
The range of subsystems parameter uncertainties to demonstrate
SMLFC −16.1907, −3.7481, −0.0464 the robustness of designed controller is referred from (Lim et al.,
−1.9328 ± 4.0504i 1996), (Mi et al., 2013). The simulation results are obtained for all four
scenarios; nominal parameters, matched uncertainty, unmatched un-
NLSMLFC −2.0020 ± 2.0389i,
certainties, and generation rate constraints (Mi et al., 2013). The
−1.9979 ± 1.5023i
controller design is also implemented on New England Test System
(IEEE 39 bus system) large multi-area power system whose parameters
linear function and finally, NLSMLFC Eq. (30) is obtained. are obtained from (Bevrani, 2014).

Scenario 1. In this subsection, the response of the plant is

The flow chart of the proposed design scheme is illustrated in Fig. 2.
5. Simulation results and discussion
To validate the effectiveness of the proposed control; NLSMLFC in
three area power system, the simulations is evaluated with MATLAB®.
demonstrated at nominal parameters without considering load
disturbances and parameters uncertainties. The frequency deviation
in three areas is shown in Fig. 3(a) without proposed controller. It is
evident that the conventional PI controller is settling down the system
slowly.
Scenario 2. The robustness of the designed NLSMLFC is demonstrated

87
S. Prasad et al. Control Engineering Practice 61 (2017) 81–92

Fig. 5. Power system responses for Scenario 3 with the NLSMLFC.

15
2
10
1
5
u1 (pu)
u1 (pu)

0
0

-5 -1

-10
-2 b) SMLFC
a) SMLFC
-15
0 1 2 3 4 5 0 1 2 3 4 5
8 0.3
6
4 0.2
u1 (pu)
u1 (pu)

2
0.1
0
-2 0

-4 c) NLSMLFC d) NLSMLFC
-0.1
0 1 2 3 4 5 0 1 2 3 4 5
Time (sec) Time (sec)

Fig. 6. Comparative control signal for Scenario 3.

Fig. 7. Comparative Δfi for Scenario 3.

88
S. Prasad et al. Control Engineering Practice 61 (2017) 81–92

Table 3
Performance index in control AREA-1.

Scenario Methods Performance indices

ISE IAE ITAE

2 SMLFC 0.1616 0.5694 0.9350

NLSMLFC 0.0033 0.0611 0.4633

3 SMLFC 0.0579 0.3391 0.4266

NLSMLFC 0.0019 0.0353 0.0133

4 SMLFC 0.0579 0.3391 0.8266

NLSMLFC 0.0019 0.0353 0.0133

1 s. The computed eigenvalues for area-1 are given in Table 2. It can be

seen that extreme rightmost conjugate eigenvalue (Italicized) has
increased value of real part and reduced natural frequency of oscilla-
tion for the system with the proposed controller. This leads to
comparatively faster response time for NLSMLFC.

Scenario 3. The three area power system is simulated at load

conditions; ΔPd1(t ) = 0.02pu , ΔPd2(t ) = 0.015pu and ΔPd3(t ) = 0.01pu
with unmatched uncertainties. The power system response obtained
with designed controller is shown in Fig. 5. The variation of tie-line
power deviation (third subfigure) is illustrated in said figure. Unlike,
the response [Fig. 7 of Ref. Mi et al. (2013)], the tie-line power
deviation obtained using proposed controller design does not have any
sustained oscillation. In addition, the said response has reduced
magnitude of steady-state error.
Next, the comparison of control effort with respect to control design (Mi
et al., 2013) is discussed. The switching coefficient gains are selected as
q1 = 3.2, β1 = 50000,K1 = [1.1709 1.3851 1.1100 15.4390 ],d1 = 2900 for
area-1.and the one proposed here is shown in Fig. 6. The said figure shows
the comparative required control effort with 0.1 and zero initial state
Fig. 8. Comparative Δfi for Scenario 3 with random load disturbances. (a) Load conditions. Fig. 6(a) and (c) shows the results for 0.1 initial state conditions,
disturbances profile in three areas. (b) Frequency deviation in three-area power system. while Fig. 6(b) & (d) for zero initial state conditions. Thus, it is clear that
the required control effort by proposed controller; NLSMLFC is less as
with the same matched parameter uncertainty as considered in (Lim compared to SMLFC. It is evident that in the proposed control design, the
et al., 1996; Mi et al., 2013) and represented as: effect of chattering has significantly reduced as compared to (Mi et al.,
2013). Thus, devastating effects caused due to wear-out of actuators (valve)
⎡ 0 0 0 0 0⎤
⎢ ⎥ in steam turbine is minimized. The corresponding variation in frequency
⎢ 0 0 0 0 0⎥ deviation response of proposed control design and SMLFC (Mi et al., 2013)
ΔA1 = ⎢ −2.26 cos(t ) 2 cos(t ) −2.604 cos(t ) 3 cos(t ) 0⎥
⎢ are shown Fig. 7. A comparatively reduced magnitude of undershoot and
0 0 0 0 0⎥
⎢ 0 ⎥⎦ settling time is achieved by proposed design, NLSMLFC against SMLFC
⎣ 0 0 0 0
(Mi et al., 2013). The frequency deviation is significantly suppressed using
The variation of Δfi without NLSMLFC applied in three area power NLSMLFC. Thus, it is evident that the proposed controller adaptively meets
system is shown in Fig. 3(b). It is observed that the frequency deviation the robust performance against uncertainties. An improved damping is
in each area is accompanied with high overshoot and fails to reach contributed to achieve desired response of frequency deviation and control
steady-state condition. The variation in frequency deviation, governor effort.
valve position, tie-line power deviation (ΔPtie ), switching surface and It is evident that the proposed controller design gives satisfactory
control signal (area-1) with NLSMLFC is shown in Fig. 4. It clearly performance accompanied with fast settling time, and undershoots/
depicts that the frequency deviation settles down quickly in less than overshoots in the range of ±0.05Hz . Thus, the proposed control design

Fig. 9. Frequency responses with GRC in three- area power system. a) With SMLFC. (b) With NLSMLFC.

89
S. Prasad et al. Control Engineering Practice 61 (2017) 81–92

Fig. 10. Variation in Δfi with non-linearity in the system. (a) With Governor Deadband and GRC. (b) Responses at different GRC limit in area-1.

Fig. 11. IEEE 39 bus system (Bevrani, 2014).

establishes the superiority as well as applicability for the LFC problem.
Next, random step disturbances applied in each area as shown in
Fig. 8(a) is considered. Mi et al. (2013) in their study has not
demonstrated system response using proposed control technique at
random load disturbances. The pattern of load disturbance applied on
the system is shown in Fig. 8(a) and the corresponding variation of
frequency deviation profile are depicted in Fig. 8(b). It is evident that
the response has low oscillation, reduced overshoot/undershoot and
small settling time by the proposed design. It is suggested that the
proposed control design performs satisfactory against random load
disturbances. Thus, it is evident that the same controller is indeed
capable of being robust against parameter uncertainties and achieves
superior performance as compared to the controller proposed in (Mi
et al., 2013). Thus, the proposed NLSMLFC minimizes the frequency
deviation and tie-line power exchange oscillation in interconnected
power systems even in the presence of parameters unmatched un-
certainties and random load disturbance.
It is apparent that very small frequency deviations with the
proposed control scheme have less impact on the plant reserve capacity
and power market. The devastating effect caused due to wear-out of

90
S. Prasad et al.
Fig. 12. Frequency deviations, tie-line power flow deviations and control efforts. (a)
Frequency deviation profile. (b) Control effort.
actuators (valve) in steam turbine is reduced. The controller maintains
robust performance and grid stability by minimizing the effects of
various disturbances. The presented controller design steps validate the
applicability of NLSMLFC in practice.
Scenario 4. This scenario is analyzed with inclusion of generation rate
constraint (GRC) effect in three area power system. The GRC [Fig. 1]
having its limits (Lim et al., 1996) is considered. Fig. 9 illustrates the
comparative dynamic response between NLSMLFC and SMLFC,
considering unmatched uncertainties in NLSMLFC. The variaion in
frequency deviations are suppressed and remain within permissible
limits..
Now, the performance indices; integral square error (ISE), integral
absolute error (IAE), integral time absolute error (ITAE) are computed
for frequency deviation in area-1 and given in Table 3, with GRC
included in area-1. These indices obtained from proposed controller
are comparatively low (shown in bold face) for the Scenarios 2, 3 and 4
considered in study. The system dynamic responses and performance
indices exhibit better performance as evidenced by overshoots/under-
shoots, settling time and transient frequency oscillations using the
proposed control (NLSMLFC). Thus, it is evident that the proposed
controller is robust and effective.
Scenario 5. Furthermore, another non-linearity in the form of dead
band (Fig. 1) in the governor operation is introduced with the limits of
0.06pu (Tripathy, Balasubramanian, & Nair, 1992). The frequency
deviation obtained by proposed control strategy with deadband and
GRC are shown in Fig. 10(a). The performance of proposed controller
for 20% and 40% decrease in GRC limits are shown in Fig. 10(b). This
demonstrates the satisfactory performance of proposed controller in
bringing the frequency deviation profile settle quickly.
Scenario 6. Now, the proposed controller is validated on large power
91
Control Engineering Practice 61 (2017) 81–92
system as considered in (Bevrani, 2014). This IEEE 39 bus system is
the well-known 10-machine New-England power system. The single-
line diagram of the IEEE 39 bus system is shown in Fig. 11. The control
area-1 and area-2 have three generators, and area-3 has four
generators, which are operating in parallel. The connected loads in
area-1, area-2 and area-3 and total power generation are 265.5, 233,
125 MW and 842 MW respectively. The parameters of the IEEE 39 bus
system can be referred from (Bevrani, 2014). The dynamic response of
said IEEE 39 bus system with proposed controller is simulated against
random step load disturbance applied in each area. The frequency
deviations and tie-line power exchange oscillation in the presence of
parameters unmatched uncertainties in each area is illustrated in
Fig. 12. It is noted that the proposed NLSMLFC improves the
transient and steady dynamics of the power system. The method has
significantly improved the frequency stability. The controller maintains
robust performance and grid stability by minimizing the effects of
various uncertainties and disturbances. Thus, the proposed controller
achieves the nominal frequency and thereby zero tie-line power
exchange deviation with minimum undershoot/overshoot and
minimum settling time. As discussed above, the designed control
signal provides significantly reduced chattering effect resulting in
decreased wear-out of actuators (valve) in steam turbine.
6. Conclusion
In this paper, a NLSMLFC has been designed by transforming each
control area into regular form. The proposed controller was designed to
take into account matched and unmatched uncertainties and to ensure
satisfactory performance under varying operating conditions of the
power system. In this study, a single effective control law was designed
for both matched and unmatched uncertainties. Thus, the proposed
controller has ability to vary closed-loop system damping character-
istics according to present uncertainties and load disturbances in the
system. The non-linear sliding surface was applied to each area to
improve the dynamic response. The frequency deviation profile was
observed with reduced magnitude of overshoot/undershoot and set-
tling time. In addition, significant reduction in chattering was noted in
the control signal, designed by NLSMLFC. The said controller design
achieved satisfactory response with parameter uncertainties. Also, the
proposed controller performed well with limits of GRC and GDB
included in the system. It is evident that undershoot/overshoot and
settling time are reduced by approximately 30% with linear SMC (Mi
et al., 2013). The performance indices are also calculated and
compared with linear SMC. The frequency oscillations are reduced to
zero and the enhanced system stability was achieved. Furthermore, the
NLSMLFC has been successfully implemented on IEEE 39 bus
(Bevrani, 2014) large power system for the random load disturbances.
Hence, it can be concluded that the proposed controller is not only
robust in the presence of uncertainties but also can be successfully
applied to a large complex multi-area power system.
References
Barisal, A. K. (2015). ‘Comparative performance analysis of teaching learning based
optimization for automatic load frequency control of multi-source power systems’.
(March)International Journal of Electrical Power and Energy System, 66, 67–77.
Bevrani, H. (2014). Robust power system frequency control (2nd edition). Springer.
Bevrani, H., & Hiyama, T. (2005). Robust decentralized PI based LFC design for time
delay power systems. Energy Conversion and Management, 49(2), 193–204.
Bevrani, H., Mitani, Y., & Tsuji, K. (2004). Robust decentralised load-frequency control
using an iterative linear matrix inequalities algorithm. IEE Proceeding -Gener
Transition Distric, 151(3), 347–356.
Camblong, H., Vechiu, I., Etxeberria, A., & Martínez, M. I. (2014). Wind turbine
mechanical stresses reduction and contribution to frequency regulation. Control
Engineering Practice, 30, 140–149.
Chandra Sekhara, G. T., Sahua, R. K., Baliarsinghb, A. K., & Pandaa, S. (2016). ‘Load
frequency control of power system under deregulated environment using optimal
firefly algorithm’. (Jan)International J of Electrical Power & Energy Sysem, 74,
195–211.
S. Prasad et al.
Chen, B. M., Lee, T. H., Peng, K. et al. (2003). Composite nonlinear feedback control for
linear systems with input saturation: theory and application’. IEEE Transition on
Autom Control, 48(3), 427–439.
Chen, H., Ye, R., Wang, X., & Lu, R. (2015). ‘Cooperative Control of Power System Load
and Frequency by Using Differential Games’. IEEE Transition on Control Systems
Technology, 23(3), 882–897.
Ersdal, A. M., Imsland, L., Uhlen, K., Fabozzi, D., & Thornhill, N. F. (2016). Model
predictive load–frequency control taking into account imbalance uncertainty.
Control Engineering Practice, 53, 139–150.
Fosha, C.E., Elgerd, O.I. ‘The megawatt-frequency control problem: new approach via
optimal control theory’, IEEE Trans. on Power Apparatus and Syst. PAS-89, 4, pp
563–567.
Fulwani, D., Bandyopadhyay, B., & Fridman, L. (2011). Non-linear sliding surface:
towards high performance robust control. IET Control Theory Appl, 6(2), 235–242.
Goshaidas, R., Sitansu, D., & Bhattacharyya, T. K. (2004). Multi-area load frequency
control of power systems: a decentralized variable structure approach. Elect Power
Compon System, 33(3), 315–331.
Hassan, Y. (2015). ‘Adaptive fuzzy logic load frequency control of multi-area power
system’. (Jun)International Journal of Electrical Power & Energy System, 68,
384–395.
Hsu, K. C. (1998). Decentralized variable structure model-following adaptive control of
interconnected systems with series nonlinearities. International Journal of System
Scienc, 29, 365.
Ibraheem, I., Kumar, P., & Kothari, D. P. (2005). Recent philosophies of automatic
generation control strategies in power systems. IEEE Transition Power System,
20(1), 346–357.
Ismayil, C., Ramdas Sreerama, K., & Thiruthimana Krishnan, S. (2015). Optimal
fractional order PID controller for automatic generation control of two-area power
systems. IEEE Transition on Power Systems, International Transition Electrical
Energy System, 25(12), 3329–3348.
Khodabakhshian, A., & Edrisi, M. (2008). A new robust PID load frequency controller.
Control Engineering Practice, 16, 1069–1080.
Khooban, M. H., & Niknam, T. (2015). A new intelligent online fuzzy tuning approach for
multi-area load frequency control: Self adaptive modified bat algorithm.
International Journal of Electrical Power and Energy System, 71, 254–261.
Kundur, P. (1994). Power system stability and control New Delhi: McGraw-Hill.
Lim, K. Y., Wang, Y., & Zhou, Y. (1996). Robust decentralized load-frequency control of
multi-area power systems. Proceedings Inst Elect Eng Gene, Trans, Distr, 143(5),
377–386.
Mi, Y., Fu, Y., Wang, C., & Wang, P. (2013). Decentralized sliding mode load frequency
92
Control Engineering Practice 61 (2017) 81–92
control for multi-area power system. IEEE Trans on Power Systems, 28(4),
4301–4309.
Prakasha, S., & Sinhab, S. K. (2015). ‘load frequency control of multi-area power systems
using neuro-fuzzy hybrid intelligent controllers’. (April)IETE Journal of Research,
5(61), 526–532.
Prasad, S., Purwar, S., & Kishor, N. (2016). On design of a non-linear sliding mode load
frequency control of interconnected power system with communication time delay.
IET General Transition District, 10(11), 2771–2784.
Rerkpreedapong, D., Hasanovic, A., & Feliachi, A. (2003). Robust load frequency control
using genetic algorithms and linear matrix inequalities'. IEEE Transition on Power
System, 18(2), 855–862.
Saxena, S., Hote, & Yogesh, V. (2013). Load frequency control in power systems via
internal model control scheme and model-order reduction. IEEE Transition on
Power Systems, 28(3), 2749–2758.
Siaramakrishana, A. Y., Hariharm, M. V., & Srisailam, M. C. (1984). ‘Design of variable
structure load-frequency controller using pole assignment technique’. (Sep)
International Journal Control, 40(3), 487–498.
Tarek Hassan Mohamed, T. H., Morel, J., Bevrani, H., Abd-Eltawwab Hassan, A.,
Mohamed, Y. S., & Hiyama, T. (2012). ‘Decentralized model predictive-based load-
frequency control in an interconnected power system concerning wind turbines'.
IEEJ Transition, 7, 487–494.
Tarkeshwar, M., & Mukherjee, V. (2015). ‘Quasi-oppositional harmony search algorithm
and fuzzy logic controller for load frequency stabilization of an isolated hybrid power
system’. (April)IET General Transition District, 9(5), 427–444.
Tripathy, S. C., Balasubramanian, R., & Nair, P. S. C. (1992). ‘Effect of superconducting
magnetic energy storage on automatic generation control considering governor dead
band and boiler dynamics’. (Aug)IEEE Transition on Power System, 7(3),
1266–1273.
Utkin, V. I. (1992). Sliding mode in control optimization. Springer-Verlag.
Vrdoljak, K., Peric, N., & Petrovic, N. (2010). Sliding mode based load frequency control
in power systems. Electrical Power System Research, 80(5), 514–527.
Yinsong, W., Shizhe, Li, Jingyu, T., & Zheng, Z. (2016). Performance assessment of
thermal power plant load control system based on covariance index. Control
Engineering Practice, 54, 58–69.
Yousef, H. A., AL-Kharusi, K., Albadi, M. H., & Hosseinzadeh, N. (2014). ‘Load frequency
control of a multi-area power System: An adaptive fuzzy logic approach’. (Jul)IEEE
Transition on Power Systems, 29(4), 1822–1831.
Zribi, M., Al-Rashed, M., & Alrifai, M. (2005). Adaptive decentralized load frequency
control of multi-area power systems’, elect. Power Energy System, 27(8), 575–583.