Electrical Power and Energy Systems 104 (2019) 178–193

Contents lists available at ScienceDirect

Electrical Power and Energy Systems

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

Load frequency regulation using observer based non-linear sliding mode T

control
⁎
Sheetla Prasada, , Shubhi Purwarb, Nand Kishorc
a
School of Electrical, Electronics and Communication Engineering, Galgotias University, Greater Noida, India
b
Department of Electrical Engineering, Motilal Nehru National Institute of Technology, Allahabad, India
c
Department of Engineering Sciences, University of Agder, Grimstad, Norway

A R T I C LE I N FO A B S T R A C T

Keywords: In this paper, the generalized extended state observer (GESO) and non-linear sliding mode control (SMC) are
Generalized extended state observer merged together to study the frequency deviation problem in multi-area power system. In study, the GESO is
Non-linear sliding mode control used not only for state and disturbance estimation but also for disturbance rejection of the system. The said
Disturbance rejection control observer ensures accurate estimation of the actual states leading to convergence of estimation error to zero. The
Load frequency control
non-linear SMC is used to increase the damping ratio of the system whenever, any perturbations occurs. The
Linear matrix inequality (LMI)
proposed observer based controller is compared with an existing two-layer active disturbance rejection control
(ADRC) and also validated on large power system at random load disturbance. Further, the proposed controller
performance is tested on the minimum and non-minimum phase systems. In addition to fast response in terms of
settling time and reduced over/undershoots, the proposed control scheme satisfactorily compensates the dis-
turbance in the system. The proposed scheme is further tested in the presence of power system non-linearities
such as generation rate constraints and governor dead-band. The simulation results illustrate the robustness of
proposed controller when subjected to load disturbances and non-linearities.

1. Introduction [9–11]. The changing operating characteristic of generating unit re-

The modern power system requires an intelligent LFC strategy to be
developed in order to maintain the frequency of each area and to keep
tie-line power exchange within pre-specified tolerance [1]. This led
researchers to review the control techniques [2,3] for the conventional
LFC scheme (PI/PID based LFC scheme [4]). In multi-area power
system, the frequency deviation occurs mainly due to load disturbance.
Due to the varying load composition with time, weather, temperature
and uncertainty in load characteristics, it is difficult to accurately model
the load for the conventional LFC studies. The secure operation of in-
terconnected power systems with varying loads have been a challenge
for the researchers [5,6].
The LFC and tie-line power flow control have become increasingly
important world-wide with the aggressive development of the smart
grid. In past few years, several control techniques for the LFC appli-
cation have been proposed in literatures [7,8]. These studies have
evolved by modifying the conventional LFC according to the changing
operating conditions of the power system. In order to overcome the
limitations of conventional LFC scheme, numerous techniques have
been investigated and improved performance reported in the literatures
quires the adjustment of servo-motor (actuator) to regulate the water/
steam flow rate in the turbine. Therefore, the parameter uncertainty is
an important issue in the controller design [12]. Several authors
[12–14] have applied variable structure theory for the design of LFC but
it has poor ability of handling system changes caused by wear and tear
over prolonged use. The sliding mode control (SMC) is a form of vari-
able structure control. The applications of SMC for LFC are given in
[15,16] and the references therein. Mi et al. [15] has presented dis-
turbance observer based sliding mode load frequency controller in
presence of different load disturbances, wind power, parameter varia-
tion and generation rate constraint (GRC). Vrdoljak et al. [16] proposed
a discrete-time SMC for LFC application. The non-linear switching
surface [17] is applied on delay based LFC problem to improve the
system dynamic performance. In most of these SMC techniques, the
sliding surface is a linear combination of the system states constructed
using appropriate time-invariant coefficients. As a result, sliding surface
remains static even in the presence of system uncertainties, exhibiting
robustness to system uncertainties. However, it leads to a sharp tran-
sient response around the equilibrium state of the system frequency.
An effort has been made by the researchers to compensate the

⁎
Corresponding author.
E-mail address: sheetla.prasad@galgotiasuniversity.edu.in (S. Prasad).

https://doi.org/10.1016/j.ijepes.2018.06.035
Received 7 August 2017; Received in revised form 5 June 2018; Accepted 18 June 2018
Available online 11 July 2018
0142-0615/ © 2018 Elsevier Ltd. All rights reserved.
S. Prasad et al. Electrical Power and Energy Systems 104 (2019) 178–193

Area-1 (Thermal)
1
B1
R1 Governor d1
Turbine
+
ACE1 K 1
− DB v1
1
m1
− K p1 1
E1 1
1+ sTp1
+
s − +
1+ sTg1 + −
sTch1
+ − Rotating mass
GRC and load
u1 System
Non-linear SMC GESO
+
Input and
Output 2 1
u2 Tie-line
12 s
−
−ACE + +
2 K E2
2
1
v2
1 m2 K p2
− 1+ sTg2 sTch2 1+ sTp2
+
s
− − + − +
− Rotating mass2
DB
GRC and load
1 Governor Turbine d2
B2
R2
Area-2 (Thermal)

(a) LFC scheme for two-area power system

d
y(t)
u(t) Two-area power
system

Non-linear
SMC
GESO

d̂(t)
x̂(t)

Control Center

(b) Schematic diagram of the proposed control scheme

Fig. 1. LFC scheme and proposed control scheme.

uncertainties/non-linearities in the system by active disturbance re-
jection control (ADRC). An active disturbance rejection control [18]
uses an extended state observer (ESO) to estimate and compensate the
total disturbance. These disturbances are all internal uncertainties of
the process dynamics and external disturbances in real time [19]. The
inherent feature of ADRC is that the uncertainties and disturbances in
the controlled operation of the system can be viewed as a total dis-
turbance. Thus, it can be estimated and compensated via the ESO. The
ADRC approach has achieved successful applications in real time. Due
to inherent feature of disturbance rejection, ADRC becomes more
popular and suitable in industries mainly for control of actuators
[20,21], motion control [22], power systems [23,24], fault diagnosis
[25] etc. This is due to the fact that the ESO is restrictive as (i) it can
only be applied to system which can be represented by the integral
chain form [18], and (ii) it can estimate the disturbances due to mat-
ched uncertainties only. A generalized extended state observer (GESO)
proposed in [26], can be applied to a wider class of systems. Moreover,
it is capable to estimate and compensate matched and mismatched
uncertainties of the system. The two-layer ADRC [24] is designed using
equivalent input disturbance (EID) compensator and internal model
control (IMC). The main drawback of this approach is that stability of
the closed loop depends on proper design of the low-pass filter in terms
of high cut-off frequency and thus becomes more complex. The un-
certainties in load disturbance of the power system affect the fixed
frequency bandwidth of low-pass filter which may degrade the dis-
turbance rejection capability or even lead to system instability. A non-
linear SMC is designed in [33] assuming all states to be measurable
which is not always practically possible. The effectiveness of the control
scheme is verified in the presence of the uncertainties and system non-
linearities. The control scheme however requires high control effort
during the transient time interval.
The different control schemes in power systems assume that all the
state variables are available for controller design. However, deviations
in governor valve position in power system are not easy to measure.
Therefore, in this paper, the states and load disturbances are estimated
by the GESO and used in the design of non-linear SMC. The objective of
non-linear SMC based LFC is to achieve minimum overshoot/under-
shoot and settling time, on the controlled output of the closed-loop
system in the presence of system non-linearities/uncertainties such as
GRC, GDB and different load disturbance patterns. The limitation of the
control scheme presented in [33] is removed completely in this study
using the LMI approach. The positive-definite matrix and final feedback

179
S. Prasad et al. Electrical Power and Energy Systems 104 (2019) 178–193

gain are obtained using LMI based optimization. The proposed control 1 1 1 1
ΔPvi̇ (t ) = − Δf (t )− ΔPvi (t )− ΔEi (t ) + ui (t )
scheme improves both the steady state and transient response of the Ri Tgi i Tgi Tgi Tgi (5)
closed-loop system simultaneously in the presence of system dis-
turbances and non-linearities with a significant reduction in chattering. ΔPij (t ) = −ΔPji (t ) = ΔPtie ; where i, j = 1, 2 and i ≠ j.
The low chattering as discussed later is due to reduction in unmodeled The state space representation of the above two-area power system
dynamic effect and small switching gains. Thus, devastating effects equations (1)(5) is now given as:
caused due to wear-out of actuators (valve) in steam turbine is mini- x ̇ (t ) = Ax (t ) + Bu u (t ) + DΔPd (t )
mized. y (t ) = Cx (t ) (6)
To the best of author’s knowledge, GESO is proposed for the first
time for the LFC problem in our work. The contributions of the pro- where,
posed GESO based non-linear SMC are the following: x (t ) = [ΔP12 (t )Δf1 (t )ΔPm1 (t )ΔE1 (t )ΔPv1 (t )Δf2 (t )ΔPm2 (t )ΔE2 (t )ΔPv2 (t )]T
is the state variables and system matrices are given as:
• Estimate states and load disturbance accurately by GESO.
• Minimize the effect of load disturbances on system dynamics via its ⎡ 0 2πT1 0 0 0 2πT1 0 0 0 ⎤
⎢ −K p1 −1 −K p1 0 0 0 0 0 0 ⎥
compensation. ⎢ Tp1 Tp1 Tp1 ⎥
• Achieve minimum over/undershoot and reduced settling time si- ⎢
⎢ 0 0 −1 −1
0 0 0 0 ⎥
⎥
multaneously with minimum chattering effect. Tchl Tchl

• Decreases the chattering phenomenon in the sliding mode control

⎢ ⎥
⎢ K1 k1 B1 0 0 0 0 0 0 0 ⎥
due to unmodeled dynamic effect and small switching gain varia- ⎢ 0 −1
0 −1 −1
0 0 0 0 ⎥
A=⎢ R1 Tg1 Tg1 Tg1 ⎥, Bu
tions. ⎢ K p2 ⎥
−1 −K p2
⎢ Tp 0 0 0 0 0 0 ⎥
Tp2 Tp2
Remaining sections of the paper is organized as follows; Section 2 ⎢ 2 ⎥
⎢ 0 −1 −1 ⎥
0 0 0 0 0 0
describes briefly the modelling of two-area power system, followed by ⎢ Tchl Tch2 ⎥
GESO design in Section 3. The formulation of GESO based non-linear ⎢ − k2 0 0 0 0 k2 B2 0 0 0 ⎥
⎢ −1 −1 −1 ⎥
SMC scheme is proposed in Section 4. LMI optimization analysis is also ⎢ 0 0 0 0 0 0
R2 Tg2 Tg2 Tg2 ⎥
included in this section. The stability analysis on GESO based non-linear ⎣ ⎦
SMC is presented in Section 5, followed by results and discussion in ⎡0 0⎤ ⎡ −0K 0 ⎤
Section 6. Lastly, the conclusions drawn from the presented work are ⎢0 0⎥ ⎢ pl 0 ⎥ ⎡1 0 0 0 0 ⎤T
given in Section 7. ⎢0 0⎥ ⎢ Tpl ⎥ ⎢0 1 0 0 0⎥
⎢0 0⎥ ⎢ 0 0 ⎥ ⎢0 0 1 0 0⎥
⎢ 1 ⎥ ⎢ 0 0 ⎥ ⎢0 0 0 0 0⎥
0⎥ ⎢ ⎥
2. Two-area power system = ⎢ Tg1 , D= ⎢ 0 0 ⎥, C = ⎢0 0 0 0 0⎥
⎢ ⎥ ⎢0
⎢0 0⎥ ⎢ 0 −Kp2
⎥ ⎢ 0 0 1 0⎥⎥
In the operation of multi-area interconnected power system, in ad- ⎢0 0⎥ ⎢ Tp2 ⎥
⎢0 0 0 0 1⎥
⎢0 0⎥ ⎢ 0 0 ⎥ ⎢0 0 0 0 0⎥
dition to primary speed control loop, supplementary control action in ⎢ 0
⎢0 1 ⎥ 0 ⎥ ⎣0 0 0 0 0⎦
secondary loop is established to schedule the generation. As such, the ⎢ Tg 2 ⎥ ⎢ ⎥
⎣ ⎦ ⎣ 0 0 ⎦
power network although being a complex non-linear system, can be
considered linearized for the study of LFC problem due to its slow In general, the nomenclatures of the variables/parameters are given
convergence rate. The linearized model of two-area power system is in Table 1. The dimension of system matrices (6) are considered as
shown in Fig. 1(a). The equations that govern the dynamics of two-area A ∈ Rn × n , Bu ∈ Rn × k ,D ∈ Rn × r and C ∈ Rm × n respectively, and
system for i = 1, 2 is given as x ∈ Rn × 1, u ∈ Rk × 1, y ∈ Rm × 1 and ΔPd ∈ Rr × 1 are considered as “n”
states, “k” input, “m” output, “r” load disturbance vectors respectively.
1 K pi Kp Kp
Δfi ̇ (t ) = − Δf (t ) + ΔPmi (t )− i ΔPij (t )− i ΔPdi The design of the GESO based non-linear SMC for aforementioned
Tpi i Tpi Tpi Tpi (1) general system matrices is considered in the following section.

̇ (t ) = − 1 ΔPmi (t ) + 1 ΔPvi
ΔPmi 3. Design of generalized extended state observer
Tchi Tchi (2)
In practice, it is very difficult to directly measure the load dis-
ΔEi̇ (t ) = KEi Bi Δfi (t ) + KEi ΔPij (3)
turbances in power system due to its unknown characteristics. Hence, it
can be estimated from the limited available measurements. The GESO is
ΔPij̇ (t ) = 2πT1 (Δfi (t )−Δf j (t )) (4) used to estimate system states and extended state variables of the power

Table 1
Nomenclature of variables/parameters for i area and i ≠ j , where i = j = 1, 2 .
State variables/Parameters Nomenclature State variables/Parameters Nomenclature

Δfi Deviation in frequency (Hz) Tgi Governor time constant (sec.)

ΔPmi Deviation in generator output (p.u.MW) Tchi Turbine-generator time constant (sec.)
ΔEi Deviation in integral control (p.u.MW) T1 Interconnection tie-line gain (p.u.MW/rad/s.) between area-i and j
ΔPij = ΔPtie Deviation in tie-line power flow (p.u.MW) x State variables vector
ΔPdi Deviation in load disturbance (p.u.MW) u Input vector
ΔPvi Deviation in governor valve position (p.u.MW) y Output vector
Kpi Power system and machine gain (Hz/p.u.MW) xd Extended state variables vector
Ri Speed regulation coefficient (Hz/p.u.MW) h Derivative of load disturbance
KEi Integral control gain constant x Modified observer state variables
Bi Frequency bias factor (p.u.MW/Hz) x ,̂ x ̂ Estimated state variables
Tpi Power system and machine time constant (sec.) x d̂ = d ̂ Estimated load disturbances

180
S. Prasad et al. Electrical Power and Energy Systems 104 (2019) 178–193

system because it is more sensitive towards load disturbance variations. convergence to the surface. The dynamic performance and robustness
The design process of GESO is similar to the generalized extended state during the reaching phase against the uncertainties is guaranteed by
observer based control (GESOBC) [26]. To estimate the load dis- selection of non-linear switching surface [27–29] given as:
turbances in the power system, the extended state variables are defined
z1 (t ) ⎤
as: σ (t ) = sz (t ) = [ s1 s2 ] ⎡
⎢
⎣ 2 (t ) ⎥
z ⎦ (12a)
[x d]r × 1 = [ΔPd]r × 1
T
where, s1 = K − ψ (y ) A12 P, s2 = Ip × p , and K ∈ Rp × (n − p) is full state
h (t ) = [ΔPḋ ]r × 1 (7) ⏟ 
linear
non-linear
Then, the system equation (6) is modified to: feedback controller gain matrix. The ψ (y ) ∈ Rp × p is non-linear in
nature where, y is frequency deviation in areas. Whenever, the fre-
x ̇ = A x + Bu u + Eh (t ) quency excursion occurs and during the transient period, the non-linear
y = Cv x (8) function (Eq. (12a)) modifies the damping ratio/ratios of the closed
xn× 1 An × n Dn × r ⎤ loop system. Subsequently, as the frequency settles down, the linear
where, [x ] = ⎡ ⎤ , A =⎡ , part gets into action. Thus, the non-linear function comes into action
⎣[x d]r × 1⎦(n + r ) × 1 ⎢ 0r × n 0r × r ⎦
⎣ ⎥(n + r ) × (n + r )
(only when frequency deviation occurs) otherwise linear part of the
Bu 0n × r ⎤ Cm × n ⎤
Bu = ⎡ n × k ⎤ ,E=⎡ , Cv = ⎡ switching surface always remains in action. Hence, the ψ (y ) is used to
⎢
⎣ 0r × k ⎥
⎦(n + r ) × k ⎣ 1r × r ⎦(n + r ) × r ⎢ 0m × r ⎦
⎣ ⎥m × (n + r )
vary the damping ratio of the system whenever frequency deviation
Assumption 1. ( A, Bu ) is controllable implies ( A , Cv ) is observable. occurs in the system. The non-linear function ψ (y ) is characterized by
following properties:
The detailed steps in the ESO design for the system equation (8) may
be referred from [18].
• Changes from 0 (or a very low negative value) to −β as frequency i
deviation occurs in the systems, where β > 0.
x ̇ ̂ = A x ̂ + Bu u + L (y−y )̂
• Differentiable with respect to time to ensure the existence of the
i

y ̂ = Cv x ̂ (9) sliding mode.

where, x ̂ = [x n̂ × 1 [x d̂ ]r × 1]T(n + r ) × 1, and y ̂ is the estimated state of the
state variables x = [x n × 1 [x d]r × 1]T(n + r ) × 1and output respectively. The non-linear function ψi (yi ) is defined as [17,28]: where, i = 1, 2.
MatrixL ∈ R (n + r ) × r is the observer gain, which has to be designed and is βi 2
ψi (yi ) = − (e−(1 − ((yi − x 0)/(r − x 0)) )−e−1)
discussed in section 5. 1−e−1 (12b)

The non-linear functions with two-areas are given as:

4. Design of Non-linear sliding mode control
This section presents the design of non-linear sliding surface and
non-linear SMC for LFC application in power system. The non-linear
SMC improves the transient and steady state response simultaneously of
the closed-loop system in the presence of disturbances and non-linear-
ities such as GRC and governor dead band [17]. It enhances the ability
of power system to maintain steady frequency following a severe dis-
turbance between generation and load. The chattering phenomenon is
also minimized using non-linear sliding surface. Thus, devastating ef-
fects caused due to wear-out of actuators (valve) in steam turbine is
minimized. Before describing the design of non-linear SMC, first the
state-space Eq. (6) is converted into regular form [17,27,28] using
z = Tcr x ,̂ such that Tcr TcrT = I and Tcr Bu = [0(n − k ) × k [Bu2]k × k ]T , where
k = rank (Bu ) and z = [[z1](n − k ) × k [z2]k × k ]T .
The state-space equation of the system (6) in regular form is given
as:
⎡ z1̇ ⎤ = ⎡ A11 A12 ⎤ ⎡ z1 ⎤ + ⎡ 0 ⎤ u + ⎡ D1⎤ ΔPd
⎣ z2̇ ⎦ ⎣ ⎥ ⎣ z 2 ⎦ ⎣ Bu 2 ⎦
⎢ A21 A22 ⎦ ⎣0⎦
y = CTcr z = [C1 C2] z
Eq. (10) is also represented as:
z1̇ = A11 z1 + A12 z2 + D1 ΔPd
z2̇ = A21 z1 + A22 z2 + Bu2 u
Assumption 2. The matrix pair ( A, Bu ) of interconnected system is
controllable as per Assumption 1. Then it implies that the pair ( A11 , A12 )
is also controllable in regular form [17,28].
4.1. Selection of non-linear switching surface
The choice of switching surface plays a vital role in the design of
traditional SMC [15,16]. The switching surface is meant to drive system
states towards a predefined sliding surface. On the other hand, control
law guarantees area states trajectory to the sliding surface and maintain
ψ (y ) = diag (ψ1 (y ), ψ2 (y )) and β = diag (β1, β2) (12c)
where, x 0 is initial value of the corresponding state, r is a small non-
negative scalar and β is a tuning parameter that determines the selec-
tion of final damping ratio (δ2 ) along with the matrix P via LMI scheme
detailed in next subsection. It is important to note that the selection of
ψ (y ) is not unique and any function with the above properties can be
considered in the design.
4.2. Stability of non-linear switching surface
The sliding mode occurs with the state trajectories confined to
σ (t ) = 0 for t > 0 , [13], hence from (12a)
z2 (t ) = −s2−1 s1 z1 (t ) = −s1 z1 (t ) (13)
Substituting (13) into (11) gives
z1̇ (t ) = (A11 −A12 s1) z1 (t ) + D1 ΔPd (t ) (14)
During the sliding mode, the stable dynamics of the system (14) is
(10) decided mainly by switching surface s1 from (12a) where s1 consists of
linear and non-linear terms. The linear term comprising of feedback
gain K is designed at low damping ratio (δ1) to improve steady state
response and non-linear term is a function of ψ (y ) and positive definite
(11) matrix P as detailed below.
Remark 1. Consider frequency deviation begins following any
perturbation in the power system. Consequently, non-linear term is
also in action due to ψ (y ) approaching to intermediate value between
zero to −β . The overall system damping ratio now increases from low
value (δ1) to final damping ratio (δ2 ) using non-linear switching surface,
where, δ2 > δ1. At this condition, the non-linear function value reaches
−β . This leads to system dynamics damped out with minimum under/
overshoots and improved the transient response. On attaining desired
final damping ratio (δ2 ), ψ (y ) starts to decrease from non-positive
intermediate value to zero. In the next step, corresponding damping
ratio also starts to reduce from final damping ratio (δ2 ) to low damping

181
S. Prasad et al. Electrical Power and Energy Systems 104 (2019) 178–193

ratio (δ1) with minimum settling time. Therefore, closed-loop system ⎧ P>0 ⎫
operates at low damping ratio (δ1) with feedback gain K . Thus, ⎪[(A11 −A12 K )T P + P (A11 −A12 K )] < 0 ⎪
proposed non-linear SMC has ability to achieve minimum over/ ⎨ εI M ⎬
⎪ [ T ] >0 ⎪
undershoot and reduced settling time simultaneously with minimum ⎩ M εI ⎭ (24)
chattering effect.
As discussed in the above Remark 1, the switching surfaces1 switches Remark 2. The effect of unmodeled dynamics which occurs due to
from linear switching surface s1 = K (when frequency deviation is zero) actuators and sensors is reduced as SMC uses estimated states instead of
T
to combined linear and non-linear switching surface s1 = K −ψ (y ) A12 P measured states directly from the plant. Secondly, switching gains vary
(when frequency deviation is there). Hence, the stability of the Eq. (14) in a small range from initial feedback gain K to final feedback gain F. As
depends on both switching surfaces. The detailed analysis is given in the gain is always within this range, the magnitude of the chattering
the following two steps: reduces.

i. Whenever frequency deviation is zero. As the non-linear function is

zero, the switching surface s1 becomes linear s1 = K . Thus, state
equation (14) becomes:
z1̇ (t ) = (A11 −A12 K ) z1 (t ) + D1 ΔPd (t )
The feedback gain K is designed through pole assignment [13] to en-
sure that (A11 −A12 K ) should be Hurwitz matrix and the dominant poles
of the system are considered as initial low-damping ratio (δ1). The
stability of the equation (15) is analyzed using linear matrix inequality
(LMI) given in equation (16).
⎧ P>0 ⎫
T
⎩[(A11 −A12 K ) P + P (A11 −A12 K )] < 0 ⎬
⎨ ⎭
4.3. Design and stability analysis of SMC
Theorem 1. The control law for non-linear sliding surface given by:
(15)
dψ (y ) T
u (t ) = −Bu−21 [sAz (t )− A12 Pz1 + κσ (t ) + ρsign (σ (t ))−d ]̂
dt (25)
forces the trajectory of (11) to slide from any initial condition to the sliding
surface in finite time and remain on it thereafter. The scalar κ > 0 , ρ ⩽ ρ
and ρ is taken from the maximum bound of the load disturbance as given in
Assumption 3 in the next section. The variable d ̂ = x d̂ is the estimated load
disturbance of the two-area power system as given in Section 3.
(16)
Proof. For a quadratic function,

ii. Whenever, frequency deviation occurs due to the perturbation in v1 (t ) = 0.5σT (t ) σ (t ) (26)
load disturbance and parameter uncertainty etc. then, ψ (y ) increases Its time derivative,
from zero to some non-positive intermediate value (zero to −β ), and
increases the damping ratio of the system. The switching surface s1 v1̇ (t ) = σT (t ) σ̇ (t ) (27)
T
changes from linear to non-linear s1 = K −ψ (y ) A12 P as ψ (y ) increases From (12), equation (27) becomes
until ψ (y ) approaches its final value −β .
v1̇ (t ) = σT (t )[s1 z1̇ (t ) + z2̇ (t ) + s1̇ z1 (t )] (28)
Then, equation (14) can be written as:
T
Substituting (11) and time derivative of (12a) in (28) and after solving,
z1̇ (t ) = (A11 −A12 K −βA12 A12 P ) z1 (t ) + D1 ΔPd (t ) (17) it gives
The system referred to (17) decides the final damping ratio (δ2 ) and is dψ (y ) T
written as: v1̇ (t ) = σT (t ) ⎡sAz (t )− A12 Pz1 + s1 D1 ΔPd (t ) + Bu2 u (t )⎤
⎣ dt ⎦
T
z1̇ (t ) = (A11 −A12 (K + βA12 P )) z1 (t ) + D1 ΔPd (t ) (18)
Using (25), in the above equation:
Suppose the desired feedback gain isF at non-linear function value
v1̇ (t ) ⩽ σT (t )[−κσ (t )−ρsign (σ (t )) + s1 ‖D1 ‖ρ + d ]̂
-β.Then, the constraint equation is written as:
T
F = K + βA12 P (19) v1̇ (t ) ⩽ −σT (t ) ρsign (σ (t ))−λ min κ‖σ (t )‖2 + ‖σ (t )‖‖s1 ‖‖D1 ‖ρ + ‖σ (t )‖d ̂
The above equation is written as: The first term in the above equation is always negative and v1̇ (t ) < 0 if
F −K the following condition is satisfied.
T
A12 P− =0
β (20) 1
‖σ (t )‖ > [‖s1 ‖‖D1 ‖ρ + d ]̂
λ min κ (29)
To achieve the desired final damping ratio δ2 , a matrix P is needed
which satisfies (20). Now considering steps (i) & (ii), the stable Hence, the switching surface is always globally attractive.
dynamics of the system (14) can be ensured when the matrix P
Remark 3. The condition (29) can be easily satisfied by a judicious
satisfies (15) and (20) simultaneously. However, it may not always be
choice ofκ. The root locus plot of the nominal system and closed system
possible for the matrix P to satisfy constraints (15) and (18)
for area 1 is shown in the Fig. 2. It is evident that the poles movement
simultaneously. Then constraints (20) may be relaxed as [17,28]:
for the closed loop system always lies in the left half of the s-plane
‖M ‖ ⩽ ε (21) which is not the case for the nominal system.
for a sufficient small ε > 0 , where,
T F −K 5. Stability analysis
M = A12 P−
β (22)
The stability of the overall system as shown in Fig. 1(b) is presented
The above constraint equation can be converted to a linear inequality
in this section. The proposed control scheme aims at the ultimate
by using Schur complement as follows:
boundedness of all the signals of interest. The ultimate bounds can be
⎡ εI M ⎤ > 0 defined by appropriate choice of control parameter without any prior
T
⎣ M εI ⎦ (23) knowledge of the bounds of the uncertainties.
However, the optimization problem can be formulated with the relaxed Assumption 3. The disturbances and rate of change of disturbances are
constraints (23): Choose P at given β to minimize ε such as: bounded, i.e., ‖ΔPd (t )‖ ⩽ ρ and ‖h (t )‖ ⩽ λ . The error of the system is

182
S. Prasad et al. Electrical Power and Energy Systems 104 (2019) 178–193

Root Locus
40

Imaginary Axis
20 Nominal system response
0
-20
-40
-35 -30 -25 -20 -15 -10 -5 0 5 10 15
Real Axis
Root Locus
20
Imaginary Axis

Closed loop system

response
0

-20
-3.5 -3 -2.5 -2 -1.5 -1 -0.5 0 0.5
Real Axis

Fig. 2. Root locus movement of the system for area-1.

defined as: with radius ((2‖PD1 ‖ρ )/ μ) . Therefore V̇ < 0 , which proves the asymp-
totic stability of the overall system.
ex = x −̂ x ⎫ ex
̂ x
≻ e = ⎡e ⎤ = x −
̂ ΔPd ⎬
ed = d − ⎣ d⎦
(30) 5.1. Summary of the proposed control design steps
⎭
The error dynamics of the observer (9) is defined using the (8), (9) and
Detailed design of the proposed approach is summarized in the
(30):
following steps:
e ̇ = (A −LCv ) e−Eh (t ) (31)
By the virtue of Assumption 1 and Assumption 2, it is possible to select Step 1) Obtain the state-space model in regular form using transfor-
the observer gain L and feedback gain K in such way that the matrices mation from the linear model of the LFC and test the con-
(A −LCv ) and (A11 −A12 K ) will have any desired eigenvalues in the left trollability condition as given in Assumption 2.
half s-plane. Step 2) Design GESO with proper selection of the observer gainL,such
that matrix(A −LCv ) has negative real part of its eigenvalues.
The design of the feedback gain K is discussed in the previous Step 3) Chose K such that(A11 −A12 K ) has stable eigenvalues and the
section. For the observer gain design, it is assumed that all the selected dominant poles have low damping ratio.
eigenvalues of (A −LCv ) matrix are negative. Then determine the posi- Step 4) The real symmetric positive-definite matrix P is designed by
tive definite matrix P2 such that it satisfies the Lyapunov’ equation for solving the LMI (24) using “feasp” and “mincx” solvers in MA-
the given positive definite matrix Q2 : TLAB® Robust Toolbox [31] with proper selection of non-linear
(A −LCv )T P2 + P2 (A −LCv ) = −Q2 (32) function and its final valueβ.
Step 5) Use non-linear sliding surface (12), to obtain non-linear SMC
For the stability analysis of the GESO based non-linear SMC, consider a (25).
Lyapunov function,
V = z1T Pz1 + eT P2 e (33) The detailed flow chart of the proposed control technique is given in
Fig. 3.
The time derivation of the Lyapunov function,
V̇ = z1Ṫ Pz1 + z1T Pz1̇ + eṪ P2 e + eT P2 e ̇ (34) 6. Numerical simulation results and discussions

Combining the (14), (31) with (34), yields To demonstrate the performance of the proposed control scheme
through MATLAB® simulations are conducted on two-area power
V̇ = z1T [(A11 −A12 K )T P + P (A11 −A12 K )] z1 + 2z1T PD1 ΔPd
system as shown in Fig. 1(a). Two-area power system is considered as
+ 2ψ (y ) z1T PA12 A12
T
Pz1 + eT [(A −LCv )T P2 + P2 (A −LCv )] e linearized model for LFC study, its time constant varies from seconds to
−2eT P2 Eh (t ) minute. Hence, each area of the two-area power system is modeled by
two linear equivalent parameters and its parameters are given in
Form the (16) and (32), it follows as:
Table 2. Following simulations are performed to test the performance of
V̇ = 2z1T ψ (y ) PA12 A12
T
Pz1−z1T Q1 z1−eT Q2 e−2eT P2 Eh (t ) + 2z1T PD1 ΔPd designed controller:

V̇ ⩽ 2z1T ψ (y ) PA12 A12

T
Pz1−μ‖z1 ‖2 + 2‖z1T ‖‖PD1 ‖‖ΔPd ‖−eT Q2 e
• Validation of GESO with proper selection of observer gainL.
−2‖eT ‖‖P2 E‖‖h (t )‖ • Design GESO based non-linear SMC and test for step load dis-
turbance.
Therefore, we have
• Comparisons of proposed controller with two-layer ADRC [24] at
V̇ ⩽ 2z1T ψ (y ) MM T z1−‖z1 ‖[μ‖z1 ‖−2‖PD1 ‖ρ ]−eT Q2 e−2‖eT ‖‖P2 E‖λ random load disturbances.
T
As PA12 A12 P = MM T ⩾ 0 and the function ψ (y ) is negative by defi- • Validation of the proposed control scheme on large power system at
random load disturbance.
• Effectiveness of the proposed controller with GRC, governor dead-
T
nition, therefore matrix 2ψ (y ) PA12 A12 P is a negative semi-definite
square matrix and system state trajectories enter into bounded region
band and random load disturbances.

183
S. Prasad et al. Electrical Power and Energy Systems 104 (2019) 178–193

Start 6.1.2. Performance of GESO based non-linear SMC

• Input the system parameters and
outputs
• Input the estimated parameters of
the GESO:
• Give desired low damping ratio :
Estimated states of the GESO are used to test the performance of the
non-linear SMC design. The frequency deviation in both the areas, non-
linear function ψ1 (y ),switching surface σ1 for area-1 are shown in Fig. 5.
It indicates that the non-linear SMC provides additional damping
characteristic to the system and reduces the settling time with
1

• Define the accuracy terms: 0

minimum under/overshoot and the chattering. Hence, the GESO based
• Select the non-linear function and non-linear SMC is effective and realizable.
its parameters: Remark 4: In the GESO based non-linear SMC, the system frequency
deviation stabilizes rapidly even in the presence of the step load dis-
Initial guess of K turbances. The norm of the instantaneous gain matrix varies in between
initial gain matrix norm ‖K ‖ = 258.5344 and final desired gain matrix
norm ‖F‖ = 285.3934 . Hence, small range of gain matrices are indication
Modify K by No Is desired of reduced chattering effect in the control signal. So, proposed con-
altering pole damping ratio
location achieved?
troller is realizable.

Yes 6.2. Performance of GESO based non-linear SMC in the presence of

uncertainty
Change the No
Is (24) feasible?
value of β Further, the some errors occur in the power system due to linear-
Yes
ization, unmodeled dynamics and etc., which may lead to the un-
certainty of the parameters. In order to illustrate the robustness of the
No observer based non-linear SMC, the subsystem parameters Ri , Tgi and Tpi
0 ?
are obtained by changing by ± 20% from their typical values. The un-
Yes certainty in the governor valve position is considered and represented
by cosine function as follows:
Obtain P using (24) The deviations in frequency with the observer based non-linear SMC
and non-linear SMC [33] on load disturbance are shown in Fig. 6. It is
seen that the frequency deviation response oscillates at same frequency,
Obtain non-linear but the magnitude of oscillation is found comparatively low in case of
switching surface (12) observer based non-linear SMC against [33]. This demonstrates that the
proposed method has satisfactory performance even in presence of the
matched uncertainties and load disturbances.
Obtain control law (25)
ΔA
End ⎡0 0 0 0 0 0 0 0 0 ⎤
⎢0 0 0 0 0 0 0 0 0 ⎥
Fig. 3. Flow chart of the proposed controller. ⎢0 0 0 0 0 0 0 0 0 ⎥
⎢0 0 0 0 0 0 0 0 0 ⎥
=⎢
⎢
0 40cos(t ) 0 1.6cos(t ) 1.6cos(t ) 0 0 0 0 ⎥
⎥
Table 2 ⎢0 0 0 0 0 0 0 0 0 ⎥
Plant Parameters. ⎢0 0 0 0 0 0 0 0 0 ⎥
⎢0 0 0 0 0 0 0 0 0 ⎥
Area Tch TG Kp Tp R B KE T1 ⎢0
⎣ 0 0 0 0 0.46cos(t ) 0 6cos(t ) 6cos(t ) ⎥
⎦
1 0.3 0.1 1.0 10 0.05 41 0.5
2 0.17 0.4 0.67 8 0.05 81.5 0.5 3.77
6.3. Comparison analysis of the proposed method with non-linear SMC and
two-layer ADRC
6.1. Validation of GESO, proposed method and comparison study in the
presence of step disturbance Now, a comparison study is tested on the two-area thermal power
system in the presence of the step load disturbance and parametric
In this subsection, the convergence of GESO and comparisons of uncertainty. The variations in frequency are shown in Fig. 7. Thus, it is
non-linear SMC, GESO based non-linear SMC and two-layer ADRC evident that the proposed control scheme is indeed capable of be robust
through simulation is tested by applying the load disturbances of against parameter uncertainties and achieve superior performance
magnitude ΔPd1 = 0.045, ΔPd2 = 0.02 (p.u.MW) and system uncertainty compared to the two-layer ADRC in [24].
to the two-area power system. The GESO gains obtained using
MATLAB® function “place” with the selection of observer real pole lo- 6.4. Performance of proposed method and comparison study at random
cations are as [−30−21−98−22−25−36−22−39−45−50−55]. load disturbances/unmatched uncertainty:

6.1.1. Performance of GESO
First, the convergence of GESO is tested without the proposed
controller. The actual and estimated frequency deviation in both areas,
tie-line power deviation and turbine-generator output power deviation
in area-1 are shown in Fig. 4. It is evident that the GESO converges
quickly and provides accurate estimates. The said observer ensures
accurate trajectory of the actual states and thus maintains tracking
error converge to zero.
Now, the proposed controller is tested by comparison with non-
linear SMC [33] and traditional two-layer ADRC [24] at random load
disturbance. Detailed simulations and discussions are given in the fol-
lowing subsections:
6.4.1. Effectiveness of proposed method and comparison with two-layer
ADRC at random load disturbances
The actual random load disturbance applied in both areas as shown

184
S. Prasad et al. Electrical Power and Energy Systems 104 (2019) 178–193

0.04 0.05
Actual
Estimated
0.02

Δ f2 (Hz)
Δ f1 (Hz)
0 0

-0.02

-0.04 -0.05
0 10 20 30 40 50 0 10 20 30 40 50
0.6 0.8

0.6
0.4
0.4
Δ Ptie (puMW)

Δ Pm1 (puMW)
0.2
0.2

0 0

-0.2
-0.2
-0.4

-0.4 -0.6
0 10 20 30 40 50 0 10 20 30 40 50
Time (sec) Time (sec)

Fig. 4. Actual and estimated system variables.

-4 -4
x 10 x 10
5 5

0
0

-5
Δ f1 (Hz)

Δ f2 (Hz)

-5
-10

-10
-15

-20 -15
0 10 20 30 40 50 0 10 20 30 40 50
-3
x 10
1
0

0.5
ψ1(y) (pu)

-5
σ 1 (pu)

-10
-0.5

-1 -15
0 10 20 30 40 50 0 10 20 30 40 50
Time (sec) Time (sec)

Fig. 5. Δfi , non-linear switching surface and non-linear function response at step load disturbance.

in sub-figure of Fig. 8 (solid blue line). The estimated load disturbances
by GESO (dotted red line) are also compared with actual load dis-
turbances (solid blue line) in the system as shown in sub-figure of Fig. 8.
It suggests that the GESO is capable to estimate the load disturbances in
the system accurately using measured signals. The frequency deviation
in both areas is shown in sub-figure of Fig. 8. It can be seen that the
frequency deviation with proposed controller is kept at minimum
magnitude without any oscillation following the load disturbance. It
suggests that the proposed control scheme is capable to estimate and
compensate the load disturbances in the system accurately using mea-
sured signals compared with two-layer ADRC [24]. Similarly, tie-line
power deviation, turbine-generator power deviation in area-1 and
control effort required to achieve frequency deviation settle down
quickly using proposed and two-layer ADRC are shown in Fig. 9. The
tie-line power deviation and turbine-generator power deviation also has
negligible oscillatory response. In other words, the proposed controller

185
S. Prasad et al. Electrical Power and Energy Systems 104 (2019) 178–193

-4 -4
x 10 x 10
5 5

0 0

Δ f1 (Hz)

Δ f2 (Hz)
-5 -5

-10 Observer based Non-linear SMC -10

Non-linear SMC

-15 -15
0 10 20 30 40 50 0 10 20 30 40 50
Time (sec) Time (sec)

Fig. 6. Δfi , non-linear switching surface and non-linear function response in presence of the uncertainty and step load disturbance.

improves the system damping characteristic. 6.5. Effectiveness of proposed method in presence of GRC and governor
It should be noted that the chattering effects are eliminated com- dead band
pletely due to fact that the non-linear switching law is the basis of the
robustness of the dynamic response. Thus, the proposed control scheme This subsection presents analysis with inclusion of
performs well compared to two-layer ADRC [24] in the presence of the ± 0.0023puMW /sec generation rate constraint (GRC) limit [15] in both
random load disturbances. generator units and governor dead-band [30]. The frequency deviation,
tie-line power deviation and turbine-generator power deviation in area-
1 with GRC and govornor dead-band are shown in Fig. 11.
6.4.2. Comparison of proposed method with non-linear SMC [33] and two- It suggests that effectiveness of the proposed control scheme is
layer ADRC [24] found satisfactory even in the presence of GRC and governor dead-
In this subsection, the proposed control scheme is tested and com- band. The integral square error (ISE) is obtained for frequency devia-
pared with the non-linear SMC [33] and two-layer ADRC [24] in the tion in area-1 with GRC and governor dead-band as shown in Table 3. It
presence of random load disturbance as shown in Fig. 10. The random is evident that the proposed controller performs satisfactorily even in
load disturbance patterns are same as given in Fig. 8. The variation in presence of GRC and governor dead-band.
frequency deviation remains consistent as discussed above by the pro-
posed control scheme.

-3 -3
x 10 x 10
1 1

0.5
0
0
Δ f1 (Hz)

Δ f2 (Hz)

-1 -0.5
Two-layer ADRC
-1
Non-linear SMC
-2
Proposed Method -1.5
-3 -2
0 10 20 30 40 50 0 10 20 30 40 50
Time (sec) Time (sec)
a) Frequency deviations in the presence of step load disturbance
-3 -3
x 10 x 10
1 1
0.5 0.5
0 0
Δ f1 (Hz)

Δ f2 (Hz)

-0.5 -0.5
-1 Two-layer ADRC -1
-1.5 Non-linear SMC
-1.5
Proposed Method
-2 -2
0 10 20 30 40 50 0 10 20 30 40 50
Time (sec) Time (sec)
b) Frequency deviations in the presence of uncertainty and step load disturbance
Fig. 7. A comparison study with non-linear SMC and two-layer ADRC.

186
S. Prasad et al. Electrical Power and Energy Systems 104 (2019) 178–193

-3 -3
x 10 x 10
4 4
Two-layer ADRC
Proposed control
2 2

Δ f1 (Hz)

Δ f2 (Hz)
0 0

-2 -2

-4 -4
0 10 20 30 40 50 0 10 20 30 40 50

0.06 0.05
Estimated
0.04
Actual

Δ Pd2 (puMW)
0.02
Δ Pd1 (puMW)

0 0

-0.02

-0.04

-0.06 -0.05
0 10 20 30 40 50 0 10 20 30 40 50
Time (sec) Time (sec)

Fig. 8. Variation in Δfi , actual and estimated load disturbance.

0.1 0.05

0.05
ΔP m1 (puMW)
ΔP tie (puMW)

0 0

-0.05

-0.1 -0.05
0 10 20 30 40 50 0 10 20 30 40 50
0.1 0.1

0.05 0.05
u2 (pu)
u1 (pu)

0 0

-0.05 -0.05
Proposed control
Two-layer ADRC
-0.1 -0.1
0 10 20 30 40 50 0 10 20 30 40 50
Time (sec) Time (sec)

Fig. 9. Variation in ΔPtie , ΔPm1 and control effort at random load disturbances.

6.6. Comparison study of the proposed method with non-linear SMC [33] 6.7. Disturbance rejection capability of proposed control scheme at random
and two-layer ADRC in presence of GRC[24] and governor dead band loads disturbances:

This scenario is analyzed with inclusion of generation rate con-
straint (GRC) and governor dead band (GDB) effect in two area power
system. Fig. 12 illustrates the comparative dynamic responses of the
proposed control scheme, non-linear SMC [33] and two-layer ADRC
[24], considering GRC and governor dead band. The variaion in fre-
quency deviations remains consistent within limits as discussed above
by the proposed ccontrol scheme.
The effectiveness of the disturbance compensation accuracy is
shown in the Fig. 13. Here, the estimated states and load disturbances
by the GESO are considered as input to non-linear SMC for the case of
disturbance compensation (solid line). While in the uncompensated
case, only estimated states are considered as input to non-linear SMC
(dotted line). It is seen that the proposed controller with disturbance
compensation has minimum under/overshoots and as well as settling
time. This demonstrates that disturbances are satisfactorily compen-
sated without implementation of EID and internal model controller.

187
S. Prasad et al. Electrical Power and Energy Systems 104 (2019) 178–193

-3 -3
x 10 x 10
4 6

2 4

0 2
Δ f 1 (Hz)

Δf2 (Hz)
-2 0
Two-layer ADRC
-4 Non-linear SMC
-2
Proposed control
-6 -4
0 10 20 30 40 50 0 10 20 30 40 50
Time (sec) Time (sec)
Fig. 10. A comparison study of proposed method with non-linear SMC [33] and two-layer ADRC [24].

-3 -3
x 10 x 10
4 4

2 2
Δ f1 (Hz)

Δ f2 (Hz)
0 0

-2 -2

-4 -4
0 10 20 30 40 50 0 10 20 30 40 50
0.1 0.06

0.04
0.05
Δ Pm1 (puMW)

0.02
Δ Ptie (puMW)

0 0

-0.02
-0.05
-0.04

-0.1 -0.06
0 10 20 30 40 50 0 10 20 30 40 50
Time (sec) Time (sec)

Fig. 11. Δfi , ΔPtie , ΔPm1 responses with GRC and governor dead-band at random load disturbance.

Table 3 6.8. Effectiveness of the proposed control method on thermal-hydro two-

ISE for frequency deviation in area-1. area power system
Performance index GRC GDB GRC & GDB
Finally, the GESO based non-linear SMC is demonstrated for non-
−04 −05
ISE 3.7101 × 10 1.9547 × 10 1.9947 × 10−05 minimum phase system (i.e. combined thermal-hydro two-area power
system). The control area-1 is a linearized non-reheat turbine type

-3 -3
x 10 x 10
4 6

2 4

0 2
Δ f1 (Hz)

Δ f2 (Hz)

-2 0

-4 Two-layar ADRC -2
Non-linear SMC
-6 Proposed Method -4

0 10 20 30 40 50 0 10 20 30 40 50
Time (sec) Time (sec)
Fig. 12. A comparison study of proposed method with non-linear SMC [33] and two-layer ADRC [24] in presence of GRC and GDB.

188
S. Prasad et al. Electrical Power and Energy Systems 104 (2019) 178–193

-3 -3
x 10 x 10
4 6
without compensation
with compensation via GESO
4
2

Δ f1 (Hz)

Δ f2 (Hz)
0
0

-2
-2

-4 -4
0 10 20 30 40 50 0 10 20 30 40 50

0.06 0.1

0.04
0.05

Δ Pm1 (puMW)
ΔPtie (puMW)

0.02
0
0
-0.05
-0.02

-0.04 -0.1
0 10 20 30 40 50 0 10 20 30 40 50
Time (sec) Time (sec)
Fig. 13. Simulation with and without disturbance compensation.

thermal power generation system (similar to area-1 Fig. 1(a)) with Table 4
different parameters. The linearized hydro power generation system in Thermal-hydro Plant Parameters [34].
area-2 is shown in Fig. 14. The parameters of the thermal-hydro two- Hydro unit RHY T12,HY KP,HY TP,HY TW T22,HY Ki1 & Ki2 TR
area power system are given Table 4 [34]. The observer real poles are area-2
selected as −30, −21, −98, −22, −25, −36, −22, −39, −45, − 2.4 48.7 s 120 20 s 1.0 s 5.0 s 0.009 0.6 s
50, −55, −80 . Thermal unit RTH TG1 KP,TH TP,TH TT1 Ks12 BHY BTH
The dynamic responses of combined thermal-hydro two-area power area-1
system using proposed control scheme are given in Fig. 15 at same 2.5 0.08 s 115 20 s 0.33 s 0.545 0.409 0.425
random load disturbances as described in previous subsection. It is
evident that the frequency deviation with proposed controller is kept at
linearities for minimum phase system as well as non-minimum phase
minimum magnitude without any oscillation following the load dis-
system. The performance indices; integral square error (ISE), integral
turbance. This suggests that the proposed control scheme is capable to
absolute error (IAE), integral time absolute error (ITAE) computed for
estimate and compensate the load disturbances accurately using mea-
frequency deviations in area-1 are given in Table 5. These indices ob-
sured signals. In other words, the proposed controller improves the
tained from proposed method are comparatively low. Thus, it is evident
system damping characteristic.
that the proposed controller is robust and effective (see Table 5).
The control effort with reduced chattering and estimated load dis-
turbances by GESO is shown in Fig. 16. It is evident that the proposed
control scheme is effective in the suppression of the chattering phe- 6.9. Validation of the proposed control method on large power system
nomenon even in the presence of the disturbances, system non-
The proposed control strategy is validated on the New England real

Thermal unit area-1

u1 Δ1
Thermal-hydro
GESO based non-linear SMC plant states, input
and output
−
Tie-line Δ tie12
Ks12
+
u2
− ΔE2 + 1 ΔX gh 2 ΔX g 2 Δ − K
1 + sTR 1 − sT g2
− Ki 2 + 1 + sT12 ,HY 1 + sT
+ − 1 + sT22 ,HY 1 + 0.5sT + −

Δ 2
BHY RHY
Δ d2

Fig. 14. Thermal-hydro two-area power system [34].

189
S. Prasad et al. Electrical Power and Energy Systems 104 (2019) 178–193

0.1 0.1

0.05 0.05

Δ f1 (Hz)

Δ f2 (Hz)
0 0

-0.05 -0.05
0 10 20 30 40 50 0 10 20 30 40 50
Time (sec) Time (sec)
0.02 0.1

0.01 0.05
Δ Ptie12 (pu Mw)

Δ Pg1 (pu Mw)

0 0

-0.01 -0.05

-0.02 -0.1
0 10 20 30 40 50 0 10 20 30 40 50
Time (sec) Time (sec)
Fig. 15. Deviations in frequency, tie-line power and power generations of thermal-hydro two-area power system.

test system (IEEE 39 bus system) [32] at random load disturbance and Table 5
parameters are given in Table 6. This IEEE 39 bus system is the well- Thermal-hydro Plant Performance Indices.
known 10-machine New-England power system. The single-line diagram Methods ISE IAE ITAE
of the IEEE 39 bus system with tie-line (bold line) is shown in Fig. 17.
The control area-1 and area-2 have three generators, and area-3 has four Proposed Control Scheme 0.0029 0.0463 0.0319
generators, which are operating in parallel. Each area generator is op- Non-linear SMC [33] 0.0033 0.0458 0.0320

erated in non deregulated environment. The generators in each area are

represented as a single equivalent generator and order of the system and
system is given in Table 3. The observer real poles are selected as −
observer are 9 and 10 respectively. Hence, total order of the system and
30, −21, −98, −22, −25, −36, −48, −39, −45, −50 for area-1,
observer are 27 and 30 respectively. The connected loads in area-1, area-
−30, −21, −98, −42, −25, −36, −29, −39, −45, −90 for area-2 and
2 and area-3 and total power generation are 265.5 MW, 233 MW,
−30, −51, −98, −42, −25, −36, −29, −39, −45, −90 for area-3
125 MW and 842 MW respectively. The parameters of the IEEE 39 bus

0.4 4

0.2 2
u1 (pu)

u2 (pu)

0 0

-0.2 -2

-0.4 -4
0 10 20 30 40 50
0 10 20 30 40 50
Time (sec) Time (sec)
1 0.05

0
0
ψ1 (y) (pu)

Δ Pd1 (pu)

-1
-0.05
-2
-0.1 ΔP d1 Estimated
-3

-4 -0.15
0 10 20 30 40 50 0 10 20 30 40 50
Time (sec) Time (sec)

Fig. 16. Control efforts, non-linear function for area-1 and estimated disturbance for area-1 of thermal-hydro two-area power system.

190
S. Prasad et al. Electrical Power and Energy Systems 104 (2019) 178–193

Table 6
IEEE 39 Parameters in per unit [32]
Generators (bus No.) M (Moment of inertia of the D (Generator’s damping Tg (Governor time Tch (Turbine-generator Linearization parameters of the governor
generator) ratios) constant) time constant) characteristic

Kt eT r

G1 (39) 3.0 4.0 0.25 0.2 250 39.4 19

G2 (31) 2.5 4.0 0.25 0.2 250 39.4 19
G3 (32) 4.0 6.0 0.25 0.2 250 39.4 19
G4 (33) 2.0 3.5 0.25 0.2 250 39.4 19
G5 (34) 3.5 3.0 0.25 0.2 250 39.4 19
G6 (35) 3.0 7.5 0.25 0.2 250 39.4 19
G7 (36) 2.5 4.0 0.25 0.2 250 39.4 19
G8 (37) 2.0 6.5 0.25 0.2 250 39.4 19
G9 (38) 6.0 5.0 0.25 0.2 250 39.4 19
G10 (30) 4.0 5.0 0.25 0.2 250 39.4 19

respectively.
The variations in frequency, tie-line power and turbine-generator
power responses at different random load disturbance are shown in
Fig. 18. The controller maintains robust performance and grid stability
by minimizing the effects of the load disturbances. Thus, the proposed
controller achieves the nominal frequency and thereby reduces tie-line
power exchange deviation to zero with minimum under/overshoots and
reduced settling time.
The performance indices are also calculated for the New England
power system as given in Table 7. It evident that the system dynamics
exhibit better performance as evidenced by overshoots/undershoots,
settling time and transient frequency oscillations with the proposed
control scheme.
6.10. Interpretations on the simulation studies
proposed GESO based non-linear SMC. The system dynamics of the
proposed controller is compared to two-layer ADRC [24], the work
presented in [33] and also validated on large power system [32] and
non-minimum phase thermal-hydro two-area power system. The un-
certainties in load disturbance of the power system affect the fixed
frequency bandwidth of low-pass filter which may degrade the dis-
turbance rejection capability of EID. The performance of the proposed
controller in terms of under/overshoot and settling time are simulta-
neously significantly reduced. The proposed controller is also found
satisfactory even in presence of GRC, governor dead-band and random
load disturbances. The detailed convergence analysis for the proposed
control scheme demonstrates the asymptotic stability of the system. The
designed control signal provides very less chattering effect signifying
reduced wear-out of actuators (valve) in steam turbine. Thus, very
small deviations in frequency with the proposed non-linear SMC have
less impact on the plant reserve capacity and power market.

Simulation studies are carried out to verify the effectiveness of the

G8
37 Area 2
G10
26 28 29
30
25
38
2
G9
27

18
1 17

24
G6
3 Tie-lines 16 35
15
G1 22
4
Area 1 21
5 14
12 19
39
6
23
11 13 33
7
20 G4
36
8 10
G7
34
9 32
31 G5 Area 3
G3
G2

Fig. 17. New England test system (IEEE 39 bus system).

191
S. Prasad et al. Electrical Power and Energy Systems 104 (2019) 178–193

ΔPd1 ΔPd2 ΔPd3 Area 1 Area 2 Area 3

0.06 0.04

0.04 0.02

Δ Pdi (puMW)

Δ fi (Hz)
0.02 0

0 -0.02

-0.02 -0.04

-0.04 -0.06
0 10 20 30 40 50 0 10 20 30 40 50
-3
x 10
0.03 5

0.02
Δ Ptiei (puMW)

Δ Pmi (puMW)
0.01 0

-0.01 -5

-0.02

-0.03 -10
0 10 20 30 40 50 0 10 20 30 40 50
Time (sec) Time (sec)
Fig. 18. Simulated results of New England large system [32] at random load disturbance.

Table 7
New England Plant Performance Indices.
Methods ISE IAE
Proposed Control Scheme 0.0011 0.0874
Non-linear SMC [33] 0.0013 0.0327
7. Conclusion
In this study, the proposed GESO based non-linear SMC scheme has
been designed for the LFC. The GESO was used to estimate the states
and disturbances in the system. The design of non-linear SMC was based
on estimated states and disturbance using GESO. The LMI technique
applied in the design of the non-linear SMC optimized the feedback gain
for desired damping ratio to achieve the improved performance. The
system responses: frequency deviation, tie-line power deviation, tur-
bine-generator power deviation and control efforts have been evaluated
in two-area and large power system in the presence of the load dis-
turbances and system non-linearities such as GRC and governor dead-
band. On one hand, the proposed control scheme maintained minimum
under/overshoot and reduced settling time and on other hand, the load
disturbance compensation was also achieved. The simulated results
demonstrate feasibility, effectiveness and robustness of the proposed
control scheme even in the presence of the uncertainties/non-linear-
ities. Based on the study presented in this paper, future work will focus
on the load disturbance compensation for the power system in the
presence of the sensor noise etc. and communication delay.
Appendix A. Supplementary data
Supplementary data associated with this article can be found, in the
online version, at http://dx.doi.org/10.1016/j.ijepes.2018.06.035.
References
[1] Kundur P. Power system stability and control. New York: McGraw-Hill; 1994.
[2] Ibrabeem PK, Kothari DP. Recent philosophies of automatic generation control
strategies in power systems. IEEE Trans Power Syst 2005;20(1):346–57.
[3] Sondhi Swati, Hote Yogesh V. Fractional order PID controller for perturbed load
frequency control using Kharitonov’s theorem. Int J Electr Power Energy Syst June
2016;78:884–96.
ITAE [4] Shayeghi H, Shayanfar HA, Jalili A. Load frequency control strategies: a state-of-the-
art survey for the researcher. Energy Convers Manage 2009;50(2):344–53.
0.1157 [5] Alizadeh Mousavi O, Cherkaoui R, Bozorg M. Blackouts risk evaluation by Monte
0.1287 Carlo Simulation regarding cascading outages and system frequency deviation.
Electr Power Syst Res August 2012;89:157–64.
[6] Elsisi M, Soliman M, Aboelela MAS, Mansour W. Bat inspired algorithm based op-
timal design of model predictive load frequency control. Int J Electr Power Energy
Syst December 2016;83:426–33.
[7] Bevrani H, Mitani Y, Tsuji K. Robust decentralized AGC in a restructured power
system. Energy Convers Manage 2004;45(15):2297–312.
[8] Bhatt P, Roy R, Ghoshal SP. Optimized multi-area AGC simulation in restructured
power systems. Int J Electr Power Energy Syst 2010;32(4):311–22.
[9] Dhillon Sukhwinder Singh, Lather JS, Marwaha S. Multi objective load frequency
control using hybrid bacterial foraging and particle swarm optimized PI controller.
Int J Electr Power Energy Syst 2016;79:196–209.
[10] Ahmadi Adel, Aldeen Mohammad. Robust overlapping load frequency output
feedback control of multi-area interconnected power systems. Int J Electr Power
Energy Syst 2017;89:156–72.
[11] Zribi M, Al-Rashed M, Alrifai M. Adaptive decentralized load frequency control of
multi-area power systems. Int J Electr Power Energy Syst 2005;27(8):575–83.
[12] Hsu KC. Decentralized variable structure model-following adaptive control of in-
terconnected systems with series nonlinearities. Int J Syst Sci 1998;29:365–73.
[13] Siaramakrishana AY, Hariharm MV, Srisailam MC. Design of variable structure
load-frequency controller using pole assignment technique. Int J Control Sept.
1984;40(3):487–98.
[14] Khodabakhshian A, Edrisi M. A new robust PID load frequency controller. Control
Eng Pract 2008;16:1069–80.
[15] Mi Yang, Yang Fu, Li Dongdong, Wang Chengshan, Loh Poh Chiang, Wang Peng.
The sliding mode load frequency control for hybrid power system based on dis-
turbance observer. Int J Electr Power Energy Syst 2016;74:446–52.
[16] Vrdoljak K, Peric N, Petrovic N. Sliding mode based load frequency control in power
systems. Electr Power Syst Res 2010;80(5):514–27.
[17] Prasad S, Purwar S, Kishor N. H-infinity based non-linear sliding mode controller
for frequency regulation in interconnected power systems with constant and time-
varying delays. IET Gener Transm Distrib 2016;10(11):2771–84.
[18] Han J. From PID to active disturbance rejection control. IEEE Trans Ind Electron
2009;56(3):900–6.
[19] Feng Hongyinping, Guo B-Zhu. Active disturbance rejection control: old and new
results. Ann Rev Control 2017. http://dx.doi.org/10.1016/j.arcontrol.2017.05.003.
[20] Cortés-Romero John, Jimenez-Triana Alexander, Coral-Enriquez Horacio, Sira-
Ramírez Hebertt. Algebraic estimation and active disturbance rejection in the
control of flat systems. Control Eng Pract 2017;61:173–82.
[21] Li SL, Yang X, Yang D. Active disturbance rejection control for high pointing ac-
curacy and rotation speed. Automatica 2009;45(8):1854–60.
[22] Sun Li, Hua Qingsong, Li Donghai, Pan Lei, Xue Yali, Lee Kwang Y. Direct energy
balance based active disturbance rejection control for coal-fired power plant. ISA
Trans July 2017. http://dx.doi.org/10.1016/j.isatra.2017.06.003.

192
S. Prasad et al.
[23] Tan Wen, Hao Yuchun, Li Donghai. Load frequency control in deregulated en-
vironments via active disturbance rejection. Int J Electr Power Energy Syst
2015;66:166–77.
[24] Liu Fang, Li Yong, Cao Yijia, She Jinhua, Wu M. A two-layer active disturbance
rejection controller design for load frequency control of interconnected power
system. IEEE Trans Power Syst 2016;31(4):3320–1.
[25] Yan B, Tian Z, Shi S, Weng Z. Fault diagnosis for a class of nonlinear systems via
ESO. ISA Trans 2008;47(4):386–94.
[26] Shihua Li, Jun Yang, Wen-Hua Chen, Xisong Chen. Generalized extended state
observer based control for systems with mismatched uncertainties. IEEE Trans Ind
Electron 2012;59(12):4792–808.
[27] Utkin VI. Sliding mode in control optimization. Berlin Heidelberg: Springer-Verlag;
1992.
[28] Prasad S, Purwar S, Kishor N. On design of a non-linear sliding mode load frequency
control of interconnected power system with communication time delay. 2015 IEEE
Conference on Control Applications (CCA), Sydney, Australia 2015. p. 1546–51.
193
Electrical Power and Energy Systems 104 (2019) 178–193
[29] Chen BM, Lee TH, Peng K, Venkataramanan V. Composite nonlinear feedback
control for linear systems with input saturation: theory and application. IEEE Trans
Autom Control 2003;48(3):427–39.
[30] Tripathy SC, Balasubramanian R, Nair PSC. Effect of superconducting magnetic
energy storage on automatic generation control considering governor dead band
and boiler dynamics. IEEE Trans Power Syst 1992;48(3):1266–73.
[31] Boyd S, Ghaoui EL, Feron E, et al. LMI in systems and control theory. Philadelphia:
SIAM; 1994. [version 5.2].
[32] Bevrani H. 2 ed. Robust Power System Frequency Control Springer; 2014.
[33] Prasad Sheetla, Purwar Shubhi, Kishor Nand. Non-linear sliding mode load fre-
quency control in multi-area power system. Control Eng Pract 2017;61:81–92.
[34] Bhatt Praghnesh, Ghoshal SP, Roy Ranjit. Coordinated control of TCPS and SMES
for frequency regulation of interconnected restructured power systems with dy-
namic participation from DFIG based wind farm. Renewable Energy
2012;40:40–50.