Electric Power Systems Research 160 (2018) 348–361

Contents lists available at ScienceDirect

Electric Power Systems Research

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

An improved dynamic performance of bidirectional SEPIC-Zeta

converter based battery energy storage system using adaptive sliding
mode control technique
P.K. Gayen ∗ , P. Roy Chowdhury, P.K. Dhara
Electrical Engineering Department, Kalyani Government Engineering College, Kalyani, Nadia, 741235, India

a r t i c l e i n f o a b s t r a c t

Article history: An improved dynamic performance of DC–DC bidirectional SEPIC-Zeta converter based battery energy
Received 27 July 2017 storage system (BESS) has been achieved using adaptive sliding mode control (SMC) technique. The micro-
Received in revised form 14 January 2018 grid concept is gaining importance for integrating various nonconventional energy sources throughout
Accepted 22 March 2018
the world. Nowadays, both DC and AC micro-grid systems are employing BESS for getting various advan-
tages such as minimization of power fluctuations, excess or emergency power support etc. In this
Keywords:
application, controlled charging and discharging operation of the concerned battery bank is extremely
Micro-grid
needed. In this scenario, DC to DC bidirectional converter such as combined SEPIC-Zeta converter is con-
Battery energy storage system
Bidirectional SEPIC-Zeta converter
sidered to facilitate dynamic charging and discharging operation of the concerned battery. Conventional
Adaptive sliding mode control proportional-integral-derivative (PID) controller based control strategy for the bidirectional converter
can not always meet the dynamic performance requirements under charging and discharging operation
cycles of the battery. Therefore, adaptive SMC action has been proposed to improve both transient and
steady state behavior of the BESS. In addition to this, the proposed model can be operated satisfactorily
under changed operating condition. The proposed concept is validated via various simulation studies
using MATLAB-SIMULINK software. The experimental study is also performed on a prototype of the BESS
to validate adaptive SMC case.
© 2018 Elsevier B.V. All rights reserved.

1. Introduction mentioned power management activities are discussed in several

literatures [9,10]. In this respect, dynamic charging and discharging
In practice, the various distributed generations (DGs) [1] are operations of concerned rechargeable battery bank are extremely
integrated into low voltage (LV) electrical network [2], commonly necessary. The life and performance of the battery bank [11] signif-
known as micro-grid [3]. To begin with, the micro-grid is used to icantly depends on the dynamic nature of charging and discharging
supply local loads. In recent scenario, it is also integrated into grid. current and the degradation phenomenon of grid connected battery
Thus, both grid connected and grid isolated operations of micro- is also discussed in Ref. [12].
grid are performed in practice. In this respect, various architectures In practical situation, various DC–DC bidirectional converters
[4] of micro-grid are available in existing literatures. In general, [13–15] are used to support both the charging and discharging
micro-grid can be classified as DC and AC micro-grid. The battery actions of battery. In this regard, the used primitive converters are
energy storage system (BESS) is used to store surplus electrical buck-boost topology, Cuk converter [16]. In advancement, SEPIC
energy [5,6], generated by DGs, which fulfills the requirements of [17] and Zeta converters [18] are developed for their advantages as
any micro-grid. These requirements are minimization of the power discussed in the paper [19]. The combined SEPIC-Zeta bidirectional
fluctuations [7] caused by DGs and supply of the extra or emer- DC–DC converter is chosen in this paper so that it can support wider
gency electrical power demand [8] whenever necessary. The above range of operation of BESS. Besides, the current and voltage ripples
of bidirectional SEPIC-Zeta converter are lesser in comparison with
other bidirectional converters. It is expected that the dynamic con-
trolling action of the DC–DC bidirectional converter in BESS should
∗ Corresponding author. be suitably done so that charging and discharging current settles to
E-mail addresses: [email protected], [email protected] the respective desired values quickly with less transient peaks. At
(P.K. Gayen), [email protected] (P. Roy Chowdhury),
first, traditional proportional-integral-derivative (PID) based con-
[email protected] (P.K. Dhara).

https://doi.org/10.1016/j.epsr.2018.03.016
0378-7796/© 2018 Elsevier B.V. All rights reserved.
 P.K. Gayen et al. / Electric Power Systems Research 160 (2018) 348–361 349

Nomenclature

Abbreviations
AC Alternating current
BESS Battery energy storage system
DC Direct current
PID Proportional-integral-derivative
PWM Pulse width modulation
SM Sliding mode
SMC Sliding mode control
SOC State-of-charge

Symbol
A, B, C, D Constants for general state matrix
A1 , B1 , C1 , D1 Constants for on-state state matrix
A2 , B2 , C2 , D2 Constants for off-state state matrix
C1 , C3 Battery side and output side capacitor respectively
C2 , e Flying Capacitor and error signal respectively
DS , DZ Anti-parallel body diodes of SZ and SS respectively
d Duty ratio
Ibat , Ibus Battery side and output side current respectively Fig. 1. AC and DC micro-grid with DC–DC bidirectional converter based BESS.
Io , IC Output current and capacitor current respectively
IC1 , IC3 Battery side and output side capacitor current The system studied is discussed in “Section 2”. The state space
respectively modeling of SEPIC and Zeta converter is described in “Section 3”.
IL1 , IL2 Current through the inductors of the bidirectional The PID controller based system is presented in “Section 4”. “Section
converter 5” is devoted to present proposed adaptive sliding mode control
k1 , k2 , k3 Adjusted coefficients for adaptive SMC strategy for BESS with the corresponding block diagram. “Section 6”
kP , kI , kD Gains of PID controller shows the results of various simulation and experimental studies.
L1 , L2 Inductors of bidirectional converter The conclusion is presented in “Section 7”. Appendices follow the
Rin , Ro Dynamic resistance looking from battery side and conclusion.
output side
Rl Resistance of line between converter and DC bus
2. Description of the system studied
SS , SZ Power semiconductor switch during SEPIC and Zeta
operation respectively
The overall circuit diagram of BESS studied in this paper is shown
U, Vo , Vref General input, output and reference voltage
in Fig. 2(a), which facilitates both charging and discharging of the
Vbat , Vbus , Vos Battery terminal, DC bus and output side volt-
battery with the DC bus. The symbols of components and their
age respectively
short description are already given in the nomenclature section. In
VC1 , VC3 , VC2 Battery side, output side and flying capacitor
Fig. 2(a), BESS comprises of rechargeable battery and DC–DC bidi-
voltage respectively
rectional SEPIC-Zeta converter. In this scheme, SEPIC configuration
Vref1 , Vref2 Reference voltage for discharging and charging
is activated during discharging of the battery and Zeta configura-
mode respectively
tion comes into action during charging of the battery. The battery
1, 2, 3 Sliding coefficients
and operational modes of BESS studied are subsequently discussed
as follows:

2.1. Rechargeable battery

trol loop is designed for the said converter and the various dynamic
responses under charging and discharging operations are observed. The fundamentals, modeling and characteristics of rechargeable
These responses have been significantly improved by the usage battery are elaborately described in Ref. [20]. The state of charge
of proposed adaptive sliding mode control (SMC) strategy. Thus, (SOC) of battery is the important quantity to be monitored to decide
the obtained smooth charging and discharging profile can help to the initiation of charging and discharging activity of any battery.
increase life span of battery. In addition to this, the suggested strat- The 0% value of SOC indicates that the battery is of empty of charge
egy is flexible in changed operating conditions during charging and and it requires immediate charging. On the other hand, 100% value
discharging cycles. On the other hand, under changed operating of SOC points to the fully charged condition and it is ready to dis-
conditions, the retuning of PID controller gains is necessary but it charge with full capacity.
is also very difficult to obtain under unpredictable variable con-
ditions. The investigation on dynamic performances of SEPIC-Zeta 2.2. Operational modes
converter based BESS under charging and discharging cycles has
not been noticed in any publication to the best of author’s knowl- The BESS has two operating modes, namely charging operation
edge. Apart from this, usage of adaptive SMC for improving the and discharging operation. At first, charging operation is briefly
dynamic behavior of the said system is the new contribution of discussed as given below:
this article. The AC and DC micro-grid connected BESS is shown in The effective circuit diagram under charging mode of operation
Fig. 1. Here, various DGs are integrated into the common network is shown in Fig. 2(b), which is developed from Fig. 2(a) with the
via various converters. The BESS requires bidirectional DC–DC con- activation of the switch ‘SZ ’ and the anti-parallel diode ‘DZ ’. The
verter in both these schemes to support charging and discharging circuit configuration functions as Zeta converter during charging
operations. condition. This circuit configuration operates like buck converter
350 P.K. Gayen et al. / Electric Power Systems Research 160 (2018) 348–361

Fig. 2. (a) Overall bidirectional SEPIC-Zeta converter based BESS, (b) effective circuit diagram during charging condition, and (c) effective circuit diagram during discharging
condition.

as the standard DC bus voltage, which is greater than the voltage of of operation, the input-output relationship of SEPIC converter can
the battery, acts as input voltage during charging operation. In this be expressed as,
respect, the duty ratio (Dc ) of the PWM signal for switch ‘SZ ’ is con-
trolled to maintain the desired voltage at the terminal of the battery Dd
Vos = V (2)
to provide proper charging current of the battery. The input–output 1 − Dd bat
voltage relationship of Zeta converter can be expressed as,
Here, to maintain the boost operation of the SEPIC configuration,
the controlled value of Dd is expected to be greater than 0.5.
Dc
Vbat = Vos (1)
1 − Dc 3. State space averaged modeling of SEPIC and Zeta
converter
Here, to maintain the buck operation of the Zeta converter dur-
In this section, the state space averaged modeling of Zeta and
ing charging mode, the controlled value of Dc is expected to be less
SEPIC converter is presented. These discussions will be useful for
than 0.5.
designing the PID and adaptive SM controller. The state space aver-
Next, the discharging operation of the battery is discussed as
aged model of Zeta and SEPIC converter is formed from the dynamic
follows:
equations for on and off state of the respective converters.
In this mode of operation, the effective circuit diagram is shown
in Fig. 2(c), which is formed from the circuit shown in Fig. 2(a)
by activation of the switch ‘SS ’ and the anti-parallel diode ‘DS ’. 3.1. Dynamic equations
As a result, the circuit configuration behaves like SEPIC converter.
Here, the discharged current is fed to the DC bus via SEPIC config- The differential equations governing the dynamics of these con-
uration and the converter functions as boost converter because it verters are expressed as follows:
steps up the battery voltage (i.e. input voltage for SEPIC converter) In the case of Zeta converter, the dynamic equations during on
to the controlled output voltage (i.e. output side voltage of SEPIC state of switch are,
converter). The discharged current is regulated by controlling the
output voltage of the converter (Vos ), which is done by controlling diL2 Vos
= (3)
duty ratio (Dd ) of the PWM signal of switch ‘SS ’. In discharging mode dt L2
 P.K. Gayen et al. / Electric Power Systems Research 160 (2018) 348–361 351

diL1 Vos VC2 VC1 Combining Eqs. (21)–(24), the weighted state space averaged
= + − (4)
dt L1 L1 L1 equations are expressed as,
dVC2 iL1 Ẋ = [A1 d + A2 (1 − d)]x + [B1 d + B2 (1 − d)]U (25)
=− (5)
dt C2
Vo = [C1 d + C2 (1 − d)]x + [D2 d + D2 (1 − d)]U (26)
dVC1 iL1 VC1
= − (6)
dt C1 Rin C1 From Eqs. (3)–(6), the state matrix of Zeta converter for on state
Vbat is written as,
In Eq. (6), Rin = Ibat ⎡ ⎤
The governing differential equations for Zeta configuration dur- ⎡ ⎤ 0 0 0 0 ⎡ ⎤ ⎡ 1 ⎤
ing off state are given by, Ẋ1 ⎢ 1 1 ⎥ x1
⎢ ⎥ ⎢ 0 0 − ⎥ ⎢ ⎥ ⎢ L2 ⎥
diL2 VC2 ⎢ Ẋ2 ⎥ ⎢
⎢
L1 L1 ⎥ ⎢ x2 ⎥ ⎢ 1
⎥⎢ ⎥ ⎢
⎥
⎥
=− (7) ⎢ ⎥=⎢ ⎥ ⎢ ⎥ + ⎢ L1 ⎥ [U] (27)
dt L2 ⎢ Ẋ ⎥ ⎢ 0 − d 0 0 ⎥ ⎣ x3 ⎦ ⎢ ⎥
⎣ 3⎦ ⎢ C2 ⎥ ⎣0 ⎦
diL1 VC1 ⎣ ⎦
=− (8) Ẋ4 1 1 x4
dt L1 0 0 − 0
C1 Rin C1
dVC2 iL2
= (9) From Eqs. (7)–(10), the state matrix for off state becomes,
dt C2
⎡ ⎤
dVC1 iL1 VC1 1
⎡ ⎤ 0 0 − 0
= − (10) ⎢ L2 ⎥⎡x ⎤
dt C1 Rin C1 Ẋ1 ⎢ ⎥ 1
⎢ ⎥ ⎢ 1 ⎥⎢ ⎥
Next, the dynamic modeling of SEPIC converter during on state ⎢ Ẋ2 ⎥ ⎢ 0 0 0 − ⎥ ⎢ x2 ⎥
of converter switch is presented as given below, ⎢ ⎥=⎢ ⎢
L1 ⎥⎢ ⎥
⎥⎢ ⎥ (28)
⎢ Ẋ ⎥ ⎢ 1 ⎥ ⎣ x3 ⎦
diL1 V ⎣ 3⎦ ⎢ 0 0 0 ⎥
= bat (11) ⎢ C2 ⎥
dt L1 Ẋ4 ⎣ 1 1
⎦ x4
0 0 −
diL2 VC2 C1 Rin C1
= (12)
dt L2
Combining Eqs. (27) and (28) as per Eqs. (25)–(26), the state
dVC2 iL2 space averaged model of Zeta converter can be obtained as,
=− (13)
dt C2 ⎡ 1−d
⎤
⎡ ⎤ 0 0 − 0 ⎡ d ⎤
dVC3
=−
VC3
(14) Ẋ1 ⎢ L2
⎥ ⎡ x1 ⎤
⎢ ⎥ L2
dt Ro C3
⎢ Ẋ2 ⎥ ⎢ 0 d 1
⎥ ⎢ x2 ⎥ ⎢ ⎥
⎥⎢ ⎥ + ⎢ ⎥
0 −
⎢ ⎥=⎢ L1 L1
⎢ Ld ⎥ [U]
Here, Ro = Vos
⎢ ⎥ ⎢ ⎥ (29)
Ibus
⎣ Ẋ3 ⎦ ⎢ 1 − d 1 − d ⎥ ⎣ x3 ⎦ ⎢ 1 ⎥
For off state of switch, the dynamic modeling of SEPIC converter ⎢ C2 −
d
0 −
Rin C2 ⎥
⎣ ⎦
is presented as, Ẋ4 ⎣ C2
⎦ x4 0
1 1 0
0 0 −
diL1 VC2 VC3 C1 Rin C1

dt
=−
L1
−
L1
(15) ⎡ ⎤
x1
diL2 VC3 ⎢ ⎥
=− (16)  ⎢ x2 ⎥
dt L2 Vo = 0 0 0 1 ⎢⎢x ⎥
⎥ (30)
dVC2 iL1 ⎣ 3⎦
= (17)
dt C2 x4
dVC3 iL1 iL2 VC3 The state variables for Zeta converter are taken as,
= − − (18)
dt C3 C3 Ro C3
x1 = iL2 , x2 = iL1 , x3 = VC2 , x4 = VC1 ,
3.2. State space averaged model
And d = duty ratio, U = Vos , Vo = Vbat
The state space averaged model for Zeta and SEPIC converter is Similarly, the state space averaged model of SEPIC converter can
developed based on the dynamic equations given in Sub-section be obtained as,
3.1. ⎡ 1−d 1−d
⎤
⎡ ⎤ 0 0 − −
The general state space equations is written as,
Ẋ1 ⎢ L1 L1
⎥ ⎡ x1 ⎤ ⎡ d ⎤
⎢ 1−d ⎥
Ẋ = Ax + BU (19) ⎢ Ẋ2 ⎥ ⎢ 0 0
d
− ⎥ ⎢ x2 ⎥ ⎢ L1 ⎥
⎢ ⎥=⎢ L2 L2 ⎥⎢ ⎥ + ⎢ 0 ⎥ [U]
⎢ ⎥ ⎢ ⎥⎣ ⎦ ⎣ ⎦ (31)
Vo = Cx + DU (20) ⎣ Ẋ3 ⎦ ⎢ 1 − d d ⎥ x3
⎢ C2 − 0 0 ⎥ 0
The model for on state condition is represented as, Ẋ4 ⎣ C2
⎦ x4 0
1−d 1−d 1
− 0 −
Ẋ = A1 x + B1 U (21) C3 C3 C3 Ro
⎡ ⎤
Vo = C1 x + D1 U (22) x1
⎢ ⎥
For off State, the model is represented as,  ⎢ x2 ⎥
Vo = 0 0 0 1 ⎢ ⎥ (32)
⎢x ⎥
Ẋ = A2 x + B2 U (23) ⎣ 3⎦
Vo = C2 x + D2 U (24) x4
352 P.K. Gayen et al. / Electric Power Systems Research 160 (2018) 348–361

Here, it is assumed that u(t) is the input of the system, b is

constant. x, ẋ and ẍ are state of the system, derivative and double
derivative of the state respectively with time.
‘S’ is switching function (i.e. mathematical equation of the slid-
ing surface) defined as,

S = 1 e + 2 ė (34)

Here, error is e = vref − vo , vo = output state and vref is the desired

state in general, 1, 2 are sliding coefficients. sgn(S) is signum func-
tion of ‘S’. The control rule is defined as,

1 1 if S > 0
d = (1 + sgn(S)) = (35)
2 0 if S < 0

The control action (i.e. ‘d’ in Eq. (35)) operates depending on

the system’s position on two sides of the sliding manifold, S = 0.
Fig. 3. PID based control scheme.
In Fig. 4(a), system trajectory moves towards the origin point
(i.e.e = ė = 0, also called equilibrium point) rapidly, after sliding
mode occurs (i.e. S = 0). The existence condition of sliding mode
The state variables for SEPIC converter are taken as, controller is generally obtained by,

x1 = iL1 , x2 = iL2 , x3 = VC2 , x4 = VC3 LimS.Ṡ < 0 (36)

S→0

And the input variable is U = Vbat . Here, Vo = Vos . In the classical model of SMC discussed above, the operating
switching frequency is not fixed and very high [22] according to
the change in input and the output requirements of any system.
4. PID based control scheme
As a result, the current and voltage stress can be developed on the
battery, if classical model of SMC is taken as the controller for the
The block diagram, operation and tuning of PID based control
BESS. This condition is highly unfavorable to the life of a battery as
scheme is presented in this section.
a power source. To make better profile of charging and discharging
operations, adaptive SMC model is taken into the consideration in
4.1. PID control scheme the study of this paper.

The block diagram representation of the PID based control 5.2. Proposed adaptive sliding mode control logic for BESS
scheme is shown in Fig. 3. It also shows the connection diagram
of PID controller based system during charging and discharging In the proposed model of adaptive SMC, the main focus is to sup-
operation. In Fig. 3, position ‘1’ and ‘2’ of each selector switch port smooth charging and discharging of the battery under variable
is activated for discharging and charging operation respectively. operating conditions. In the DC bus connected BESS application,
Accordingly, the reference, feedback and control signals for PID the suggested controller operates on the basis of proposed modi-
based scheme are altered depending on the type of operation i.e. fied switching function, which is an advancement of the classical
charging and discharging action. During charging and discharg- model of SMC. The modified switching function is expressed in the
ing mode of operations, Zeta and SEPIC configured converters are form given below,
controlled respectively by the tuned PID controller.
S = 1 P1 + 2 P2 + 3 P3 (37)
4.2. Tuning of PID where, 1 , 2 and 3 are sliding coefficients. The states P1 , P2 , P3
are defined as, 
The tuning procedure of PID controller is discussed here on the P1 = (Vref − Vo ), P2 = (Vref − Vo )dt and P3 = dtd
(Vref − Vo ) =
basis of state space averaged model given in Section 3. At first, the d −i
− dt VC = CC , as Vo = VC = output side capacitor voltage for converter
MATLAB program is written to obtain the transfer function from the
in general.
state space averaged model. Then, Routh array is formed from the
And 1 , 2 , 3 are chosen in such way, that the switching func-
obtained transfer function. The critical gain and frequency is calcu-
tion becomes,
lated from the Routh table. These two parameters are used to find 
the gains of the PID controller using conventional Ziegler–Nichols k1 Io
S= (Vref − Vo ) + k2 (Vref − Vo )dt − k3 iC (38)
tuning chart. The designed gains are provided in Appendix A. Vo

where, Vref = reference voltage, Vo = output volt, U = input voltage,

5. Sliding mode control schemes iC = capacitor current, VC = output side capacitor voltage, Io = output
current and k1 , k2 and k3 are chosen coefficients used in the adaptive
Both classical SMC and proposed adaptive SMC are described in SM controller.
this section as follows: k I
In Eq. (38), the sliding coefficient ‘1 ’ is taken as V1oo to facilitate
the adaptive controller to counteract the disturbances, which are
5.1. Classical sliding mode controller introduced during changed loading conditions. The ‘2 ’ and ‘3 ’ are
chosen as coefficients k2 and −k3 respectively in Eq. (38), which
Let us consider the design of a classical sliding mode controller influence the dynamic response of the system. In this context, the
[21] for the following second order system: necessary conditions for deciding the values of k1 , k2 and k3 are
presented below. Here, the necessary conditions are formed from
ẍ = f (x, ẋ, t) − bu(t) (33) the existence and stability criteria of SMC.
 P.K. Gayen et al. / Electric Power Systems Research 160 (2018) 348–361 353

Fig. 4. (a) Sliding surface for classical SMC, (b) sliding surface for adaptive SMC and (c) proposed adaptive SMC scheme.

The controlled duty ratio of the converter is expressed in terms is enabled to control SEPIC and Zeta configured converter for dis-
of the switching function (S) as, charging and charging conditions respectively. During discharging
1 mode of operation, SEPIC converter comes into action. Hence, the
d= {1 + sgn(S − |k|)} (39) switching function for controlling of SEPIC converter is formed from
2
Eq. (38) as,
where, 2k is the band of oscillation of switching and sign of k is
negative, when sign of S is also negative. Fig. 4(b) shows that once 
the system dynamics reaches the sliding manifold (i.e. S = 0), the k1
SS = (V − Vos ) + k2 (Vref 1 − Vos )dt − k3 iC3 (40)
system oscillates with a narrow band ‘2k’around the line, S = 0 and R0 ref 1
reaches rapidly to the stable point where all the states P1 , P2 , P3 are
equal to 0 (i.e. equilibrium point).
where, Vref1 = reference output voltage of SEPIC, Vos = output volt-
The block diagram of adaptive SM controller for the bidirectional
age of SEPIC, and IC3 = output side capacitor current used in the SM
SEPIC-Zeta converter in connection with charging and discharging
controller for SEPIC converter. Taking derivative of Eq. (40), the
of battery is shown in Fig. 4(c). Here, the control signal is generated
transformed equation is given by,
according to the control action given in Eq. (39). During discharg-
ing mode, all the selector switches S1 − S6 are placed in terminal
1 simultaneously and in case of charging mode, all these switches k1 d d
ṠS = (V − Vos ) + k2 (Vref 1 − Vos ) − k3 iC3 (41)
are placed in terminal 2. In this way, the suggested control circuit Ro dt ref 1 dt
354 P.K. Gayen et al. / Electric Power Systems Research 160 (2018) 348–361

Fig. 5. (a) Output current and (b) output voltage responses during normal operation.

By substituting Vos = VC3 and iC3 = ((1 − d)iL1 − (1 − d)iL2 − ios ), Eq. 1 1 1 1
(41) modifies to + VC3 − V − VC2 − VC3 )] (45)
L1 L1 bat L2 L2
k1 d
ṠS = − VC3 + k2 (Vref 1 − VC3 ) The Eq. (45) can be represented in compact form as,
Ro dt
d ṠS = f (g) + df (h) + d2 f (p) (46)
−k3 ((1 − d)iL1 − (1 − d)iL2 − ios ) (42)
dt
where,

Rearranging Eq. (42) and putting ios = Vos

=
VC3
we get, (k1 − k3 ) 1 1 1 1 1
Ro Ro (− iL1 + iL2 + VC3 ) − k3 ( VC2 − VC3 )
Ro C3 C3 C3 Ro L1 L1
k1 k3 d 1
ṠS = (− + ) VC3 + k2 (Vref 1 − VC3 ) +k3 ( VC3 ) + k2 (Vref 1 − VC3 ) = f (g) ,
Ro Ro dt L2
d d
−k3 (1 − d) iL1 + k3 (1 − d) iL2 (43) (k1 − k3 ) 1 1 2 2 2
dt dt ( iL1 − iL2 ) + k3 ( VC2 + VC2 − VC3
Ro C3 C3 L2 L1 L1

d d d 1
By putting the values of dt VC3 , dt iL1 and dt iL2 from state space + V ) = f (h)
averaged matrix of SEPIC converter given in Eq. (31), Eq. (43) L1 bat
becomes 1 1 1 1 1
and k3 (− VC2 + VC3 − V − VC2 − VC3 ) = f (p)
k1 k3 1 − d 1−d 1 L1 L1 L1 bat L2 L2
ṠS (− + )( iL1 − iL2 − VC3 )
Ro Ro C3 C3 C3 Ro So from the local reachability condition, LimS.Ṡ < 0, the exis-
S→0
1−d 1−d d tence of sliding mode during discharging condition is derived as,
−k3 (1 − d)( VC2 − VC3 − V )
L1 L1 L1 bat ⎧ k k3 VC3
d 1−d ⎪ 1
⎪ (− Ro + Ro )(− C Ro ) + k2 (Vref 1 − VC3 ) < 0
+k3 (1 − d)( VC2 − VC3 ) + k2 (Vref 1 − VC3 ) (44) ⎪
⎪
L2 L2 ⎪
⎨ (k − k ) i
3

1 3 i L1 V L2 C3
(− + + ) (47)
⎪
⎪
Ro C3 C3 C3 Ro
Again, rearranging Eq. (44), it is expressed as, ⎪
⎪
⎪
⎩ −k ( VC2 − VC3 + VC3 ) + k (V − V ) > 0
3 2 ref 1 C3
(k1 − k3 ) 1 1 1 1 1 L1 L1 L2
ṠS = (− iL1 + iL2 + VC3 ) − k3 ( VC2 − VC3 )
Ro C3 C3 C3 Ro L1 L1
The equivalent control signal by the SMC is derived for charging
1 (k1 − k3 ) 1 1 condition by making ṠS = 0 in Eq. (46) as,
+k3 ( VC3 ) + k2 (Vref 1 − VC3 ) + d[ (+ iL1 − iL2 )
L2 Ro C3 C3 
−f (h) ± f 2 (h) − 4f (p)f (g)
2 2 2 1 1 deqS = (48)
+k3 ( VC2 + VC2 − VC3 + V )] + d2 [k3 (− VC2 2f (p)
L2 L1 L1 L1 bat L1
 P.K. Gayen et al. / Electric Power Systems Research 160 (2018) 348–361 355

Fig. 6. (a) Terminal voltage, (b) terminal current and (c) SOC of battery at normal condition.

For finding the stability condition of SEPIC converter, by putting The characteristics equation is formed by doing |sI − A| = 0 as,
the equivalent control signal ‘deqS ’ in the ‘A’ matrix of the state space (s is the Laplace operator)
averaged model in Eq. (31) to as,
⎡ ⎤ s4 + s3 z + s2 (−tw − uy − qv − xr) + s(−twz − qvz)
1 − deqS 1 − deqS
0 0 − −
⎢ L1 L1 ⎥ + quvy − quxw + rtvy − rtvw = 0 (51)
⎢ ⎥
⎢ deqS 1 − deqS ⎥
⎢0 0 − ⎥
⎢ L2 L2 ⎥ The condition for the stability of the system is
A=⎢ ⎥ (49)
⎢ 1 − deqS deqS ⎥ ⎧
⎢ − 0 0 ⎥
⎢ C2 C2 ⎥ ⎪
⎪
z>0
⎣ ⎦ ⎪ xr > uy
⎨
1 − deqS 1 − deqS 1
− 0 − quvy + rtvy > quvw + rtxw (52)
C3 C3 C3 Ro ⎪
⎪
⎪
⎩ −twz + qvz > quvyz − quxwz + rtvyz − rtxwz
For simplification, by introducing new symbols for the corre-
−uy + xr
sponding elements of the matrix ‘A’, Eq. (49) is transformed into
following form, During charging mode of operation, Zeta converter comes into
⎡ ⎤ action. Hence, the switching surface for controlling the Zeta con-
0 0 −q −r
⎢0 0 verter is formed from Eq. (38) as,
−t −u ⎥
A=⎢ ⎥ (50) 
⎣ −v −w 0 0 ⎦ k1
SZ = (V − Vbat ) + k2 (Vref 2 − Vbat )dt − k3 iC1 (53)
−x −y 0 −z Rin ref 2
356 P.K. Gayen et al. / Electric Power Systems Research 160 (2018) 348–361

Fig. 7. (a) Output current and (b) output voltage responses at changed operating condition.

where, Vref2 = reference output voltage of Zeta, Vbat = battery termi- So from the local reachability condition, LimS.Ṡ < 0, the exis-
nal voltage of Zeta, and IC1 = battery side capacitor current used in S→0
tence of sliding mode during charging condition is derived as,
the adaptive SM controller for Zeta converter. Taking derivative of
Eq. (53), it can be expressed as,
⎧ (k − k ) i VC1 VC1
⎪
⎪
1 3 L1
⎪ [ Rin (− C1 + C1 Rin ) + k3 L1 + k2 (Vref 2 − VC1 )
k1 d d
ṠZ = (V − Vbat ) + k2 (Vref 2 − Vbat ) − k3 iC1 (54) ⎪
Rin dt ref 2 dt ⎪
⎪
⎪
⎪ VC2 Vos
⎪
⎨ −k3 ( L1 + L1 )] < 0
By substituting Vbat = VC1 and iC1 = (iL1 − ibat ), Eq. (54) becomes
(59)
⎪
⎪ (k1 − k3 ) 1
(− iL1 +
1
ṠZ = −
k1 d d
VC1 + k2 (Vref 2 − VC1 ) − k3 (iL1 − ibat ) ⎪
⎪
VC1 )
(55) ⎪
⎪
Rin C1 C1 Rin
Rin dt dt ⎪
⎪
⎪
⎩ +k 1 V + k (V − V ) > 0
Vbat VC1 3 C1 2 ref 2 C1
Rearranging Eq. (55), and taking ibat = Rin
= Rin
we get, L1

k1 d d d VC1
ṠZ = − VC1 + k2 (Vref 2 − VC1 ) − k3 iL1 + k3 (56) The equivalent control signal by the SMC is derived for charging
Rin dt dt dt Rin
condition by making ṠZ = 0 in Eq. (55) as,
d d
By putting the values of dt VC1 and dt iL1 from state averaged
matrix of Zeta converter [i.e. Eq. (29)], the equation is derived as,
f (i)
deqZ = (60)
k1 k3 1 1 f (j)
ṠZ = (− + )( iL1 − VC1 ) + k2 (Vref 2 − VC1 )
Rin Rin C1 C1 Rin
d 1 d For finding the stability condition of Zeta converter, by putting
−k3 ( VC2 − VC1 + Vos ) (57)
L1 L1 L1 the equivalent control signal ‘deqZ ’ in the ‘A’ matrix of the state space
averaged model in Eq. (29) to as,
By rearranging Eq. (57), it can be represented as,
⎡ 1 − deqZ
⎤
(k1 − k3 ) 1 1 1 0 0 − 0
ṠZ = (− iL1 + VC1 ) + k3 VC1 + k2 (Vref 2 − VC1 ) ⎢ L2 ⎥
Rin C1 C1 Rin L1 ⎢ ⎥
⎢ deqZ 1 ⎥
d d ⎢0 0 − ⎥
−d[k3 ( VC2 + Vos )] = f (i) − df (j) (58) ⎢ L1 L1 ⎥
L1 L1 A=⎢ ⎥ (61)
⎢ 1 − deqZ deqZ 1 − deqZ ⎥
⎢ − 0 − ⎥
where,
(k1 −k3 )
(− C1 iL1 + 1
V ) + k3 L1 VC1 + k2 (Vref 2 − VC1 ) = ⎢ C2 C2 Rin C2 ⎥
Rin 1 C1 Rin C1 1 ⎣ ⎦
f (i) 1 1
0 0 −
And k3 ( Ld VC2 + d
V )
L1 os
= f (j) C1 Rin C1
1
 P.K. Gayen et al. / Electric Power Systems Research 160 (2018) 348–361 357

Fig. 8. (a) Terminal voltage, (b) terminal current and (c) SOC of battery under changed operating condition.

For simplification, by introducing new symbols for the corre- The values of k1 , k2 and k3 are decided based on the relations
sponding elements of the matrix ‘A’, Eq. (61) is transformed into of the inequalities Eqs. (47), (52), (59) and (64). In this way, the
following form, existence condition and the stability of BESS is guaranteed after
⎡ ⎤ hitting the system trajectory in the sliding manifold, S = 0 with these
0 0 −a 0 designed coefficients in the proposed adaptive SMC technique.
⎢ 0 0 −b −c ⎥
A=⎢ ⎥ From the above discussion, it can be summarized that the
(62)
⎣ −d −e 0 −f ⎦ notable features of the proposed control scheme are
0 −g 0 −h
1. Constant switching frequency operation is established.
The characteristics equation is formed by doing |sI − A| = 0 as,
2. Adaptability to the variation of input (i.e. charging current) and
s4 + s3 h + s2 (−be − cg − ad) + s(−adh − beh + bgf ) + acdg = 0 (63) output current (i.e. discharging current) is assured.
3. The designed tuned values of k1 , k2 and k3 maintains stability of
The condition for the stability of the system is the BESS under changed operating condition.
⎧
⎪ h>0
⎪
⎪
⎪
⎪ acdg >0 6. Simulation and experimental study
⎨
bgf
cg < − (64) 6.1. Simulation study
⎪
⎪ h
⎪
⎪
⎪
⎩ adh + beh − bgf > acdgh
2
The overall simulation study is carried out using MATLAB-
cyh + bgf SIMULINK software. The specifications of various components and
358 P.K. Gayen et al. / Electric Power Systems Research 160 (2018) 348–361

Fig. 9. (a) Experimental set-up and (b) fabricated bidirectional SEPIC-Zeta converter.

devices used in the simulation model are provided in Appendix A. (b) respectively. In Fig. 6(a) and (b), better transient and steady
The different dynamic responses are compared between PID and state performances are maintained by using the designed adaptive
adaptive SMC controller based scheme during dynamic charging SMC methodology. The SOC under both case studies are plotted
and discharging conditions. The comparative results are presented in Fig. 6(c) to show the starting and ending of the charging and
as follows: discharging action. Thus, the adaptive SMC case study shows supe-
rior dynamic performance over the PID case study under combined
6.1.1. Results under normal charging and discharging mode of charging and discharging conditions. The smoother charging and
operation discharging profile can increase the life span of the battery by
In this section, combined charging and discharging operations adopting adaptive SMC technique.
of BESS are performed to validate the improved dynamic control
capability of adaptive SMC technique in comparison to the PID con- 6.1.2. Results under changed operating condition
troller. The threshold values of SOC for deciding the initiation of In this simulation study, the reference output voltage is decided
charging and discharging are taken as 40% and 100% respectively. as 560 V, whereas output voltage is taken as 540 V in the study
Here, the reference output voltage during discharging condition presented in Section 6.1.1. Thus, battery discharges faster with
is taken as 540 V. The DC-bus voltage is taken as 500 V. The resis- discharge current of 3.0 A in comparison to the study of Section
tance of line between converter and DC bus is decided as 20 Ohm. In 6.1.1. (i.e. +50% higher). Under this changed operating condition,
this condition, the BESS will deliver 2.0 A current into DC bus. This the various responses under discharging and charging cycles are
discharge action will stop when SOC reaches at the value of 40%. observed in both PID and adaptive SMC case studies. In this operat-
Then, charging action will be initiated and the charging current of ing situation, the output grid side current and voltage profiles are
the battery is to be taken from DC bus. This charging current can presented in Fig. 7(a) and (b) respectively. These figures reveal that
be controlled by suitably chosen external battery terminal voltage. PID controller based system used in the study of Section 6.1.1 fails to
For this purpose, the reference external battery terminal voltage is maintain charging and discharging operations. On the other hand,
decided as 130 V for the charging action. After completion of charg- proposed adaptive SMC based system used in the study of Sec-
ing (i.e. SOC reaches 100%), the discharge action starts again. In this tion 6.1.1 is capable to perform the charging and discharging cycles
way, discharging and charging cycle continues. This cycle opera- of DC bus connected BESS satisfactorily under changed operating
tion is implemented using both PID and adaptive SM controller condition.
based schemes. The various dynamic responses are observed and The corresponding battery terminal voltage, current and SOC of
presented here. Fig. 5(a) shows that the DC bus side dynamic cur- both PI and adaptive SMC case studies are shown in Fig. 8(a)–(c)
rent under charging and discharging cycles have been significantly respectively to indicate the incapability of PID controller to main-
improved for the proposed adaptive SMC. Fig. 5(a) shows lesser tain the operation of BESS under changed situation.
transient peak and oscillation of the current response in the case of From the various simulation studies presented in this section,
adaptive SMC in comparison to the PID case. The respective output it can be stated that significant improvement of dynamic perfor-
voltage is shown in Fig. 5(b). Fig. 5(b) reveals that the dynamic volt- mance of the DC bus connected BESS has been achieved with the
age profile under adaptive SMC case study is very much smoother proposed adaptive SMC technique in comparison to the PID case.
than the PID case study. Here, the charging and discharging profile is more acceptable for
The corresponding battery terminal voltage and current under the battery with the usage of adaptive SMC technique. It leads to
charging and discharging cycles is represented in Fig. 6(a) and increase of life span of battery as an electrical power source. Simul-
 P.K. Gayen et al. / Electric Power Systems Research 160 (2018) 348–361 359

Fig. 10. Experimental results of (a) output current and (b) voltage responses using adaptive SMC.

taneously, the extension of operating range of the bidirectional 7. Conclusion

DC–DC converter based BESS is possible by using adaptive SMC
technique. This paper aims at improvement of dynamic performances of
DC–DC bidirectional SEPIC-Zeta based BESS under charging and dis-
charging operations. At the beginning, the conventional PID based
6.2. Experimental study control scheme is designed for this converter. Oscillations with
transient peaks are observed in the voltage and current responses
The experimental work on the prototype of the BESS is done in in BESS, while this control scheme is used. Also, the PID controller
the laboratory to validate proposed adaptive SMC based system. based scheme fails to maintain charging and discharging operations
The overall experimental set up is shown in Fig. 9. The PC inter- of the battery during changed operating conditions. Then, adaptive
faced dSPACE 1104 is used as real time controller board to generate SMC action is implemented for controlling the converter of BESS. In
the controlled switching signals using adaptive SMC for the power this context, the existence and stability conditions are also derived
metal oxide semiconductor field effect transistors (2SK3878) of the to design the adaptive SMC. After applying this proposed tech-
bidirectional SEPIC-Zeta converter. The voltage and current sen- nique, the overall dynamic performances have been significantly
sors used in the experiment for measuring voltage and current are improved and operating zone under changed conditions has also
hall-effect sensors such as LV25P and LTS25NP respectively. The been extended. In this way, this paper justifies that the superior-
lead-acid battery cell used in the experiment has the rating of 12 V, ity of the employed control scheme for bidirectional SEPIC-Zeta
1.6 Ah. The dc bus voltage is taken as 30 V, which is regulated dc converter based BESS under various operating conditions. The var-
power supply. Output voltage and current under dynamic charging ious schemes are simulated using MATLAB-SIMULINK software.
and discharging operations are shown in Fig. 10. Fig. 11 shows the The applicability of the proposed adaptive SMC scheme is verified
battery side current and voltage responses. through scaled down model of the proposed system in hardware
360 P.K. Gayen et al. / Electric Power Systems Research 160 (2018) 348–361

Fig. 11. Experimental results of (a) battery side current and (b) voltage responses using adaptive SMC.

platform. Thus, the novelty of this paper is to improve dynamic References

performance of SEPIC-Zeta based BESS during variable operating
conditions by using adaptive SMC. [1] T. Ackermann, G. Andersson, L. Soder, Distributed generation: a definition,
Electr. Power Syst. Res. 57 (2000) 195–204.
[2] T. Adefarati, R.C. Bansal, Integration of renewable distributed generators into
the distribution system: a review, IET Renew. Power Gener. 10 (7) (2016)
Appendix A. Parameters of system studied
873–884.
[3] N. Hatziargyriou, Microgrids Architectures and Control, first ed., John Wiley &
Components Specifications Sons Ltd, The Atrium, Southern Gate, Chichester, West Sussex, PO 19 8SQ,
United Kingdom, 2014.
Rechargeable battery Lead acid [4] Q. Sun, J. Zhou, J.M. Guerrero, H. Zhang, Hybrid three-phase/single-phase
120 V nominal voltage, 10 Ah microgrid architecture with power management capabilities, IEEE Trans.
C1 , C2, C3 88 ␮F, 56 ␮F, 88 ␮F Power Electron. 30 (10) (2015) 5964–5977.
L1 , L2 4.0 mH, 8.0 mH [5] N. Eghtedarpour, E. Farjah, Distributed charge/discharge control of energy
storages in a renewable-energy-based DC micro-grid, IET Renew. Power
Gener. 8 (1) (2014) 45–57.
Appendix B. Gains of PID controller [6] K. Divya, J. Østergaard, Battery energy storage technology for power systems
— an overview, Electr. Power Syst. Res. 79 (4) (2009) 511–520.
[7] X. Li, D. Hui, X. Lai, Battery energy storage station (BESS)-based smoothing
KP KI KD control of photovoltaic (PV) and wind power generation fluctuations, IEEE
Trans. Sustain. Energy 4 (2) (2013) 464–473.
0.001 0.1 0.025 [8] S.W. Mohod, M.V. Aware, Micro wind power generator with battery energy
storage for critical load, IEEE Syst. J. 6 (1) (2012) 118–125.
[9] Y. Jiao, W. Qiao, A hierarchical power management strategy for multiple
Appendix C. Coefficients of adaptive SMC single-phase roadway microgrids, IEEE PES General Meeting (2013) 1–5.
[10] D.S. Koussa, M. Koussaa, A. Rennanea, S. Hadji, A. Boufertella, A. Balehouane, S.
Bellarbi, Hybrid diesel-wind system with battery storage operating in
k1 k2 k3 standalone mode: control and energy management — experimental
investigation, Energy 130 (2017) 38–47.
2000 50 100
 P.K. Gayen et al. / Electric Power Systems Research 160 (2018) 348–361 361

[11] K. Uddin, A.D. Moore, A. Barai, J. Marco, The effects of high frequency current [16] M.H. Rashid, Power Electronics: Circuits, Devices and Applications, Pearson
ripple on electric vehicle battery performance, Appl. Energy 178 (2016) Education, 2004.
142–154. [17] J. Falin, Designing DC/DC converters based on SEPIC topology, power
[12] M. Dubarry, A. Devie, K. Stein, M. Tun, M. Matsuura, R. Rocheleau, Battery management, Analog Appl. J. (2008) 18–23, Texas Instruments.
energy storage system battery durability and reliability under electric utility [18] J. Falin, Designing DC/DC converters based on ZETA topology, power
grid operations: analysis of 3 years of real usage, J. Power Sources 338 (2017) management, Analog Appl. J. (2010) 16–21, Texas Instruments.
65–73. [19] C.D. Denny, M. Shahin, Analysis of bidirectional SEPIC/Zeta converter with
[13] M.K. Kazimierczuk, Pulse-Width Modulated DC–DC Power Converters, 2nd coupled inductor, IEEE Int. Conf. on Technological Advancements in Power &
Ed., John Wiley & Sons Ltd, The Atrium, Southern Gate, Chichester, West Energy (2015).
Sussex, PO 19 8SQ, United Kingdom, 2016. [20] O. Tremblay, L.A. Dessaint, A.I. Dekkiche, A generic battery model for the
[14] L.S. Yang, T.J. Liang, Analysis and implementation of a novel bidirectional dynamic simulation of hybrid electric vehicles, Proc. IEEE Veh. Power
DC–DC converter, IEEE Trans. Ind. Appl. 59 (l) (2012) 422–434. Propulsion Conf. (VPPC) (2007) 284–289.
[15] I. Aharon, A. Kuperman, D. Shmilovitz, Analysis of bidirectional buck-boost [21] K.D. Young, V.I. Utkin, U. Ozguner, A control engineer’s guide to sliding mode
converter for energy storage applications, in: Proc. of IEEE IECON Conf., control, IEEE Trans. Control Syst. Technol. 7 (3) (1999) 328–342.
Vienna, 2013, pp. 858–863. [22] A. Sabanovic, Variable structure systems with sliding modes in motion control
— a survey, IEEE Trans. Ind. Inform. 7 (2) (2011) 212–223.