ISA Transactions 52 (2013) 638643

Contents lists available at ScienceDirect

ISA Transactions
journal homepage: www.elsevier.com/locate/isatrans

Analysis and design of sliding mode controller gains for boost power
factor corrector
Abdelhalim Kessal a,n, Lazhar Rahmani b
a
b

Mohammed el bachir el ibrahimi University, Bordj Bou Arrridj, Algeria

Automatic Laboratory of Stif (LAS), Ferhat Abbas University, Algeria

art ic l e i nf o

a b s t r a c t

Article history:
Received 26 March 2013
Accepted 3 May 2013
Available online 2 June 2013
This paper was recommended for publication by Jeff Pieper

This paper presents a systematic procedure to compute the gains of sliding mode controller based on an
optimization scheme. This controller is oriented to drive an ACDC converter operating in continuous
mode with power factor near unity, and in order to improve static and dynamic performances with large
variations of reference voltage and load. This study shows the great inuence of the controller gains on
the global performances of the system. Hence, a methodology for choosing the gains is detailed. The
sliding surface used in this study contains two state variables, input current and output voltage;
the advantage of this surface is getting reactions against various disturbancesat the power source, the
reference of the output, or the value of the load. The controller is experimentally conrmed for steadystate performance and transient response.
& 2013 ISA. Published by Elsevier Ltd. All rights reserved.

Keywords:
PFC
Power factor
Sliding mode control
Switching frequency

1. Introduction
Usually, traditional PID controllers are used for the control of
power converters [13]. Simple models of converters are generally
obtained from signals averaging and linearization techniques;
these models may then be used for control design [4,5]. On the
other hand, PID controllers failed to satisfactorily perform constrained specications under large parameter variations and load
disturbances [2]. Another choice for controlling power converter is
to use the sliding control techniques. Sliding mode control (SMC)
of variable structure systems such as power converters is particularly interesting because of its natural robustness, its capability of
system order reduction, and suitability for the nonlinearity aspect
of power converters [57]. However, despite being a popular
research subject, SMC is still rarely applied in practical ACDC
converters. It is mainly due to the fact that no systematic
procedure is available for the design of SMC in practical applications [8]. For example, the inuence of the controller gains on the
closed loop system performances for a given application is not
properly claried, and most of the previous works are limited to
the study of the inuence of these parameters only on the
existence and stability of sliding mode [9,10]. In other cases an
empirical approach is adopted for selecting these gains of SMC;
computer simulation and experiments were performed to study
the effect of the various control gains on the response of the
output voltage [10]. Therefore in this paper, analysis and design of

Corresponding author. Tel.: +213 66404 8090; fax: +213 35 674543.

E-mail address: Abdelhalim.kessal@yahoo.fr (A. Kessal).

SMC for power factor corrector (PFC) are studied. After studying
and analyzing different existing solutions for sliding mode control
of PFC, a control mode that allows a direct control of the voltage of
boost converter is proposed. The performances of the controller in
terms of robustness and dynamic response will be improved. Most
literature works are concerned with the study of hitting, existence
and stability conditions of the SMC. The contribution of this paper
goes beyond this direction by involving the study of the inuence
of control parameters on system performances. In this context, an
optimization algorithm is developed in order to choose the
controller parameters based on a predened specication for a
given real application.
Accordingly, this paper is oriented in the application of the
sliding modes for control of the bench of the power factor
corrector (PFC). Principle of control by sliding modes is described
briey. Thereafter, the application of this principle for the control
of the bench of PFC will be evoked. Based on the choice of the
sliding surface, various modes of control will be studied. Then a
mode of control based on a sliding surface utilizing all the
variables of state are studied; this is in order to improve the
performances of the closed loop. The important concepts associated to this type of control such as the convergence conditions,
existence, or stability of the sliding mode, are considered carefully.
This paper proposes a systematic analysis, design and digital
implementation of the proposed controller, composed by linear
controller in the DC voltage loop and sliding mode controller in
the current loop. This controller is veried by detailed MATLAB/
Simulink based on simulations through the use of a continuous
time plant model and a discrete time controller. Design is
comprehensive in the sense that it accounts for sampling effects,

0019-0578/$ - see front matter & 2013 ISA. Published by Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.isatra.2013.05.002

A. Kessal, L. Rahmani / ISA Transactions 52 (2013) 638643

639

Fig. 1. Boost converter circuit.

computation delays, hardware ltering for antialiasing, and software ltering for measurement noise reduction, where necessary.
Real-time implementation is done on an experimental prototype
using the dSPACE DS1104 controller board. This controller is
experimentally compared for steady-state performance and transient response over the entire range of input and load conditions
for which the system is designed. The paper is organized as
follows. In Sections 2 and 3, a description of converter, and a
design and analysis of controllers are given. The experimental
setup is detailed in Section 4. Section 5 presents the obtained
results with discussions.

2. Mathematical model of boost converter

The basic circuit diagram of the DCDC converter with front
end solid state input power factor conditioner used in the
proposed scheme is shown in Fig. 1.
The power circuit is that of an elementary step-up converter.
When the boost switch Sw is turned on (u 1), the inductor current
builds up, and energy is stored in the magnetic eld of the
inductor, whereas the boost diode D is reverse biased, and the
capacitor supplies power to the load. This is the rst mode
operation. As soon as the boost switch is turned off (u 0), the
power circuit changes mode, and the stored energy in the
inductor, together with the energy coming from the input AC
source, is pumped to the output circuitry (capacitorload combination). This is mode 2 of the circuit. Then the state space model
for the boost PFC in continuous current mode can be found by the
circuit analysis. The output voltage and inductor current dynamics
are governed by the variable structure real switched system.
8
< C dvo 1uiL io
dt
1
: L didtL vin 1uvo
In order to obtain a sinusoidal input current in phase with the
input voltage, the control unit should act in such a way that vin
sees a resistive load equal to the ratio of vin and iL. This has been
done by comparing the actual current passing through the
inductor with a current reference, which is derived from vin and
has an amplitude determined by the output voltage controller.

Fig. 2. Boost converter circuit governed by sliding mode controller.

inductor current via sliding mode that replace classical hysteresis

current control (Fig. 2).
This control of the output voltage of ACDC converter meets the
criteria of stability and existence of sliding mode. However, it is
difcult to determine the gains of the voltage loop since sliding
mode is a highly nonlinear method [2]. Furthermore, since SMC is
only applied to current regulation, the voltage loop will be more
sensitive to high frequencies phenomena and to uncertainties in
the reference current. In order to improve the performances of the
controller, a control mode based on a sliding surface which
involves output voltage will be treated.
Let (Vequ, Iequ) be the desired equilibrium point, where Vequ is the
output voltage, and Iequ is the inductor current peak at equilibrium
point. The input current peak IL can be expressed as [13]
IL

1
Vo
2
R

S 1 vo V ref 2 iL iref

The control objectives of the PFC are twofold: regulate the

output voltage vo to a reference voltage Vref and give the input
current iL a rectied sine waveform in phase with the rectied
voltage vin. The design of sliding mode controller for PFC starts
with the choice of sliding surface. As it is shown in [11], it is clear
that direct surface voVref can tend to zero only if the current
increases continuously. Usually, a cascade control structure is used,
which leads to solve the control problem using two control loops
[12]: an outer voltage loop which generates the reference current
from voltage error and an inner current loop which controls the

where 1 and 2R+.

The control by current imposes the average power passed to
the load with the ideal PFC pre-regulators [13].
P

V SM I ref
vo io
2

The reference current peak depends on the operating point; it

can be taken as
V SM I ref
io
vo

2V ref
1cos 2t
R1cos 2t

2vo V ref
io

1cos 2t
V SM R1cos 2t

I ref
3. Design of sliding mode controller



equilibrium point becomes V equ ; I equ =2 V equ
 So, the

1=R , and the sliding surface shall be given according to the
expression

Sliding surface coefcients (1, 2) should be chosen such that

the sliding mode exists at least around the desired equilibrium
point, and the dynamics of the system will reach the surface and
lead toward the equilibrium point.
3.1. Existence condition
The existence condition of sliding mode implies that both S_ and
_S will tend to zero (when t-), which means that the system
dynamics remains on the sliding surface. The existence condition
of the sliding mode is SS_ o0 (when S-0); achieving this

640

A. Kessal, L. Rahmani / ISA Transactions 52 (2013) 638643

inequality guarantees the existence of the sliding mode around the

switching surface.
The model of PFC can be written in a state space where the
equilibrium point is the origin, whether
x1 vo V ref and x2 iL iref where iref I ref jsin tj, so

the surface is given by


 

vin =L 10 x1 V ref =2 RC
 

ueq 1 
x1 V ref =L 10 x2 iref =2 C

8
< C dx1 1ux2 iref x1 V ref
R
dt

Replacing the equivalent control (12) in the state space

model (7) and from S 0, the dynamic of x1 at the sliding regime
is given as


  
2 
vin iref x1 10 =2 x1 V ref =R
dx1
 



13

dt
C x1 V ref L iref 10 x1 =2 10 =2

: L dx2 vin 1ux1 V ref

dt

Replacing iref by its value from (6) in the expression of the

commutation surface (3), the surface becomes




2vo V ref jsin tj
S 1 vo V ref 2 iL
V SM R1cos 2t

2vo V ref jsin tj

V SM R1cos 2t


22 V ref jsin tj
vo 2 iL 1 V ref
1
V SM R1cos 2t
1 vo V ref 2 iL 2

Introducing the Lyapunov function V 1=2x21 , its derivative is

V_ x1 x_ 1 , so


vin 1 =2 v0 V ref =R
2
_


14
V x1
Cx1 V ref Liref 10 x1 =2 10 =2
The condition for V_ to be negative is



 0 
LV 2ref =Rvin 10 =2 CV ref
10 x1
1
4 0x1 4
CV ref x1 L iref
2
2
C L10 =2 2
15

This equation can be written in the coordinate system (x1, x2):





22 V ref jsin tj
22 V ref jsin tj
x1 1
V
S 1
V SM R1cos 2t
V SM R1cos 2t ref
2 x2 2 iref 1 V ref 01 x1 2 x2

where 01 1 22 V ref jsin tj=V SM R1cos 2t

The state of the switch (u{0,1}) imposes two signs of the
derivative of the sliding surface; replacing u in the state system
(7), the boundaries of the sliding area are deduced by
8
< x1 10
RC

: x1

01

RC

V ref
RC

2 vLin 4 0

L2 x2

01
C

S_ 4 0

2 vLin

V ref
0 vref
L 1 RC

iref
C o0

S_ o 0

10

To ensure that the sliding mode exists at least around the

equilibrium point (x1 x2 0), the following condition must be
satised:
1
RCvin
o
2
V ref L

11

3.2. Stability condition

To ensure stability, the system dynamics during sliding mode is
directed to the desired equilibrium point. The goal is to determine
the dynamic of x1 and x2 when the sliding regime is achieved.
Taking the state space model in (7) and the commutation surface
in (8), from S_ 0, the equivalent average control that must be
applied to the system in order that the system state slides along

12

Based on the sliding region dened by (10) and the existence

condition (11), the condition given in (15) is always satised along
the sliding region of the commutation surface. According to the
theorem of Lyapunov stability, the system is globally asymptotically stable.
3.3. Controller parameters and system performance
Inequality (11) provides only general information concerning the
existence of sliding mode. On the other hand, performances of the
closed loop system are inuenced by the choice of parameters of the
controller, especially when the system presents large variations
around nominal. To choose these parameters, the size of the sliding
part in the switching surface must be taken into consideration.
Really, sliding condition is only satised on a subpart of the
surface and not on the entire surface as shown in (10). Therefore,
the controller parameters must be carefully chosen to ensure that
the system dynamic will intercept the commutation surface in the
sliding part. For this, precautions against unwanted behaviors that
cause an overshoot response must be taken (Fig. 3).
The two points A(x1A,x2A) and B(x1B,x2B) are assumed which are
respectively the crossing of system dynamic together with the
commutation surface, and the boundary of the sliding parcel
(Fig. 3). Supposing at t 0, the output voltage is Vini (initial voltage
value) and the input current is null, since the surface S will be
negative, then u 1 such that the space vector becomes
8
< C dx1 x1 V ref
R
dt
16
: L dx2 vin
dt

Fig. 3. Sliding mode with and without overshoot.

A. Kessal, L. Rahmani / ISA Transactions 52 (2013) 638643

Taking into consideration the initial values, the resolution of

this system gives
Lx2 iref

x1 V ini e

RC:vin

V ref

17

This solution represents the dynamic of the system before

intercepting the surface at point A. From S0 and from (16) we
deduce the coordinate of point A:


Lx2A iref
10

x2A
V ini e RC:vin V ref
18
2
From the equation of the commutation surface (9) and the limit
of sliding zone dened by (10), the coordinate of point B is
 0 
 !
 
  
1 V ref =RC iref =C =2 vin V ref =L
10
19
x2B
 


2 
2
10 =2 10 =2 RC 1=C
By deduction, the system intercepts the commutation surface in
the right part if the controller parameters are selected in such a way
that
OA o OB

20

The theory of sliding mode assumes that the hysteresis band

shall be null, so frequency approaches innity. It is obvious that
this assumption could not be made owing to the frequency
limitation caused by the feature of circuit components and losses.
Generally, a hysteresis window is added around the surface to
maintain the operating frequency; from Fig. 4, the rise time ton and
fall time toff can be expressed as
2
t on
S

t of f

1
f
t on t of f

21

From (9) and (21), the hysteresis band expression is deduced, so

1
2f

terms of steady state error. So the choice of controller parameters

must be taken into account to make the closed loop system less
sensitive to an error in the current part of the surface. It is
supposed that the measured reference current (IrefMEAS) can be
expressed as the sum of the expression of reference current given
by (4) and an error term (eIref), where e is dened as the error
percentage, so that
I ref Meas I ref eI ref

23

Replacing Irefmeas in the expression of sliding surface (3)



2 eV 2ref
2 eV ref
vo V ref 2 iL I ref
S 10
vin R
vin R

24

The average value of the term 2 eV 2ref =vin R is constant, so



2 eV ref
vo V ref vo 2 iL I ref iL
25
S 10
vin R
where vo, and iL, are the voltage and the current steady state
errors respectively.
Error part in the sliding surface is given by
2

10

2 eV ref
2 eV ref
vo 2 iL
vin R
vin R

26

The control u will tend S to zero, so vo -V ref vo and

iL -I ref iL ; from the balance of input/output power
V ref vo io vin I ref iL

1



  


  

1= 10 V ref =RC 1=vin 1 2 =L vin V ref
1= 2 =L vin 10 V ref =RC

Consequently, from the expression of (1,2), a limitation on

the choice of the controller parameters is maintained. In fact,
values of these parameters must ensure that the hysteresis
window is greater than the perturbation generated by the converter, and in the same time, the band value should be limited to
guarantee the robustness. In addition, the main goal is to regulate
the output voltage, but the commutation surface (3) depends on
the current errors, so an important optimization study of controller gains should be the analysis of the sensitivity of the
controller in front of a measurement or a current reference
estimation error.
Actually, the current reference is unknown; it can be extracted
from the load current (6). However, this latter can be measured or
observed through an extended Luenberger observer. In both cases,
an error can occur which affects the response of the system in

641

27

!
22

Then, expression of vo will be

v1
0
u
2
4eV 2ref
Rvin @1 V ref eV ref u
1 V ref eV ref
A
vo

2
2 Rvin Rvin
2 Rvin Rvin
v2 R2
in

28

4. Experimental setup
In this study, the gains 1 and 2 of sliding controller are
adjusted employing the off-line iterative genetic algorithm (GA).
Thus, GA determines the controller gains which are the most
compatible and provide optimum performance. The criteria are
based on practical specications. Thus the objective of the genetic
algorithm is to determine the values of parameters that ensure,
regardless of the operating point, that the system will intercept
the sliding part of commutation surface while respecting the
following conditions:
8
RCvin
1
>
>
< 2 o min V ref L
min o Max1 ; 2 o Max
>
>
: Maxv o v
o
oMax f ore eMax

29

The power circuit is designed to meet the following

specications:

Fig. 4. Hysteresis band of sliding regime.

output voltage V0 160 V

output voltage ripple o2%

642

A. Kessal, L. Rahmani / ISA Transactions 52 (2013) 638643

input voltage VSeff 115 V, RMS

input current ripple 5%
load resistance R 210

5. Results

The experimental prototype was built around the dSPACE1104

controller board, which hosts the PowerPC 603e processor, to
examine operating characteristic of the proposed method control
for PFC. Although the PowerPC and its associated data acquisition
circuitry can run up to 1 MHz, computation delay and communication overheads only allowed for the control algorithm to be
executed at 20 kHz. The fourth-order RungeKutta solver was
chosen to discretize the controller for real-time implementation.
One Hall-effect CT's LEM (PR30) and isolation amplier HAMEG
(HZ64) were employed to detect the inductor current, input line
voltage, and the output DC-bus voltage. Control circuits were built
for offset correction and appropriate scaling. To prevent aliasing in
the sampling process, second-order low-pass Butterworth lters
were used to remove noise and switching frequency ripple in the
sensed signals. Step load changes were affected by electronically
connecting/disconnecting a parallel load.

A real-time experimental study was performed to capture the

performance of the proposed method control for PFC. First, the
steady-state performance is evaluated in terms of output voltage
regulation, THD, and power factor. Next, the transient performance
is evaluated for output voltage response on application of load
step changes that are expected in practical applications of this
circuit. All the data presented here were captured at 20 kHz using
the control desk user interface for the dSPACE1104.
1. Steady-state performance: Fig. 5 shows the corresponding
experimental results: the obtained power factor is 0.998% and
THD is 2.95%; it is important to note that at nominal line and
load condition, the method control has a THD value below 3%
even with the limited bandwidth that is allowed by the digital
implementation. Line current is very close to sine wave and in
phase with the line voltage as shown in Fig. 6; the output
voltage error is about less than 2 V. These results show that the
proposed PFC control method achieves near unity power factor
under steady state, and THD value is much better than the
adoption of IEC1000-3-2 as the EN61000-3-2 standard.

[50V/div]

[2A/div]

[100ms/div]
Fig. 5. Experimental results for steady state, grid voltage, input current and
regulated output DC voltage.

Fig. 8. Experimental results for load disturbances, regulated output DC voltage and
input current.

Fig. 6. Experimental measurements, THD, PF and phase.

220
Vref, vo
200

Increasing load
180

Decreasing load
160

140
1.4

1.6

1.8

2.2

2.4

Fig. 7. Simulation results for load changes.

2.6

2.8

A. Kessal, L. Rahmani / ISA Transactions 52 (2013) 638643

240

643

Vref, vo

220

200

180

160

1.4

1.8

2.2

2.6

3.2

Fig. 9. Simulation results for output voltage reference changes.

[100V/div]

[2A/div]
[100ms/div]
Fig. 10. Experimental results for output voltage reference variations, regulated
output DC voltage and input current.

Transient performance: To evaluate performances in transient

2. mode, step load changes are effected by disconnecting (or
connecting) parallel load. The reference current amplitude is
limited to 3.5 A in the control method designs. Figs. 7 and 8
show, respectively, simulation and experimental results of
transient response for the proposed method control for PFC
for a load resistor step, by 7 33% of the nominal value of the
load (212 ). After a short transient (about 150 ms), the DC-bus
voltage is maintained close to its reference value with a good
approximation and stability. The line currents have nearly
sinusoidal waveforms.
The dynamic behavior of the proposed method under a step
change of Vref is presented in Fig. 9 for simulation and Fig. 10 for
experimental results. After a short transient (about 100 ms), the
DC-bus voltage is maintained close to its new reference (from
180 V to 220 V and vice versa) with good approximation and
stability. The line currents have nearly sinusoidal waveforms.
6. Conclusion
In this paper, a practical design of sliding mode control for
boost power factor controller is established, using a sliding surface
which includes all state variables, output regulated voltage and
input sinusoidal current. An optimization algorithm was

developed in order to calculate the optimal values of the sliding

surface parameters based on a predened specication. Results
show excellent dynamic response of controller and robustness to
load and voltage reference with large variations around nominal
values. Experimental results show excellent dynamic response,
good output regulation, low harmonic distortion, and high power
factor can be achieved with the proposed single-stage converter
and control scheme based on the proposed sliding mode controller. Finally, to verify the PFC function, the harmonic distortions
are measured and compared to the international standards as EN
61000-3-2 and IEEE 519, the power factor is near unity and the
THD is less than 3%.
References
[1]

Mingzhi, H, Jianping, X. Nonlinear PID in digital controlled buck converters.

In: Proceedings of the applied power electronics conference APEC 2007,
Anahein, CA, USA; 2007. pp. 14615.
[2] Guo Liping, Hung John Y, Nelms RM. Comparative evaluation of sliding mode
fuzzy controller and PID controller for a boost converter. Electric Power
Systems Research 2011;81(1):99106.
[3] Arulselvi S, Uma G, ChidambaramM. Design of PID controller for boost
converter with RHS zero. In: Proceeedings of the 4th International Power
Electronics and Motion Control Conference (IPEMC 2004), China; 2004.
[4] Yang Z, Sen PC. A novel technique to achieve unity power factor and fact
transient response in ac-to-dc converters. IEEE Transactions on Power
Electronics 2001;16(6):76478.
[5] Rashid M. Power electronics handbook. USA: Academic Press; 2001.
[6] Utkin VI. Sliding modes and their applications on variable structure system.
Moscow Russia: MIR publisher; 1978.
[7] Utkin VI, Chang H. Sliding mode control in electromechanical systems.
Mathematical Problems in Engineering 2002;8(45):45173.
[8] Tan SC, Lai YM, Cheung MKH, Tse CK. On the practical design of a sliding
mode voltage controlled buck converters. IEEE Transactions on Power
Electronics 2005;20(2):42537.
[9] Ben Saad K, Sahbani A, Benrejeb M. Design procedure and implementation of
a robust fuzzy sliding mode controller for buck converters. International
Review of Automatic Control 2008;1(3):30310.
[10] Hu ZB, Zhang B, Du GP, Zhong L , Deng WH. Fast transient three-level
converters with sliding-mode control. In: Proceedings of the applied power
electronics conference and exposition (APEC), vol. 3; 2005. pp. 143640.
[11] He Zhaolan, Zhang Shuli, Wu Junfeng. Sliding mode control of switched linear
systems based on common Lyapunov function. In: Proceedings of the 3rd
international symposium on systems and control in aeronautics and astronautics (ISSCAA); 2010. pp. 3004.
[12] Navarro-Lpez Eva M, Corts Domingo, Castro Christian. Design of practical
sliding-mode controllers with constant switching frequency for power
converters. Electric Power Systems Research 2009;79(5):796802.
[13] Kessal Abdelhalim, Lazhar Rahmani, Gaubert Jean-Paul, Mohammed
Mostefai. Analysis and design of an isolated single-phase power factor
corrector with a fast regulation. Electric Power Systems Research 2011;81
(9):182531.