This article has been accepted for publication in a future issue of this journal, but has not been
fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2018.2879881, IEEE Access
Backstepping Sliding Mode Control of FC-UC
based Hybrid Electric Vehicle
Muhammad Saud Khan, Iftikhar Ahmad∗ , Hammad Armghan, and Naghmash
Abstract—Depleting fossil fuels, and subsequently, high envi-
ronmental pollution has made hybrid electric vehicles (HEVs),
a vital replacement to its conventional counterpart. Electric
and HEVs offer a compact, lightweight & effective means of
transportation that is economical and has sufficient life cycle.
In this paper, the HEV under consideration uses two energy
sources to operate its traction motor: a fuel cell (FC) works as
the main source, whereas, an ultra-capacitor (UC) is used as
an auxiliary source. Each source is connected to the DC bus
via a power conditioning circuitry, which comprises electronic
power converters. The power conditioning unit harnesses energy
from the sources according to the load demand of the vehicle.
A nonlinear control scheme, called Backstepping Sliding mode
control, is employed on the power converters in the power
conditioning unit. The reason for applying this nonlinear control
scheme is due to the systems nonlinear nature, where this
control law provides a robust and precise output in presence
of these nonlinearities. Formal stability analysis of the proposed
HEV system is provided in this study and the simulations are
performed in MATLAB/Simulink environment. The simulation
results show perfect regulation of the DC bus voltage with very
fast convergence, low overshoots and almost zero steady state
error, which is our primary objective.
Index Terms—Hybrid electric vehicle (HEV), Backstepping
Sliding mode control, Fuel cell (FC), Ultra-capacitor (UC)
I. I NTRODUCTION
The internal combustion engines (ICEs) have been the main
driving unit of vehicles since their invention. ICEs use fossil
fuels to generate energy and use this energy to drive the
vehicle. Unfortunately, the depletion of conventional energy
resources and a sudden increase in the world’s temperature
due to global warming have made renewable energy sources,
a pressing need in vehicular technology. A major source of
greenhouse gases (GHG) emissions is transportation. A study
shows that 28% of the total GHG in the U.S. during 2016
was produced by emissions from vehicles and other forms of
transport [1].
Of the latest renewable energy systems, the hybrid electric
vehicle (HEV) system has proven to be most efficient and
capable in the scenario discussed above [2]. Most of the
previous work on HEVs show that these systems comprise
of two or more input sources, one being the main source and
the other as an auxiliary source. The main source has a high
energy density and provides the overall average power to the
vehicle. In most works, a fuel cell (FC) has been used as
the main source, because it works on hydrogen fuel and has
∗ Iftikhar Ahmad (corresponding author, iftikhar.rana@seecs.edu.pk) and all
others authors are with the School of Electrical Engineering & Computer
Science (SEECS), an institute of National University of Science & Technology
(NUST), Islamabad, Pakistan.
2169-3536 (c) 2018 IEEE. Translations and content mining are permitted for academic research only. Personal use is also permitted, but republication/redistribution requires IEEE permission. See
http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
zero carbon emissions, hence it is environmentally friendly
and proves to be economically sufficient when compared with
the traditional ICEs [3].
The auxiliary source provides power during high load
demands like vehicle start-up, acceleration, and uphill climb.
Therefore, a high power density source, such as an ultra-
capacitor (UC) is needed [4]. Moreover, the energy obtained
through regenerative braking can be utilized by reverse char-
ging the UC [5]. The combined energy of both the FC and UC
sources provide power to a DC Bus. Since traction motors used
in HEVs are mostly induction-type AC motors, an inverter is
placed between the DC Bus and the vehicle motor. The sole
purpose of this inverter power stage is to convert DC to AC,
which drives the traction induction motor. The highlights of
the operation of the FC-UC HEV are:
1) The FC provides power throughout the whole vehicular
operation. It can work continually as long as the hydrogen
fuel is available.
2) UC assists the main source (FC) during transients of the
load.
3) UC can be recharged through regenerative braking or
during the downhill drive.
4) The lifespan of a UC is 10-100 times more than batte-
ries, which makes them an ideal choice for continuous
charging and discharging during load transients.
In [6], different topologies and configurations are explained
using batteries, UCs and FCs. However, it does not highlight
the importance of a control system in an HEV. Also, there is
no distinction defined between the use of battery and UC in
this study.
The HEV system proposed in [7], [8], contains multiple
inputs, with some sources linked directly to DC bus, whereas
the others connected through bi-directional DC-DC converters.
But this system lacks DC bus voltage regulation. In studies
like [9]–[13], each source is connected through its own DC-
DC converter and equipped with an individual controller. This
scheme proves to be costly and complicated, as it requires
additional hardware and computational power. An economical
solution is the multi-input HEV, which is introduced in [14],
[15]. In multi-input HEV, all the inputs are controlled by a
single controller, making it relatively simple and economical.
However, active power sharing between multiple sources is
not considered in these studies.
In [16], [17] multi-input converters are employed with only
a linear PI controller in a feedback loop. The mathematical
model of the DC-DC boost converter used in these topologies
is inherently nonlinear, however, the model is linearized using
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2018.2879881, IEEE Access
small signal approximation. This linearization around a speci-
fic equilibrium point restricts the operating range of the DC-
DC converter and decreases its robustness [18]. An FC/UC
combination with nonlinear Lyapunov control is presented
in [19]. The proposed system obtains all of its error values
through an energy management system (EMS). The controller
then diminishes these errors to obtain stability and desired
output. In other words, the control scheme heavily relied on
the EMS for its operation. The EMS usually uses sensors
to measure quantities like the battery or UC’s state-of-charge
(SOC) and/or instantaneous power measurements of FC and
UC. These measurements can include some uncertainties in
their readings due to inaccurate sensors. Moreover, additional
components like data acquisition systems and decision-making
system are also required to implement the EMS algorithm.
The Backstepping Sliding mode control is a nonlinear
control technique, which provides the advantages of both the
Backstepping control and Sliding mode control. Studies like
[20], [21] employ this technique on DC-DC buck converters,
but can also be extended to the system under considera-
tion. Conventional Backstepping control is a recursive control
technique where the system is stabilized not only using the
system’s original control input but also using certain states as
virtual control inputs. However, simple Backstepping control
lacks robustness in presence of parametric variations, such as
in presence of varying output load current (Io ). Unlike simple
Backstepping control, Backstepping Sliding mode control can
be used to handle more than one system inputs and provides
more robustness in presence of varying system parameters.
This article is arranged as follows: Section II describes the
working principle of the FC, UC, and the DC-DC converters
employed in the selected HEV system. Section III expresses
the global mathematical model and the details of the para-
meters used in the construction of this model. Section IV
explains the design of the controller, with a description of its
control strategy. Section V illustrates and analyzes the results
obtained from the simulation. Section VI gives the conclusion
and future scope of this work.
II. O PERATIONAL A NALYSIS OF THE E LECTRIC
C OMPONENTS
A. Description & Behavior of Energy Sources
In this study, a parallel-type HEV system is selected, which
consists of an FC as the primary source and UC as the
auxiliary source as shown in Fig. 1. This section describes
the operation and power flow of each energy source, in detail.
Fig. 1: Layout of the selected FC-UC HEV
2169-3536 (c) 2018 IEEE. Translations and content mining are permitted for academic research only. Personal use is also permitted, but republication/redistribution requires IEEE permission. See
http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
A mathematical model of both FC and UC sources, along
with their respective DC-DC converters, is also derived and
analyzed below. The development and analysis of the mathe-
matical models are based on the FC’s and UC’s basic working
principle and their usage in the selected HEV system.
1) Fuel Cell (FC) Operation: A fuel cell (FC) generates
electricity from a chemical fuel, such as hydrogen fuel when
the hydrogen ions from the hydrogen fuel react with oxygen
or another oxidizing agent. In contrast to batteries, an FC can
provide electricity, as long as it is provided with its fuel. An
FC system consists of stacks of cells which are connected in
series. The series combination increases the overall voltage
of the FC system. There are different types of FCs, generally
classified by the electrolyte used for the chemical reaction.
The FC system considered in this work is proton exchange
membrane FC (PEMFC), which is a prime candidate for this
HEV application due to its compactness and its lowest start-up
time among other types of FCs.
The inherent characteristics of the FC produce a non-uniform
output voltage when FC current is varied [22]. Specifically,
the output voltage decreases as the output current increases.
Hence FC, in contrast to a constant voltage source, produces
an unsteady voltage output. Therefore, the control system in
an FC-HEV must possess the following characteristics:
i. The voltage of the FC system must be boosted to a high
DC bus voltage level, as the FC voltage is usually low
when compared to the DC Bus voltage. This is done by
introducing a DC-DC boost converter between FC and
DC bus.
ii. Usually, the auxiliary source(s), of an HEV system, pro-
duce a steady and unvarying output voltage in comparison
with FC. Therefore, the DC-DC converter also ensures a
steady and regulated FC voltage.
iii. The reason why FC is used as the main power source
in this configuration (Fig. 1) is the unlimited amount of
charge it can provide, as long as its hydrogen fuel is
available. This is also known as charge density. Because
of high charge density of the FC, most HEVs nowadays
offer a distance of 60 miles during normal cruising mode.
2) Ultra-capacitor (UC) Operation: The UC source is
very different from FC, in terms of its working principle,
power density, and the way it generates electricity. In its
origin, the UC is closer to its chemical substitute, the battery.
However, no chemical reaction takes place inside a UC and the
charge is stored electrostatically on the surface of its dielectric
material and its plates. There are many advantages of UCs
over batteries, such as higher power density (charging and
discharging rate), a higher number of life cycles [23], and
higher specific power i.e. higher energy storage with smaller
size and weight. Although UCs are more reliable, in a sense
that they produce more regulated output voltage as compared
to FCs, however, there are a number of reasons why a DC-
DC converter is used with the UC in HEVs. Some of them
are described below:
i. The output voltage of a UC is low in comparison with
the DC Bus voltage of an HEV.
ii. A UC can discharge and recharge automatically. Howe-
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2018.2879881, IEEE Access

ver, to control the time of discharging and recharging, a u1 : off, D1 : on =⇒ hiCo i(1−u1 )Ts = iL1 − i1
DC-DC converter is necessary.
Averaging over one complete time period, the state equations
iii. Active power sharing between UC, FC and the DC
are obtained as:
Bus is only possible in the presence of suitable DC-DC
converters. d vf c R1 vdc
hvL1 iTs = if c = − if c − (1 − u1 ) (1)
dt L1 L1 L1
B. Choice of Power converters d iL1 i1
hiCo iTs = vdc = (1 − u2 ) − (2)
The configuration, in which the two power sources and their dt Co Co
subsequent DC-DC power converters are arranged, is shown 2) UC Boost-buck converter: As the UC source has the
in Fig. 2. capability to be recharged by regenerative braking, a bi-
directional DC-DC converter is used between UC and DC
bus. This converter is also known as ”Boost-buck” because
it performs the operation of the boost converter when power
is transferred from the UC to DC bus and acts as a buck
converter, when the power is transferred vice versa. Since the
UC voltage is generally lower than the DC bus voltage, this
converter is the perfect solution in this situation. It contains
two switches S2 & S3 , with one inductor L2 and its internal
resistance R2 , as shown in the Fig. 2. Here, u2 is the PWM
gate signal on switch S2 & u3 is the PWM gate signal on
switch S3 . The boost and buck modes of the UC boost-buck
converter are given as,
u2 ∈ (0, 1) , u3 = 0 (off) (=⇒ boost mode)
u2 = 0 (off) , u3 ∈ (0, 1) (=⇒ buck mode)
The averaged state equations for each sub-cycle are given as:
d
For Boost mode: hvL2 iTs = iuc
dt
vuc R2 vdc
Fig. 2: Schematic of the DC power stage of the selected FC- = − iuc − (1 − u2 )
L2 L2 L2
HEV (3)
1) FC Boost Converter: As discussed earlier, the working d
For Buck mode: hvL2 iTs = iuc
phenomenon of a fuel cell is very different from normal dt
batteries and UCs. Unlike batteries, it cannot be recharged vuc R2 vdc
= − iuc − u3 (4)
when exposed to a compatible voltage source. Therefore, a uni- L2 L2 L2
directional DC-DC boost converter is used with the FC. The
boost converter amplifies the FC supply voltage and provides It is important to note that in steady state, the output current
a steady output voltage. Since FC operates throughout the of the UC converter (i2 ) depends upon its input current iuc
vehicular operation, the FC boost converter is used to stabilize and its duty ratio M (u2 , u3 ) by the relation:
and regulate the DC bus voltage to achieve its desired value. In i2 = M (u2 , u3 ) · iuc (5)
contrast, the auxiliary source provides the necessary amount
of current to the traction motor via DC bus and vice versa. The
III. G LOBAL M ATHEMATICAL M ODEL
FC boost converter contains a single switch S1 , in this case, a
metal oxide semiconductor field effect transistor (MOSFET), In this section, a global mathematical model is derived,
an inductor L1 with its internal resistance R1 , and an output which is necessary to develop a global control scheme for
capacitor Co , as shown in Fig. 2. The control signal of the our FC-UC HEV. It should be noted here that in section II-B,
switch S1 is represented here by u1 , which inherently takes its the equations for the system’s state variables were obtained
value between [0, 1]. By using Kirchhoff’s voltage and current by averaging the system quantities over one switching period.
laws on the circuit in Fig. 2, the state equations for each half- These system quantities were selected on the number of
cycle of FC or inductor current (if c or iL1 ) are obtained as: energy-storing devices (capacitors & inductors) used in a
power converter. The quantity, in which a passive element
u1 : on, D1 : off =⇒ hvL1 iu1 Ts = vf c − if c R1 stores its energy, becomes the state variable of the system
u1 : off, D1 : on =⇒ hvL1 i(1−u1 )Ts = vf c − if c R1 − vdc [24]. The reason behind this selection was that a change in a
state variable eventually represents a change in the system’s
For the output Bus voltage (Vdc ), the state equations, for each
energy.
sub-cycle, are obtained as:
The modes of UC can also be represented by the current flow
u1 : on, D1 : off =⇒ hiCo iu1 Ts = −i1 of the UC current iuc , i.e.

2169-3536 (c) 2018 IEEE. Translations and content mining are permitted for academic research only. Personal use is also permitted, but republication/redistribution requires IEEE permission. See
http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2018.2879881, IEEE Access

i. When the UC is supplying power to DC bus in boost IV. D ESIGN OF BACKSTEPPING S LIDING MODE
mode: iuc > 0 CONTROLLER
ii. When the UC is charging through regenerative power A. Controller Objectives
from the motor in buck mode: iuc < 0
The aims and objectives, to be achieved by the Backstepping
Hence, we can define a variable K, which represents the Sliding mode controller, are:
UC modes in the following way:
i. Steady and regulated DC bus voltage, with minimum
(
ref
steady state error under varying load conditions.
1, if Iuc > 0 V Boost mode ii. The FC supplies power throughout the vehicular opera-
K= ref
(6)
0, if Iuc < 0 V Buck mode tion and regulates the DC bus voltage to its desired value.
iii. The UC supplies and recharges its current, whenever
Hence the two separate equations for the boost and buck mode required by the motor load.
of the UC converter (3 & 4) can be combined by introducing iv. A robust and fast output response with lesser involvement
the variable K, such as, of the EMS. Shortcomings of EMS are already discussed
d vuc R2 vdc in section I.
iuc = − iuc − [(1 − K)u3 + K(1 − u2 )] (7) v. Stability of the whole proposed system.
dt L2 L2 L2
Additionally, we can redefine a virtual control input u23 for
an easier derivation of our control scheme, such that, B. Modifications in the Mathematical model
Backstepping Sliding mode control provides the advanta-
u23 = [(1 − K)u3 + K(1 − u2 )] (8) ges of both the Backstepping and the Sliding mode control
techniques. The Sliding mode control (SMC) provides finite
As we can see that the control input u23 act as the duty cycle
time convergence and better robustness in presence of any
of the PWM control signal of the UC Boost-buck converter.
parametric variation and the Backstepping control provides
This implies, in steady-state:
a recursive approach to minimize the errors using states as
M (u2 , u3 ) = U23 = hu23 i virtual controls. Also, the SMC possess a strong ability to
handle large input variations and load transients [25]. The
Using Kirchhoff’s current law on the output current node of Backstepping Sliding mode technique applies a Sliding mode
the DC-DC converters, control loop over the Backstepping controller [26]. The re-
sulting controller is capable of handling the uncertainties and
i1 = Io − i2 load disturbances correctly.
= Io − u23 · iuc (9) To apply the Backstepping Sliding mode control on the se-
lected HEV model eq.(10), the model first needs to be modified
Therefore, using equations (1), (2), (7), (8) & (9), the global into a compatible form. Let us define two new control variables
model for the FC-UC HEV system is given as: µ1 & µ2 as:
x3 R1 Vf c x3 Vf c
ẋ1 = −(1 − U1 ) − x1 + µ1 = −(1 − U1 ) +
L1 L1 L1 L1 L1
(11)
x3 R2 Vuc x3 Vuc
ẋ2 = −U23 − x2 + (10) µ2 = −U23 +
L2 L2 L2 L2 L2
x1 x2 Io Using the definitions of µ1 & µ2 from eq.(11), the model
ẋ3 = (1 − U1 ) + U23 −
Co Co Co eq.(10) can be re-written as:
The above mathematical model is in state equations form.
The state variables and control inputs are given as, R1
ẋ1 = − x1 + µ1 (12)
 T  T L1
x1 x2 x3 = hif c i hiuc i hvdc i R2
ẋ2 = − x2 + µ2 (13)
&  L2   
T T Vf c L1 µ1 x1 Vuc L2 µ2 x2 Io
ẋ3 = − + − −
 
U1 U23 = hu1 i hu23 i (14)
x3 x3 Co x3 x3 Co Co
Where, hif c i = average FC current, hiuc i = average UC Now in the above form, µ1 & µ2 are the control inputs in
current, hvdc i = average DC bus voltage, hu1 i = average value equations (12) & (13) respectively.
of control input u1 , hu23 i = average value of control input u23 . If we observe the eq.(14), we can see that all the nonlinear
terms are gathered in this equation. Therefore, in order to
Another important observation, which is eminent from the control the UC current i.e. x2 , a simple proportional-integral
model eq.(10), is that the model is a nonlinear, multi-input- (PI) controller is applicable, as eq.(13) now represents a linear
multi-output (MIMO) system. Deriving a control law is much subsystem. But the DC bus voltage (x3 ) depends upon the FC
difficult for such a complex system. current (x1 ) for its desired value. Therefore, the Backstepping
Sliding mode controller is designed using equations (12) &

2169-3536 (c) 2018 IEEE. Translations and content mining are permitted for academic research only. Personal use is also permitted, but republication/redistribution requires IEEE permission. See
http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2018.2879881, IEEE Access

(14) together, to achieve our main task of regulating the

DC bus voltage to its desired value. Now, multiplying and
expanding the terms of eq.(14), we can obtain,
 
Vf c Vuc x3 L1 µ1 L2 µ2
ẋ3 = x1 + x2 − Io − x1 − x2
x3 C o Vf c Vf c Vf c Vf c
In the above equation, it can be observed that the two
rightmost terms, inside the parenthesis, contains the control
inputs µ1 & µ2 . To account for input variations, let the
system’s input disturbance be
L1 µ1 L2 µ2
d=− x1 − x2 (15)
Vf c Vf c
Fig. 3: Phase plane diagram of SMC [27]
The above equation defines the sliding mode input distur-
bance which is not very significant when the trajectories are
away from the sliding surface. However, once the trajectories D. Stability Analysis
reach the sliding surface, the disturbance d becomes significant
The next step in designing a nonlinear controller is defining
enough that it cannot be neglected. Hence the input disturbance
an energy or energy-like function for the selected HEV system.
(d) is neutralized by the sliding mode control’s subpart: the
Let us define a Lyapunov function candidate as:
switching control, whereas the rest of the terms are stabilized
using the sliding mode control’s other subpart: the nominal 1 2
S V = (18)
control. Using the definition of the disturbance (d), the mo- 2
dified HEV model can be represented by the following set of Taking the derivative of the Lyapunov function candidate
dynamical equations: from eq.(18) with respect to time gives
R1 V̇ = S Ṡ
ẋ1 = − x1 + µ1
L1
R2 From equations (16) & (17), putting the derivative of sliding
ẋ2 = − x2 + µ2 (16) surface function S in the above equation, gives
L2  
Vf c Vuc x3 V̇ = S · ẋ3 [∵ the derivative of a constant (Vref ) is zero]
ẋ3 = x1 + x2 − Io + d
x3 C o Vf c Vf c
 
Vf c Vuc x3
=S· x1 + x2 − Io + d (19)
x3 C o Vf c Vf c
C. Selection of Sliding Surface
In the 19, the state x1 behaves as a virtual control input for the
The Sliding mode control can further be divided into two Backstepping control loop. But x1 is only a state variable and
parts: not a real control input. Thus it cannot be enforced throughout
1. The nominal control un , which is used to reach the sliding the vehicular operation t > 0. However, if x1 = Iref 1 then our
surface S, as shown in the Fig. 3 from time t0 to t1 . The desired objective is achieved. The following equations show
nominal control can be described as, our desired value for x1 ,
(
1 ⇒ ”ON”, when S > 0
un = Iref 1 = x1 = un + us
0 ⇒ ”OFF”, when S < 0
Vuc x3 x3
2. Once the trajectory reaches the sliding surface S, the =− x2 + Io − Co σS −ρsgn(S) (20)
Vf c Vf c Vf c | {z }
switching control us keeps the trajectory, as close to it | {z } switching control
as possible. It can also be seen in the Fig. 3 after the nominal control

time t1 . For this purpose, the switching control is usually where un represents the nominal control and us represents the
taken as, switching control. Substituting eq.(20) into the eq.(19), we get
us = sign(S) Vf c

x3

V̇ = S · −Co σS − ρsgn(S) + d
Hence, a simple sliding surface can be selected as, C o x3 Vf c
 
Vf c |x3 |
S = vdc − Vref ≤ |S| −Co σ |S| − |ρ| sgn(S) + |d|
Co |x3 | Vf c
or
when |ρ| ≥ |d|, the above equation yields
S = x3 − Vref (17)
V̇ ≤ −σS 2 (21)
So that when S = 0, we reach our main objective of
tracking DC bus voltage (x3 ) to its desired value (Vref ). Since V̇ is negative definite, the Lyapunov stability criterion
Further analysis and results validate our selection of the sliding is proved. The sliding surface S is globally asymptotically
surface. stable, which means that the trajectory reaches and stays at

2169-3536 (c) 2018 IEEE. Translations and content mining are permitted for academic research only. Personal use is also permitted, but republication/redistribution requires IEEE permission. See
http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2018.2879881, IEEE Access

the sliding surface (S = 0) ∀ x. Thus x3 = vdc = Vref . chattering phenomenon, and the two PI controller gains, are
The above stability is only possible when the virtual control given in Table III. The selection of the simulation parameters,
error e1 diminishes. To eliminate e1 , a PI controller is em- given in Table III, is based on the trial and error method.
ployed. The output of this PI controller is used to control There are a few systematic techniques to evaluate these gains
the inputs of the HEV model eq.(10). Therefore, in physical like genetic algorithm based gain estimation [32]. However,
implementations, the most common practice nowadays is the trial and error
Z approach as it does not require additional computational power
u1 = Kp1 e1 + Ki1 e1 dt (22) and complex algorithms. It is important to note here that all
Z of these parameters must be positive because these ensure the
u23 = Kp2 e2 + Ki2 e2 dt (23) stability of the system.

where,
e1 = Iref1 − x1
ref
e2 = Iuc − x2
As we know that the sgn(S) function switches its value
instantly at t = 0, it creates chattering along the sliding surface
(S = 0). The chattering phenomenon creates unnecessary
noise and switch stress in mechanical and power electronics
systems [28]. To alleviate this phenomenon, a saturation
function is introduced to ”smooth-out” the transients near the
sliding surface because the saturation function does not change
its value instantly [26]. This saturation function is given as,
S high motor load or acceleration mode, whereas low-speed
us = ρ , ε>0
|S| + ε
where ε is the boundary layer value. With the improved
switching function, the final control law or the reference value
B. Results from EUDC statndard
The results of this HEV system, operating under a standard
driving cycle, are presented in the Figs. 4–10. These graphs
show the UC charging and discharging under different tran-
sients of load current, with constant FC current. A constant
FC current is employed to portray our basic requirement of the
baseline power from the FC source. The proposed HEV system
is tested under the standard European extra-urban driving cycle
(EUDC) for a better and complete analysis of its performance.
In this way, EUDC assesses the effectiveness of the proposed
control scheme in the HEV system. Fig. 4 shows the vehicle
speed profile for the EUDC cycle. High-speed values represent
(24) values represent low motor load or deceleration mode.
EUDC Vehicle Speed vt (km/h)
100

Iref1 of the proposed Backstepping Sliding mode control is

80
given as,
Speed (km/h)

Vuc x3 x3 S 60
Iref1 = − x2 + Io − Co −ρ (25)
Vf c Vf c Vf c |S| + ε 40

For simulation purposes, the states x1 & x2 are considered

20
measurable, therefore, there is no need for a state observer
in this control scheme. This assumption is rightly justified 0
0 50 100 150 200 250 300 350 400
because the aforementioned states are physically inductor Time (in seconds)
current and capacitor voltage, respectively. Both of these
quantities can be measured using an ammeter and a voltmeter. Fig. 4: EUDC vehicle speed profile
Although to avoid inaccurate data readings from the sensors,
state observers can be used. The performance of a control Figure 5 shows the varying load transients for EUDC. This
scheme can be significantly improved by using a state observer load current profile Io is generated by keeping in mind the
[29]. The implementation of a state observer (sliding mode vehicle specifications, the traction induction motor characte-
based) is recommended for future work. For more details, see ristics, and a non-ideal inverter efficiency of 75%. The relation
studies like [30], [31]. of load current with vehicular specifications can be given as,
 
V. S IMULATION R ESULTS & A NALYSIS 1 1
2 d
Io = ρair vt ACx + Mt gCr + Mt vt vt
Vref 2 × ηinverter dt
A. Selection of Simulation Parameters (26)
The proposed HEV control scheme is simulated in MAT- where vt represents the speed of the vehicle, Vref is the desired
LAB/Simulink software to test its performance under varying DC Bus voltage and ρair is the air density. The rest of the
conditions. In order to highlight the comparison of this work parameters are listed in Table IV. Same parameters of vehicle
with the previous ones, the parameters and specifications of the and motor are chosen to compare the performance of the
power sources are selected the same as from [19] and are given proposed HEV system with recent works, such as [19]. Due to
in Table I. The selected parameters of the DC-DC converters the lack of computational power, the EUDC is adjusted from
are listed in Table II. The values of the simulation parame- a total time of 400 seconds to a shorter time of 5 seconds
ters: the overall stability parameter σ, disturbance rejection but the shape of load profile curve is kept the same. This
parameter ρ, the boundary layer gain ε for the alleviation of assumption is justified because the FC source is restricted to a

2169-3536 (c) 2018 IEEE. Translations and content mining are permitted for academic research only. Personal use is also permitted, but republication/redistribution requires IEEE permission. See
http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2018.2879881, IEEE Access

constant base current, whereas the UC provides power during 100

EUDC Result: x1

load transients. Hence the FC is providing the low-frequency 90

if c
iref
fc

part of the load profile and the UC is responsible for the high- 80

frequency parts of the load profile. 70

80

60
22

20

Current (A)
60 18
40
16
EUDC load current profile 50 20
14
60 0
40 0 5 10 15 20 ×10
-3
12
Iload 4.5 4.55 4.6 4.65 4.7 4.75

40 Zero line 30
Load Current profile (A)

20
20
10

0
0 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
Time (s)

-20
Fig. 7: Plot of state x1 (FC current)
-40

EUDC Result: x2
-60 40
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 iuc
Time (in seconds) iref
16.5 uc
20
Fig. 5: Applied motor load current of the vehicle Io under 16.45

EUDC 0 16.4
0.699 0.7 0.701 0.702

Current (A)
-20

EUDC Result: x3 -40 -5

-74.9
330 -10
330 -75
vdc -15
ref -75.1
320 320 vdc -60 -20
-75.2

310 0 5 10 15 ×10 -3 -75.3

310 300
4.599 4.6 4.601 4.602

-3
-80
2 4 6 8 10 12 ×10 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
300 Time (s)
Voltage (V)

290
300.2
Fig. 8: Plot of state x2 (UC current)
300
280
299.8
299.6

270 299.4
299.2

260
4.45 4.5 4.55 4.6 4.65 4.7
current changes its value abruptly. However, the controller
250
successfully regulates the system after a very short time,
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
Time (s) which shows excellent robustness in the presence of system
Fig. 6: DC Bus voltage under varying load conditions of parameter variations. It should be noted that in the HEV
EUDC simulation, no manual reference value was given to the FC
current and the system behavior is entirely controlled by the
Fig. 6 shows the DC Bus voltage vdc , tracking its reference UC current reference value. This makes the HEV system less
value of 300V DC. Initially, the transient response lasts for reliant on the EMS, which was one of our primary objectives.
about 0.04 seconds. This is the stabilization or settling time the Figures 9 & 10 show the plots of control inputs u1 & u23
controller takes to achieve its reference and optimize the HEV applied at the gate of MOSFET switches of the DC-DC power
system according to our objectives. Later on, small variations converters.
can be seen in the zoomed-in sections of the fig. 6 at time
EUDC Result: u1
instants where there is a sharp change in the load current Io 1

value, however, each time the state eventually converges to 0.9

0.8
its reference after a short period of time. A very small steady
0.7
state error can also be observed in fig. 6 but it is small enough
0.6
Duty ratio

to be negligible. 0.5
The plots of state variables iuc & if c can be seen in figures 0.4

7 & 8. It can be observed that the FC current if c is at a 0.3

constant value of around 20 A, which is used to provide the 0.2

base-line or average power to the vehicle, and the UC current 0.1

u1
supplies and recharges the extra load current. Fig. 8 shows 0
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
Time (s)
that the UC current tracks its reference value correctly. The
zero line indicates the difference between the charging and Fig. 9: Plots of Control signals u1
the discharging mode of the UC. Top half plane indicates the
discharging mode (where Io > if c ), whereas the bottom half
plane represents the charging mode of UC (Io < if c ). Slight
variations can be observed in the fig. 7 initially, which die C. Results from comparison with other control schemes
down after the system transients have passed. Later on, small To further highlight the performance of the proposed HEV
variations are also observed at time instants where the load system, a comparison with a recent study [19] is done in this

2169-3536 (c) 2018 IEEE. Translations and content mining are permitted for academic research only. Personal use is also permitted, but republication/redistribution requires IEEE permission. See
http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2018.2879881, IEEE Access

EUDC Result: u23 FC current (x1 )

1 120
iref
fc
110 Backstepping sliding if c
100
100
0.95 90

80

80 70

Current (Amps)
60
0.9 100
Duty ratio

-3
0 5 10 15 ×10 60

50 80
60
40
60
0.85 30

20 40
40
10
20
0.4 0.405 0.41 0.415 0.7 0.701 0.702 0.703 0.704 0.705
0.8
20

u23
0.75 0
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Time (s) Time (s)

Fig. 10: Plots of Control signals u23 Fig. 12: Plot of state x1 (FC current)

DC Bus voltage (x3 ): Comparison plot

450
section. For better comparison, the DC-DC converter parame-
ters are kept the same as given in the previous study. However, 400

to observe the differences in the transient responses of the two 400 420
420
350
systems, the load current profile is assumed as given in fig. 350
400

Voltage (volts)
300 400
380

11. This is an example load current profile and the values of 300
250
360
380

200

the currents are assumed here to compare the performance of 0 0.005 0.01 0.015 0.02 0.025
0.4 0.405 0.41 0.415 0.42 0.425
360
0.7 0.705 0.71 0.715

the proposed system with the previous study. This assumed 250
zoom-in view of e ss

400
load current profile is valid because the highest load current 200 398 ref
vdc
value is assumed for the vehicle in acceleration mode and the 396 Backstepping sliding vdc
Lyapunov Redesign vdc
0.52 0.54 0.56 0.58 0.6

lowest load current value is assumed for deceleration mode, 150

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Time (s)
as can be seen in fig. 11. Hence this load current profile
is consistent with the basic operation of a vehicle. For this Fig. 13: Comparative plots of state x3 (DC Bus voltage)

Iload
80

Acceleration
This proves that the HEV system, using the Backstepping
70
Sliding mode control, provides better performance than the
60 Lyapunov based control used in [19].
Normal Mode
Current (A)

50

40
30
20
10
0 0.1 0.2 0.3 0.4
Fig. 11: Example load current of the vehicle Io
part of the simulation, the UC current is kept at a constant
reference of 10 A, whereas, the FC is used to provide the
ref
extra load to the vehicle. The DC bus voltage reference (vdc
is also changed to 400 V. The comparative results of the states
x1 & x3 are given in the figs. 12 & 13. The FC current
reference (iref
f c ) is again generated here by the Backstepping
Sliding mode scheme itself, according to the demand of the
output load current (Io ) and the UC current reference (irefuc ).
Fig. 12 shows very fast convergence of the FC current (if c )
to its reference. A small overshoot can be observed initially,
however, this transient behavior is quite normal for these types
of systems. Fig. 13 shows the two DC bus voltage (vdc )
waveforms obtained using the Backstepping Sliding mode
control and the Lyapunov based control presented in [19]. The
results from the Backstepping Sliding mode controller show
fast convergence, lower overshoots and lesser steady state error
than the results obtained using the Lyapunov based control.
VI. C ONCLUSION
In this work, an HEV system has been presented, with a
Deceleration PEMFC as the main driving source and a UC as the auxiliary
source, for the optimum energy use in modern vehicular
0.5 0.6 0.7 0.8 0.9 1
systems. Power converters have been used to provide dynamic
Time (t)
energy exchange between the sources, as mentioned above. A
nonlinear control scheme, called ”Backstepping Sliding mode
control”, has been implemented on the selected HEV system.
The resulting controller provides fast reference tracking and
performs satisfactorily in presence of load variations. As a sub-
) objective, the use of an energy management system (EMS)
has been limited and the dependency of the HEV system
on its EMS has been decreased as compared to its former
works. MATLAB/Simulink is used to test the performance of
the HEV system under a standardized European extra-urban
driving cycle (EUDC). Simulation results have indicated that
the FC and UC behave satisfactorily and accurately, according
to our objectives. Hence, with the provided analytical and
simulation proofs, the system’s stability and the performance
of the whole proposed system turn out to be satisfactory.
Currently, only two DC-DC converters have been employed
to integrate both FC and UC. In future, any other suita-
ble converter topology can be applied, which reduces the
number of power converters, which in turn reduces the size,
complexity, and cost of the whole HEV system. Also, the

2169-3536 (c) 2018 IEEE. Translations and content mining are permitted for academic research only. Personal use is also permitted, but republication/redistribution requires IEEE permission. See
http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2018.2879881, IEEE Access
control technique applied in this study is a nonlinear technique,
called Backstepping Sliding mode control. However, another
nonlinear control technique, called adaptive control scheme,
can be implemented to produce a controller that does not
require the knowledge of the load current profile [33], [34]. It
may result in a more robust and stable HEV system. Moreover,
other combinations of energy sources, such as photovoltaic
(PV) array, can be used to optimize, conserve, and maximize
the energy available in an HEV system.
The scope of this work is limited to the modeling and
control of the initial power stage, which excludes the inverter
system and the traction motor connected to the DC Bus. In
the future, the inverter and the traction/induction motor can
also be modeled and can be included in the global model of
the HEV system.
R EFERENCES
[1] United States Environmental Protection Agency EPA. Sources of
greenhouse gas emissions. https://www.epa.gov/ghgemissions/sources-
greenhouse-gas-emissions [Accessed 26-04-18].
[2] Ali Emadi, Sheldon S Williamson, and Alireza Khaligh. Power electro-
nics intensive solutions for advanced electric, hybrid electric, and fuel
cell vehicular power systems. IEEE Transactions on Power Electronics,
21(3):567–577, 2006.
[3] Kaushik Rajashekara. Hybrid fuel-cell strategies for clean power
generation. IEEE Transactions on Industry Applications, 41(3):682–689,
2005.
[4] Mamadou Baı̈lo Camara, Hamid Gualous, Frederic Gustin, and Alain
Berthon. Design and new control of dc/dc converters to share energy
between supercapacitors and batteries in hybrid vehicles. IEEE Tran-
sactions on Vehicular Technology, 57(5):2721–2735, 2008.
[5] MK Yoong, YH Gan, GD Gan, CK Leong, ZY Phuan, BK Cheah,
and KW Chew. Studies of regenerative braking in electric vehicle. In
Sustainable Utilization and Development in Engineering and Technology
(STUDENT), 2010 IEEE Conference on, pages 40–45. IEEE, 2010.
[6] Ali Emadi, Young Joo Lee, and Kaushik Rajashekara. Power electronics
and motor drives in electric, hybrid electric, and plug-in hybrid electric
vehicles. IEEE Transactions on industrial electronics, 55(6):2237–2245,
2008.
[7] Shuai Lu, Keith A Corzine, and Mehdi Ferdowsi. A unique ultracapaci-
tor direct integration scheme in multilevel motor drives for large vehicle
propulsion. IEEE Transactions on Vehicular Technology, 56(4):1506–
1515, 2007.
[8] Mamadou Baı̈lo Camara, Hamid Gualous, Frederic Gustin, Alain
Berthon, and Brayima Dakyo. Dc/dc converter design for supercapacitor
and battery power management in hybrid vehicle applicationspolyno-
mial control strategy. IEEE Transactions on Industrial Electronics,
57(2):587–597, 2010.
[9] Sathish Kumar Kollimalla, Mahesh Kumar Mishra, and N Lakshmi
Narasamma. Design and analysis of novel control strategy for battery
and supercapacitor storage system. IEEE Transactions on Sustainable
Energy, 5(4):1137–1144, 2014.
[10] Alireza Payman, Serge Pierfederici, Farid Meibody-Tabar, and Bernard
Davat. An adapted control strategy to minimize dc-bus capacitors of
a parallel fuel cell/ultracapacitor hybrid system. IEEE Transactions on
Power Electronics, 26(12):3843–3852, 2011.
[11] Ahmad Saudi Samosir and Abdul Halim Mohamed Yatim. Implemen-
tation of dynamic evolution control of bidirectional dc–dc converter
for interfacing ultracapacitor energy storage to fuel-cell system. IEEE
Transactions on Industrial Electronics, 57(10):3468–3473, 2010.
[12] Abdallah Tani, Mamadou Baı̈lo Camara, and Brayima Dakyo. Energy
management based on frequency approach for hybrid electric vehicle
applications: fuel-cell/lithium-battery and ultracapacitors. IEEE Tran-
sactions on Vehicular Technology, 61(8):3375–3386, 2012.
[13] Majid Zandi, Alireza Payman, Jean-Philippe Martin, Serge Pierfederici,
Bernard Davat, and Farid Meibody-Tabar. Energy management of
a fuel cell/supercapacitor/battery power source for electric vehicular
applications. IEEE transactions on vehicular technology, 60(2):433–
443, 2011.
2169-3536 (c) 2018 IEEE. Translations and content mining are permitted for academic research only. Personal use is also permitted, but republication/redistribution requires IEEE permission. See
http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
[14] Alireza Khaligh, Jian Cao, and Young-Joo Lee. A multiple-input dc–dc
converter topology. IEEE Transactions on power electronics, 24(3):862–
868, 2009.
[15] Farzam Nejabatkhah, Saeed Danyali, Seyed Hossein Hosseini, Mehran
Sabahi, and Seyedabdolkhalegh Mozaffari Niapour. Modeling and con-
trol of a new three-input dc–dc boost converter for hybrid pv/fc/battery
power system. IEEE Transactions on power electronics, 27(5):2309–
2324, 2012.
[16] Furkan Akar, Yakup Tavlasoglu, Enes Ugur, Bulent Vural, and Ismail
Aksoy. A bidirectional non-isolated multi input dc-dc converter for
hybrid energy storage systems in electric vehicles. 2015.
[17] Luca Solero, Alessandro Lidozzi, and Josè Antenor Pomilio. Design of
multiple-input power converter for hybrid vehicles. IEEE transactions
on power electronics, 20(5):1007–1016, 2005.
[18] Rodrigo Ferreira, André Ferreira, and Pedro Barbosa. Comparative
study of linear and non-linear control techniques applied to dc-dc boost
converter as a voltage regulator.
[19] Hassan El Fadil, Fouad Giri, Josep M Guerrero, and Abdelouahad Tahri.
Modeling and nonlinear control of a fuel cell/supercapacitor hybrid
energy storage system for electric vehicles. IEEE Transactions on
Vehicular Technology, 63(7):3011–3018, 2014.
[20] Li-kui Yi, Jun Zhao, and Dan Ma. Adaptive backstepping sliding mode
nonlinear control for buck dc/dc switched power converter. In 2007 IEEE
International Conference on Control and Automation, pages 1198–1201.
IEEE, 2007.
[21] Shoei-Chuen Lin and Ching-Chih Tsai. Adaptive voltage regulation of
pwm buck dc-dc converters using backstepping sliding mode control.
In Proceedings of the 2004 IEEE International Conference on Control
Applications, 2004., volume 2, pages 1382–1387 Vol.2, Sept 2004.
[22] J Marshall and M Kazerani. Design of an efficient fuel cell vehicle
drivetrain, featuring a novel boost converter. In 31st Annual Conference
of IEEE Industrial Electronics Society, 2005. IECON 2005., pages 6–pp.
IEEE, 2005.
[23] Santiago Miret. Storage wars: Batteries vs. supercapacitors.
http://berc.berkeley.edu/storage-wars-batteries-vs-supercapacitors/, No-
vember 2013.
[24] Robert W Erickson and Dragan Maksimovic. Fundamentals of power
electronics. Springer Science & Business Media, 2007.
[25] Siew-Chong Tan, Yuk-Ming Lai, Martin KH Cheung, and CKM Tse. On
the practical design of a sliding mode voltage controlled buck converter.
IEEE Transactions on Power Electronics, 2005.
[26] Chao-Ming Lee, Shin-Han Han, Chen-Hong Zheng, and We-Song Lin.
Power split of fuel cell/ultracapacitor hybrid power system by backstep-
ping sliding mode control. In 2012 10th International Power & Energy
Conference (IPEC), pages 538–543. IEEE, 2012.
[27] ”Intech” [Online]. Basic sliding mode control principle.
http://www.intechopen.com/source/html/17687 /media/image63.png.
[Accessed 19-04-18].
[28] Vadim Utkin and Hoon Lee. Chattering problem in sliding mode
control systems. IFAC Proceedings Volumes, 39(5):1, 2006. 2nd IFAC
Conference on Analysis and Design of Hybrid Systems.
[29] Jianxing Liu, Yabin Gao, Xiaojie Su, Maxime Wack, and Ligang Wu.
Disturbance-observer-based control for air management of pem fuel cell
systems via sliding mode technique. IEEE Transactions on Control
Systems Technology, (99):1–10, 2018.
[30] Ligang Wu, Yabin Gao, Jianxing Liu, and Hongyi Li. Event-triggered
sliding mode control of stochastic systems via output feedback. Auto-
matica, 82:79–92, 2017.
[31] Jianxing Liu, Hao An, Yabin Gao, Changhong Wang, and Ligang Wu.
Adaptive control of hypersonic flight vehicles with limited angle-of-
attack. IEEE/ASME Transactions on Mechatronics, 23(2):883–894,
2018.
[32] Wei-Der Chang, Rey-Chue Hwang, and Jer-Guang Hsieh. A self-tuning
pid control for a class of nonlinear systems based on the lyapunov
approach. Journal of Process Control, 12(2):233–242, 2002.
[33] Bin Wang, Jun Xu, Binggang Cao, and Bo Ning. Adaptive mode switch
strategy based on simulated annealing optimization of a multi-mode
hybrid energy storage system for electric vehicles. Applied energy,
194:596–608, 2017.
[34] Bin Wang, Jun Xu, Rong-Jong Wai, and Binggang Cao. Adaptive
sliding-mode with hysteresis control strategy for simple multimode
hybrid energy storage system in electric vehicles. IEEE Trans. Industrial
Electronics, 64(2):1404–1414, 2017.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2018.2879881, IEEE Access

10

TABLE I: Power sources specifications

Power Source Specs
Proton exchange membrane FC 200 V, 200 A, 48 kW
(PEMFC)
UC Module 2.5 V (single cell), 141 cells (in
series), 1500 F

TABLE II: Component values of the DC-DC converters

Parameter Value
Inductors L1 &L2 3.3 mH
Inductor resistance R1 &R2 20 mΩ
Output Capacitance Co 1.66 mF

TABLE III: Values of simulation parameters

Parameter Value
σ 10
ρ 2500
ε 1000
Kp1 20
Ki1 1000
Kp2 0.1
Ki2 20

TABLE IV: Vehicle, motor & inverter specifications

Parameter Value
Total mass of the vehicle Mt 1922 kg
Aerodynamic drag coefficient Cx 0.3
Vehicle frontal area A 2.5 m2
Rolling resistance coefficient Cr 0.01
Gravitational acceleration constant g 9.8 m/s2
Inverter efficiency ηinverter 75%

2169-3536 (c) 2018 IEEE. Translations and content mining are permitted for academic research only. Personal use is also permitted, but republication/redistribution requires IEEE permission. See
http://www.ieee.org/publications_standards/publications/rights/index.html for more information.