Received: 11 February 2020 Revised: 24 February 2021 Accepted: 11 March 2021 IET Power Electronics

DOI: 10.1049/pel2.12120

ORIGINAL RESEARCH PAPER

Disturbance observer-based sliding mode control for dynamic

performance enhancement and current-sensorless of buck/boost
converter

Mei Su1,2 Weiquan Feng1,2 Taowen Jiang1,2 Bin Guo1,2 Hui Wang1,2
Chengyan Zheng1,2

1
Department of Electrical Engineering, Central Abstract
South University, Changsha, Hunan 410083, China
Bidirectional DC-DC converters are widely used in electric vehicle charger, where excellent
2
Hunan Provincial Key Laboratory of Power steady-state and dynamic performances are crucial. However, those two performances are
Electronics Equipment and Grid, Changsha, Hunan
410083, China
easily affected by external disturbances and parameter uncertainties. To solve this prob-
lem, this paper proposes a disturbance observer-based second-order sliding mode control
Correspondence (SOSMC) for the buck/boost converter. With the super-twisting algorithm (STA) second-
Hui Wang, Department of Electrical Engineering, order sliding mode control method, satisfactory steady-state and dynamic performance is
Central South University, Changsha, Hunan 410083,
achieved. In addition, to further improve the dynamic performance and reduce the control
China.
Email: [email protected] burden, a high-order sliding mode disturbance observer (HOSMDO) is employed to esti-
mate the lumped disturbances. Thus, the mismatched disturbances caused by input volt-
age fluctuation, load variation, output voltage regulation, power flow change, and model
uncertainties can be effectively suppressed. Moreover, the state variable containing induc-
tor current is also observed, which removes one current sensor, reducing system cost. The
design methods of the second-order sliding mode control and the high-order sliding mode
disturbance observer are presented as well as the stability analysed through the Lyapunov
theory. Simulation and experimental results verify the effectiveness of the proposed control
strategy.

1 INTRODUCTION In [4], the stability of linear control based on linear small-

signal model is easily disturbed by non-linear factors. And the
Bidirectional DC-DC converters are widely used in electric vehi- passivity-based control [5] and non-linear current-mode con-
cle charging system, energy storage system and many other trol [6] have weak ability to suppress large disturbance. Further-
fields due to its high power density, bidirectional power flow, more, adaptive back-stepping control [7], fuzzy logic control [8]
and easy integration. Among them, the LLC resonant con- and model predictive control (MPC) [9] need precise parameters
verter cascaded buck/boost converter is one of the most com- and mathematical model to achieve high performance. Though
monly used one [1–3]. However, the control performance of the sliding mode control (SMC) is another promising solution,
the two-stage bidirectional converter in electric vehicle charg- chattering needs to be eliminated [10–12], and the problem that
ing application is susceptible to the external disturbances the performance of the SMC would be deteriorated when con-
and the model parameter uncertainties, such as the input sidering various external disturbances needs to be solved [13].
voltage fluctuation, the load variation, the uncertain induc- In [14–16], the controller based on a super-twisting algorithm
tance and the capacitance deviation. Therefore, to tackle this (STA) is employed to avoid chattering. The second-order sliding
issue, improved control method should be developed at the mode control (SOSMC) can reduce the sensitivity to external
back-end buck/boost converter to achieve superior system disturbances. However, the model of the buck/boost converter
performance. shows that the disturbances exist in different state equations and

This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is
properly cited.
© 2021 The Authors. IET Power Electronics published by John Wiley & Sons Ltd on behalf of The Institution of Engineering and Technology

IET Power Electron. 2021;14:1421–1432. wileyonlinelibrary.com/iet-pel 1421

 17554543, 2021, 8, Downloaded from https://ietresearch.onlinelibrary.wiley.com/doi/10.1049/pel2.12120 by Algeria Hinari NPL, Wiley Online Library on [10/02/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
1422 SU ET AL.

Source/Load Bus
+ L +
S1 S3 S9
Charging S5 S7
pile D1C1 D3 C3
1:n D5 C5 D7 C7
Low-Voltage Bus
ULV Cr Lr * * UHV
C1 S10 C
High-Voltage Bus Lm
Power
AC DC DC Integration S2 S4 S6 S8
DC System of New D2C2 D4C4
DC DC - D6 C6 D8 C8 -
Energy Vehicle
AC
Source
FIGURE 2 A cascaded bidirectional topology with the combination of
Storage LLC resonant converter and buck/boost converter
System of
Battery

FIGURE 1 A general structure of the distributed DC system with power iL iload

electronic-based sources and loads + +
L
S9

is the mismatched disturbances. And the influence of the mis-

matched disturbances is not easy to be suppressed completely. ui C R uo
S10
Moreover, many voltage or current sensors need to be added,
leading to high system cost. Therefore, the disturbance observer
[17–20] and the high-order sliding mode observer (HOSMO)
[21] are adopted to reduce the sensor and achieve disturbance
- -
compensation. However, regarding these disturbance observers,
many studies have less consideration on the effect of the lumped
FIGURE 3 A buck/boost converter considering unknown disturbance of
disturbances and the situation under bidirectional power flow of power loads
the converters.
In light of the above, an SOSMC based on STA and HOS-
MDO is proposed in this paper to solve the problem of the rectifier of the power grid side. The input side of the
control performance degradation in the bidirectional DC-DC isolated DC-DC converter is connected to the low-voltage
converter caused by the external disturbances and model bus, and high-voltage bus is connected to the output of the
parameter uncertainties. The proposed STA-based SOSMC is isolated DC-DC converter. The rear-end bidirectional DC-DC
employed to alleviate chattering, which has no demand on the converter produces the source/load bus for the loads including
derivative of the sliding mode variable, simplifying the control charging pile, power integration system of new energy vehicle
structure. Meanwhile, an HOSMDO is designed to estimate the and storage system of battery.
disturbances term and the derivative term of the state variable. Figure 2 is a typical topology of the two DC-DC converters in
In that case, the system cost is reduced due to less current Figure 1. It is a cascaded bidirectional topology with the combi-
sensor, and the disturbances suppression is obtained with zero nation of LLC resonant converter and buck/boost converter.
steady-state tracking error. Moreover, considering the bidirec- The front-end LLC operating near the resonant frequency is
tional characteristics of the buck/boost topology, the proposed adopted to achieve isolation and high conversion efficiency. The
control method mainly aims at the same control target of the back-end buck/boost converter realizes wide range of output
output voltage. Therefore, the proposed control strategy can voltage regulation.
achieve a good dynamic response in bidirectional power flow The back-end stage shown in Figure 2 can be considered as
cases. Following, the analysis of the disturbances and parameter a buck/boost converter feeding a resistive load which consists
uncertainties and the system model are presented in Section 2. of the initial resistor R and unknown disturbance terms of the
In Section 3, the proposed SOSMC based on the STA and power loads. The current of the load is expressed by iload . The
the HOSMDO for the buck/boost converter are presented in equivalent and simplistic topology of the buck/boost converter
detail. Additionally, it also introduces stability analysis and the is described in Figure 3.
parameters selection method. The results of simulation and Firstly, the dynamic model of the buck/boost converter is
experiment are shown in Section 4. Finally, Section 5 gives the derived as follows:
conclusion.
d iL
L = d ui − uo
dt
2 PROBLEM DESCRIPTION d uo u
AND SYSTEM MODEL C = iL − o − iload (1)
dt R

Figure 1 shows a general structure of the distributed DC where C and L represent the capacitance and inductance values,
system. The low-voltage bus is connected to the output of respectively. ui is the input voltage, uo and iL are the output
 17554543, 2021, 8, Downloaded from https://ietresearch.onlinelibrary.wiley.com/doi/10.1049/pel2.12120 by Algeria Hinari NPL, Wiley Online Library on [10/02/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
SU ET AL. 1423

voltage and inductor current of the buck/boost converter, by the sign (̂∙). Then, the detailed design is presented as
respectively. d is the duty of the MOSFET, which is also the follows.
control law. iload is the uncertain term considered as the load
current which depends on the external load.
Considering the aforementioned, the disturbance terms usu- 3.1 Controller for buck/boost converter
ally comprise the external disturbance and the deviation of the
model parameters uncertainties. Then, the model parameters The first step in designing a sliding mode controller is to select
uncertainties such as inductance L, capacitance C and input the sliding surface. The sliding surface is designed according to
voltage ui , can be designed with bounded uncertainties as fol- the state equation of the controlled objective. Once the state
lows: reaches the sliding surface, it will reach a stable state based on
the exponential approach.
L = L0 + ΔL; C = C0 + ΔC ; ui = ui0 + Δui
(2)
|ΔL| ≤ 𝜀L ; |ΔC | ≤ 𝜀C ; |Δui | ≤ 𝜀ui
3.1.1 Sliding surface of SOSMC
where L0 , C0 and ui0 represent the nominal value of the induc-
tance, capacitance and input voltage, respectively. ∆L, ∆C and The control objective of the SOSMC for buck/boost converter
∆ui represent the deviation value of the inductance, capacitance is the output voltage uo . In order to track the voltage reference
and input voltage, respectively. 𝜀L , 𝜀C and 𝜀ui are known as pos- uo∗ with zero steady-state error and fast dynamic response, the
itive and bounded constants. error term for the output voltage and voltage reference should
Therefore, the dynamic model in Equation (1) is expressed be included in the sliding surface. Meantime, to increase the
as follows when the model parameter uncertainties are freedom of bandwidth regulation and extract the mismatched
considered. disturbances, the first-order derivative term of voltage error is
added as the state variable. Thus, the sliding surface is defined
d iL di by the following form,
L0 = d ui0 − uo+d Δui − ΔL L =d ui0 − uo+Δ𝜌1
dt dt
(3)
d uo uo d uo u s = 𝜆1 x1 + x2 (4)
C0 = iL − − i load − ΔC =iL − o − i load +Δ𝜌2
dt R dt R
where λ1 is a positive sliding coefficient, which affects the
di du
where Δ 𝜌1 = d Δui − ΔL L and Δ𝜌2 = −ΔC o are equiv- dynamic response of the operating system [22], and
dt dt
alent to the lumped uncertainties.
From Equation (3), it is found that the disturbance terms x1 = uo − uo∗
(5)
are the mismatched disturbances which are included in differ- x2 = ẋ 1 = u̇ o − u̇ o∗
ent formulas, such as ∆ρ1 and ∆ρ2 . If only some parts of the
disturbance terms are taken into account, the effect of the dis- Based on Equation (3), the first-order and second-order
turbances still exists. Therefore, the effects of the mismatched derivative of the control object uo can be obtained as the fol-
disturbances should not be neglected when designing the con- lowing form,
troller. Meanwhile, the build model should also contain buck
mode in the forward direction and boost mode in the reverse iL u i Δ𝜌2
u̇ o = − o − load + (6)
direction. The main control target of the proposed control C0 C0 R C0 C0
method is to mantain constant output voltage in both the for-
ward and reverse direction of power flow. Therefore, control (2) u x x u∗
strategies are designed and analysed as follows. uo = i0 d − 1 − 2 −
C0 L0 C0 L0 C0 R C0 L0
( ) (7)
i̇ u̇ ∗ Δ𝜌1 Δ𝜌̇ 2
+ − load − o + − = u1 + Δ𝜌
C0 C0 R C0 L0 C0
3 CONTROLLER AND DISTURBANCE
OBSERVER DESIGN where u1 is the input of the controller, which contains the con-
trol law d. Δ𝜌 is the disturbance term. It can be seen from Equa-
This section will give a detailed description of the proposed tion (7) that there are two relative degrees of the output voltage
method. Moreover, in order to clarify the problem, some uo to the control input u1 .
symbols of the function are represented as follow: the first- Therefore, the first-order derivation ṡ of the sliding surface
order derivative of variables are denoted by the sign (∙),
̇ the can be expressed as follows,
second-order derivative of variables are denoted by the sign
∗(2)
(∙)(2) , the reference value of variables are symbolized by the ṡ = 𝜆1 ẋ 1 + ẋ 2 = 𝜆1 x2 + u1 + Δ𝜌 − uo = 𝜆1 x2 + u1 + 𝜌
sign (∙∗ ), and the estimated value of variables are denoted (8)
 17554543, 2021, 8, Downloaded from https://ietresearch.onlinelibrary.wiley.com/doi/10.1049/pel2.12120 by Algeria Hinari NPL, Wiley Online Library on [10/02/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
1424 SU ET AL.

∗(2) The super-twisting control term udis is given by following,

where 𝜌 = Δ𝜌 − uo , 𝜌 is regarded as the total disturbance
terms of the system. 1
From Equations (3) and (7), it can be seen that all of the dis- udis = −a1 |s| 2 sign(s) + 𝜔 (14)
turbance factors are considered in 𝜌, such as ΔL, ΔC , Δui and 𝜔̇ = −a2 sign(s)
i̇load etc. Therefore, the influence of whole disturbance terms
can be considered in ṡ and the sliding surface s. In addition, where a1 and a2 are the designed positive constants.
it also shows that the system is susceptible to external distur- To better describe the design of the controller, there will be a
bances and model parameter uncertainties. stability analysis of the controller in the next part.

3.1.2 Design of controller-based STA 3.2 Stability analysis of the controller

From the above analysis, the expression of the dynamical sys- Assuming that each disturbance term can converge to a certain
tem is described as the following form which is consisted of the boundary value, the disturbance term 𝜌 is bounded in the fol-
disturbance term 𝜌, lowing form,

ẋ 1 = x2 |𝜌| ≤ 𝜀 (15)
ẋ 2 = u1 + 𝜌 (9)
y = x1 where ε is a known positive constant.
According to Equation (10), the disturbance term 𝜌 can
where y is the output of the dynamical system and u1 is the con- be extracted separately when the influence of disturbance is
trol law of the system, x1 and x2 are the state variables [23]. not considered in the state variable. Therefore, the first-order
Then, according to the definition, x1 = uo − uo∗ , x2 = derivative of the sliding surface s can be rewritten as follows,
u̇ o − u̇ o∗ , Equation (9) can be rewritten as follows,
ṡ = 𝜆1 ẋ 1 + ẋ 2 +𝜌 = 𝜆1 x2 + ẋ 2 +𝜌 (16)
i u i Δ𝜌2
ẋ 1 = x2 = L − o − load + − u̇ o∗
C0 C0 R C0 C0
u x x u∗ (10)
Combining Equations (11), (14) and (16), the generalized
ẋ 2 = u1 + 𝜌 = i0 d − 1 − 2 − o + 𝜌 expression of the closed-loop system is obtained as follows,
C0 L0 C0 L0 C0 R C0 L0
y = x1 [ ( ) ]
ui0 x1 x2 uo∗
ṡ = 𝜆1 x2 + ueq + udis − − − +𝜌
C0 L0 C0 L0 C0 R C0 L0
Furthermore, the STA is a control algorithm of rela- ui0
1

tive degree one system, which has the advantage to achieve = u + 𝜌 = −a3 |s| sign(s) + 𝜔 + 𝜌
2
C0 L0 dis
finite-time convergence and inhibit the smooth disturbance 𝜔̇ = −a4 sign(s)
of arbitrary circumstances [24, 25]. Considering two terms (17)
when we design the global control law u(t), the equivalent where a3 and a4 are the designed positive constants, and a3 =
ui0 u
control term ueq and the super-twisting control term udis . a1 , a4 = i0 a2 .
C0 L0 C0 L0
Thus, the total control law can be obtained by the following In order to prove the stability of the actual system, the Lya-
form, punov function can be chosen as following [26, 27].

u = ueq + udis (11) ( )2

1 1 1∕2
V = 2a4 |s| + 𝜔2 + a3 |s| sign (s ) − 𝜔 (18)
2 2
where the equivalent control term ueq can be calculated by = 𝜉 T P𝜉
solving the following expression,
1 4a + a3 −a3
2
1∕2
ṡ = 𝜆1 ẋ 1 + ẋ 2 = 𝜆1 x2 + ẋ 2 = 0 where 𝜉 T = [|s| sign(s) 𝜔 ], P = [ 4 ].
(12) 2 −a3 2
From Equation (18), Its first-order derivative is derived as fol-
Substituting Equation (10) into Equation (12), the ueq can be lows,
obtained as the following form. For the disturbance term 𝜌, its
estimated value 𝜌̂ can be obtained later by the HOSMDO. 1 T 𝜌
V̇ = − 𝜉 A𝜉 + BT 𝜉 (19)
1∕2 1∕2
[ ∗ ] |s| |s|
C0 L0 uo x 1
ueq = d = + 1 +( − 𝜆1 )x2 − 𝜌
ui0 C0 L0 C0 L0 C0 R a3 2a + a32 −a3 T a32 −a3
where A = [ 4 ], B = [ (2a4 + ) ].
(13) 2 −a3 1 2 2
 17554543, 2021, 8, Downloaded from https://ietresearch.onlinelibrary.wiley.com/doi/10.1049/pel2.12120 by Algeria Hinari NPL, Wiley Online Library on [10/02/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
SU ET AL. 1425

Referring to the elaboration in [26, 27] and combining the

boundary conditions of disturbance terms in Equation (15), fol-
lowing inequality can be obtained,

1 T ̃
V̇ ≤ − 𝜉 A𝜉 (20)
1∕2
|s|

4a
2a + a32 −( 4 + a3 )𝜀 −(a3 + 2𝜀)
̃ =
where A
a3
[ 4 a3 ]
2 −(a3 + 2𝜀) 1
From what has been discussed above, V̇ < 0 means the sys- FIGURE 4 The control block diagram of the buck/boost converter with
̃ > 0. In this case, the gains
tem will tend to be globally stable if A the proposed control strategy
of the STA need to be satisfied as follows,
Defining the estimation error e1 = x1 − x̂ 1 , e2 = x2 − x̂ 2 ,
a3 > 2𝜀
5a3 𝜀+4𝜀2 (21) and e3 = 𝜌 − x̂ 3 . Then, the correction terms are defined as
a4 > a3 follows,
2(a3 −2𝜀 )

2
Inequality in Equation (21) is the condition for the global sta- z1 = k1 |e1 | 3 sign(e1 )
bility of the system and provides a reference for the design of 1
(23)
parameters a1 and a2 . z2 = k2 |e1 | 3 sign(e1 )
z3 = k3 sign(e1 )

3.3 Design of HOSMDO where k1 , k2 , and k3 are the positive coefficients. The selection
of these parameters mentioned above will be illustrated in Sec-
From the above derivation, the disturbance terms exist in the tion 3.4.
sliding surface s, the derivation ṡ of the sliding surface and Then, the expression of error dynamics can be obtained as
the control law u. The disturbance terms can be observed by follows,
the sliding mode observer (SMO). However, from Equation (3),
2
it is also known that the disturbance terms assigned in different ė 1 = −k1 |e1 | 3 sign(e1 ) + e2
state equations. It is a mismatched disturbance. All disturbance 1
(24)
terms need to be considered by applying the second-order ė 2 = −k2 |e1 | 3 sign(e1 ) + e3
derivative of the output voltage. Meantime, to reduce the ė 3 = −k3 sign(e1 ) + 𝜌̇
current sensor, the inductance current iL needs to be observed.
Thus, the proposed HOSMDO method is designed to estimate Assuming that the disturbance term 𝜌 is a Lipschitz constant
the value of state variables. It requires the disturbance term 𝜌 and |𝜌| ≤ 𝜀. Now, assuming that |𝜌| ̇ ≤ 𝛿, where 𝛿 is a known
as a state variable and accurately observes the 𝜌. Therefore, it positive constant. Then, the stability of Equation (24) in finite-
makes compensation for the disturbance terms of the system time is already proved in [29] and [30].
and eliminates the influence of the disturbance on the system. Therefore, the HOSMDO has the following form,
In addition, it also estimates the state variable x2 . There-
2
fore, the sampling of the inductive current can be eliminated x̂̇ 1 = x̂ 2 + k1 |e1 | 3 sign(e1 )
because the inductive current is contained in the state variable x̂ u∗
1
x̂̇ 2 =
ui x1 (25)
x2 . u− − 2 − o + x̂ 3 + k2 |e1 | 3 sign(e1 )
CL CL CR CL
To suppress the influence of the disturbance terms and x̂̇ 3 = k3 sign(e1 )
reduce the cost, an HOSMDO is designed to achieve the desired
performance. Referring to Equation (9), the dynamic expres- It is concluded that the error terms e1 , e2 and e3 tend to zero in
sion of the HOSMDO containing observed errors is given as finite-time by choosing the appropriate correction coefficients
follows: k1 , k2 , and k3 , and the observed state variables can be used
to replace the actual state variables, namely, x1 = x̂ 1 , x2 = x̂ 2 ,
x̂̇ 1 = x̂ 2 + z1 and 𝜌 = 𝜌̂ = x̂ 3 [31].
Therefore, the control block diagram of the buck/boost con-
x̂̇ 2 = x̂ 3 + u + z2 (22) verter with the proposed control strategy is shown in Figure 4.
x̂̇ 3 = z3
3.4 The selection of parameters
where z1 , z2 , and z3 are the correction terms. x̂ 3 is regarded as
an observation value of disturbance term whose initial value is The six designed parameters in the actual system are λ1 , a1 ,
simplified to zero [28]. a2 , k1 , k2 and k3 , respectively, which are related to SOSMC
 17554543, 2021, 8, Downloaded from https://ietresearch.onlinelibrary.wiley.com/doi/10.1049/pel2.12120 by Algeria Hinari NPL, Wiley Online Library on [10/02/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
1426 SU ET AL.

Oscilloscope Processor

Buck/Boost

Programmable
DC source

Resistive Bidirectional programmable

load DC source

FIGURE 5 Experimental platform

and HOSMDO. They play a significant role in steady-state and MDO are obtained as follows,
dynamic performance, as well as the ability to against the distur-
bance. k1 = −𝜆̃ 2 Ld 2∕3
According to the sliding surface s, the output voltage error k2 = −𝜆̃ 1 Ld 1∕3 (27)
can be expressed as follows when s = 0 and ṡ = 0: k3 = −𝜆̃ 0 Ld

x1 (t ) = x1 (0)e−𝜆t (26) where Ld is the known Lipschitz constant, and Ld > 0. 𝜆̃ 2 , 𝜆̃ 1

and 𝜆̃ 0 are the adjustable coefficients. An appropriate selec-
Meantime, referring to Equation (26), it can be concluded tion of the parameters Ld , 𝜆̃ 2 , 𝜆̃ 1 , and 𝜆̃ 0 is mentioned in [21].
that λ1 utilized in the sliding surface function denotes the pole From literature [21], due to the HOSMDO has a third-order
of sliding mode dynamic. It influences the dynamic response equation, the coefficients can be selected as follows: 𝜆̃ 0 = 1.1,
of the closed-loop system, so the parameter λ1 should be 𝜆̃ 1 = 2.12, 𝜆̃ 2 = 2.
selected as large as possible. However, when extremely large Then, it can be learned from [21] that the parameter value
λ1 is designed, in spite of the representative point can faster to Ld significantly influences the response and disturbance rejec-
reach, bigger overshoot and unstable state may occur. Besides, a tion of the system. When the parameter Ld gradually increases,
small λ1 will deteriorate the tracking effect. the dynamic response will accelerate, but if it is too large, the
As the gain coefficients of the STA term, the system can be performance of the disturbance rejection will be deteriorated.
considered to keep stability and the error converges to zero in
finite-time by adjusting the parameters a1 and a2 . They influence
the proportion of the STA term udis in the control law. Further- 4 SIMULATION AND EXPERIMENTAL
more, the gain parameters a1 and a2 can be selected based on RESULTS
the bounded disturbance, refer to Equations (17) and (21).
Considering that the observable value of voltage û o is com- To verify the effectiveness of the proposed method, a 2.1-kW
prised of a bounded Lebesgue-measurable noise with unknown experimental platform of the buck/boost converter is devel-
features, whose third-order derivative has a known Lipschitz oped in the laboratory as shown in Figure 5. The switches used
constant. To get the ̂u̇o and robust estimations in finite-time, a in the topology are MOSFET and the digital signal processor
sliding mode differentiator (SMD) is mentioned as the HOS- is DSP (TMS320F28069) from TI. A bidirectional DC source
MDO [28, 31, 32]. According to the non-recursive form of the and programmable DC source are adopted to imitate the energy
differentiator [32], the parameter gains k1 , k2 and k3 of HOS- feedback-end and DC bus side. The combination of resistors is
 17554543, 2021, 8, Downloaded from https://ietresearch.onlinelibrary.wiley.com/doi/10.1049/pel2.12120 by Algeria Hinari NPL, Wiley Online Library on [10/02/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
SU ET AL. 1427

TABLE 1 Experimental parameters TABLE 2 Results of input voltage variation

Parameters Description Values Parameters Double-loop PI SOSMC+HOSMDO

ui DC-link voltage 400 V Overshoot (uo ) 0.6 V 0.12 V

Po Output power 2.1 kW Overshoot (iload ) 0.0145 A 0.003 A
fsw Switching frequency 20 kHz Settling time (uo ) 3.8 ms 2.5 ms
fs Sampling frequency 20 kHz Settling time (iload ) 3.8 ms 2.5 ms
L0 Filter inductor 1.5 mH
C0 Filter capacitor 880 uF
R Initial resistance 128 Ω
4.2 Test of input voltage variation
Rx Full load resistance 43 Ω
S9 -S10 MOSFET switch tube FCH072N60F For the conventional methods, PI control is widely used in the
𝜆1 SOSMC parameter 1 600 actual engineering application due to its merits of simplicity and
a1 SOSMC parameter 2 0.01 effectiveness. Therefore, to further demonstrate the superior
performance of the proposed control strategy, a double-loop
a2 SOSMC parameter 3 100
PI controller is applied to make a comparison. The parameters
Ld HOSMDO parameter 1 1 × 1010
for the double-loop PI controller are well designed to achieve a
k1 HOSMDO parameter 2 2 × (1 × 1010 )1/3 dynamic effect. The voltage-loop bandwidth is 100 Hz, and the
k2 HOSMDO parameter 3 2.12 × (1 × 1010 )2/3 current-loop bandwidth is 1000 Hz. Moreover, the current-loop
k3 HOSMDO parameter 4 1.1 × (1 × 1010 ) considers the actual voltage as a feedforward term to speed up
the response.
Figure 7 shows the output voltage and output current wave-
forms of the system when the input voltage changes from 400
to 200 V and then back to 400 V. In this case, the reference value
of the output voltage is 100 V and the load resistance is 43 Ω. It
can be seen from Figure 7 that the change rate of the input volt-
age is 1000 V/S when the external programmable DC power
supplies the input voltage. And there is no obvious overshoot
of the output voltage and output current. Therefore, Figure 8
compares the effect of the proposed method and double-loop
PI method in the case of instantaneous large-range variation of
input voltage through simulation.
From the simulation results in Figure 8, the input voltage
is instantaneous changes from 400 to 200 V and then back to
400 V. It can be clearly seen that the overshoot of the out-
put voltage and output current are both significantly improved
FIGURE 6 Steady-state performance of the system
when the input voltage is change suddenly. From the experimen-
tal results in Table 2, the overshoot is only about one fifth of the
double-loop PI method. Compared with the results obtained
by double-loop PI method, the method proposed in this paper
simulated as the load. The experimental parameters for the sys- adjusts the voltage and current value quickly within settling time
tem are listed in Table 1. less than 2.5 ms.

4.1 Steady-state performance of the system 4.3 Disturbance rejection of load variation
First of all, to verify the steady-state performance of the pro- Then, in order to evaluate the performance of disturbance rejec-
posed method, the steady-state waveforms of input voltage ui , tion under load variation, the resistive load can be increased or
output voltage uo , inductor current iL in the 400 ms scale are decreased from the initial load. Figure 9 shows the dynamic per-
shown in Figure 6. In this case, the input voltage is 400 V, the formance of the system with the load resistor varies from 128
output voltage is 300 V, and the load resistance is 43 Ω. From to 43 Ω (power from 700 to 2100 W) and from 43 to 128 Ω
Figure 6, it can be seen that the output voltage is very stable (power from 2100 to 700 W), respectively. At the front of each
under the application of the proposed method in this paper. The graphic is the zoom in on the details of each case.
enlarged figure shows the steady-state precision of output volt- It can be seen from Figure 9(a,c) that the overshoot of the
age is about ±1.5 V. It can be calculated that the error of the output voltage by the double-loop PI method is about 7 V and
output voltage is within 1%. the settling time is about 16 ms when the load resistor is changed
 17554543, 2021, 8, Downloaded from https://ietresearch.onlinelibrary.wiley.com/doi/10.1049/pel2.12120 by Algeria Hinari NPL, Wiley Online Library on [10/02/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
1428 SU ET AL.

400

350

ui(V)
300

250

200

0.03 0.035 0.04 0.045 0.05 0.055 0.06 0.065 0.07

101
100.5

uo(V)
100

99.5
99

0.03 0.035 0.04 0.045 0.05 0.055 0.06 0.065 0.07

2.36

iload(A)
2.35

2.34

2.33

2.32
0.03 0.035 0.04 0.045 0.05 0.055 0.06 0.065 0.07

Time(s)

a
400

350
ui(V) 300

250

200

0.03 0.035 0.04 0.045 0.05 0.055 0.06 0.065 0.07

101

100.5
uo(V)

100

99.5

99

0.03 0.035 0.04 0.045 0.05 0.055 0.06 0.065 0.07

2.36
iload(A)

2.35

2.34

2.33

2.32

0.03 0.035 0.04 0.045 0.05 0.055 0.06 0.065 0.07

Time(s)

b
FIGURE 8 Simulation results of the dynamic performance when the
input voltage varies instantaneously. (a) Double-loop PI, (b)
SOSMC+HOSMDO

from 128 to 43 Ω. In the proposed method, the overshoot of

the output voltage is 4 V and the settling time is about 7 ms.
As shown in Figure 9(b,d), when the load resistor changes
from 43 to 128 Ω, the overshoot of the output voltage by the
double-loop PI method is about 7 V and the settling time is
about 17 ms. For the proposed method, the overshoot of the
output voltage is 2 V, and the settling time is about 5 ms. The
FIGURE 7 The dynamic performance of the system when the input experimental results are listed in Table 3. Therefore, it can be
voltage varies by 1000 V/S rate. (a) From 400 to 200 V (Double-loop PI), (b)
concluded that the proposed method in this paper has better
from 200 back to 400 V (Double-loop PI), (c) from 400 to 200 V
(SOSMC+HOSMDO), (d) from 200 back to 400 V (SOSMC+HOSMDO) load disturbance rejection and fast dynamic performance.
 17554543, 2021, 8, Downloaded from https://ietresearch.onlinelibrary.wiley.com/doi/10.1049/pel2.12120 by Algeria Hinari NPL, Wiley Online Library on [10/02/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
SU ET AL. 1429

TABLE 3 Results of load variation

Parameters Double-loop PI SOSMC+HOSMDO

Load resistor changes from 128 to 43 Ω

Overshoot 7V 4V
Settling time 16 ms 7 ms
Load resistor changes from 43 to 128 Ω
Overshoot 7V 2V
Settling time 17 ms 5 ms

TABLE 4 Results of output voltage regulation

Parameters Double-loop PI SOSMC+HOSMDO

Output voltage changes from 100 to 300 V

Overshoot 5.5 V 2.5 V
Settling time 17 ms 6 ms
Output voltage changes from 300 to 100 V
Overshoot 6V 2V
Settling time 16 ms 5 ms

4.4 Effect of output voltage regulation

Figure 10 shows the effect of dynamic adjustment when the

output voltage is subject to wide-range regulation. Meantime,
to better exhibit the superiority of the proposed method, some
enlarged dynamic details are present at the front of each graph
when different control methods are used.
At this point, the output voltage is adjusted from 100 to
300 V and then back to 100 V. The varied step of the output
voltage is 5000 V/S. It can be seen from Figure 10(a,c) that the
effect of voltage regulation by the proposed method is obviously
improved than the double-loop PI method. The overshoot of
double-loop PI is about 5.5 V in output voltage and the settling
time is 17 ms. As a comparison, the voltage overshoot of the
proposed method is only 2.5 V and the settling time is about
6 ms. Furthermore, Figure 10(b,d) presents the dynamic perfor-
mance with the output voltage adjust to 100 V. It can be seen
that the overshoot of output voltage is about 6 V by double-
loop PI method and the settling time is 16 ms. The proposed
method has about 2 V overshoot in output voltage and tracks
the reference need about 5 ms settling time. The experimental
results are listed in Table 4. Thus, it can be concluded that the
SOSMC based on STA and HOSMDO method mentioned in
this paper, compared with the double-loop PI method, achieves
more excellent performance when the reference output voltage
changes.

FIGURE 9 The dynamic performance of the system when the load 4.5 Robustness of the control method to
variation. (a) From 128 to 43 Ω (Double-loop PI), (b) from 43 back to 128 Ω model parameter uncertainties
(Double-loop PI), (c) from 128 to 43 Ω (SOSMC+HOSMDO), (d) from 43
back to 128 Ω (SOSMC+HOSMDO)
Considering that the actual inductance and capacitance param-
eters may have tiny variations along with the running time
 17554543, 2021, 8, Downloaded from https://ietresearch.onlinelibrary.wiley.com/doi/10.1049/pel2.12120 by Algeria Hinari NPL, Wiley Online Library on [10/02/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
1430 SU ET AL.

306
ΔL=0
304
ΔL=-20%
ΔL=20%
302

uo(V)
300

298

296

294
0.015 0.02 0.025 0.03 0.035 0.04 0.045
8
ΔL=0
7
ΔL=-20%
6 ΔL=20%

iload(A)
5

1
0.015 0.02 0.025 0.03 0.035 0.04 0.045

Time(s)
a
306
ΔC=0
304 ΔC=-40%
ΔC=40%
302
uo(V)
300

298

296

294
0.015 0.02 0.025 0.03 0.035 0.04 0.045
8

7 ΔC=0
ΔC=-40%
6
ΔC=40%
iload(A)

1
0.015 0.02 0.025 0.03 0.035 0.04 0.045

Time(s)

FIGURE 11 Simulation results of disturbance rejection with uncertain

inductance and capacitance parameters when the load step variation. (a)
Different deviation values of inductance △L at 0, −20%, 20%, (b) different
deviation values of capacitance △C at 0, −40%, 40%

or environmental factors and the design of the parameters

and the measured value of components may exist a certain
small deviation, the system will suffer from the disturbance of
model uncertainties. Therefore, to demonstrate the robustness
of the proposed method against parameter uncertainties, the
simulation verification is carried out. In this paper, the values
of inductance and capacitance are respectively increased and
decreased to different degrees, so as to make a comparison.
FIGURE 10 The dynamic performance of the system when the output Figure 11(a) shows the dynamic performance of output voltage
voltage regulation. (a) From 100 to 300 V (Double-loop PI), (b) from 300 to
and output current corresponding to different inductance val-
100 V (Double-loop PI), (c) from 100 to 300 V (SOSMC+HOSMDO), (d)
from 300 to 100 V (SOSMC+HOSMDO) ues when the load resistor varies from 128 to 43 Ω and back to
128 Ω. Figure 11(b) shows the dynamic performance of output
 17554543, 2021, 8, Downloaded from https://ietresearch.onlinelibrary.wiley.com/doi/10.1049/pel2.12120 by Algeria Hinari NPL, Wiley Online Library on [10/02/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
SU ET AL. 1431

0.9

0.8

u
0.7

0.6

2000

s
0

-2000

6
4
2
0

x
-2
-4

4000
2000
0

x
-2000
-4000
0.046 0.048 0.05 0.052 0.054 0.056 0.046 0.048 0.05 0.052 0.054 0.056
Time(s)

FIGURE 13 The dynamic waveforms of control law u, sliding surface s

and sliding variables x1 , x2 for the case of forward and reverse running.
Left-side: From 2100 W in the forward direction to −2100 W in the reverse
direction. Right-side: From −2100 W in the reverse direction to 2100 W in the
forward direction

The left-side waveform of Figure 13 shows the effect of

FIGURE 12 The dynamic performance of system for forward and changing from forward 2100 W to reverse −2100 W, and the
reverse running. (a) From 2100 W in the forward direction to −2100 W in the right-side waveform of Figure 13 shows the effect of changing
reverse direction, (b) from −2100 W in the reverse direction to 2100 W in the
forward direction
from reverse −2100 W to forward 2100 W. From Figure 13, it
can be seen that the overshoot of state variable x1 is approxi-
mately 2.5 V during the direction changes. It clearly shows that
voltage and output current corresponding to different capac- the system has a short tracking time with less than 3 ms. Mean-
itance values when the load resistor varies from 128 to 43 Ω time, the simulation results also show that the control law u is
and back to 128 Ω. The deviation value of inductance is ±20%, near 0.75 and the state variable x1 , x2 are nearly zero when the
and the deviation value of capacitance is ±40%. It can be seen system is operating in steady-state. It can be concluded that the
from Figure 11 that the method mentioned in this paper can proposed method achieves stability and superior robustness of
still eliminate the influence of model parameter uncertainties. It the system.
maintains stable output of the system and quickly adjusts back
to the steady-state when the load changes suddenly.
5 CONCLUSION

4.6 Test of bidirectional characteristics The disturbance rejection performance plays an important role
in the buck/boost converter with wide range of output voltage
Figure 12 shows the dynamic performance of the system when regulation. This paper proposes an SOSMC based on STA and
the buck/boost converter running in different directions. Chan- the HOSMDO to improve the steady-state and dynamic perfor-
nels 1 to 4 corresponds to gating signal of the MOSFET switch mance of the system. Aiming at external disturbances and model
S9 , output voltage, output current, and gating signal of the parameter uncertainties, this method considers the lumped dis-
MOSFET switch S10 . As can be seen, Figure 12(a) shows the turbance terms as the observable state variables. Moreover, the
change of direction from forward 2100 W to reverse −2100 W, derivative of the error term between the output voltage and its
and Figure 12(b) shows the change of direction from reverse reference is also observed. It eliminates the disturbances by the
−2100 W to forward 2100 W. From the waveforms, it can be observation value of disturbance terms and reduces the induc-
seen that the output voltage keeps stable, and the direction of tor current sensor. The results show that the proposed method
the output current can achieve natural and smooth change. has good suppression effect on the input voltage fluctuation and
To exhibit the superior performance of the proposed load variation, superior steady-state and dynamic performance,
method, the simulation results of control law u, sliding surface s as well as robustness against the uncertainties of inductance and
and sliding variable x1 , x2 are presented. capacitance parameters.
 17554543, 2021, 8, Downloaded from https://ietresearch.onlinelibrary.wiley.com/doi/10.1049/pel2.12120 by Algeria Hinari NPL, Wiley Online Library on [10/02/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
1432 SU ET AL.

ACKNOWLEDGMENTS 16. Liu, J., et al.: Extended state observer-based sliding-mode control for
The authors acknowledge the support for this work pro- three-phase power converters. IEEE Trans. Ind. Electron. 64(1), 22–31
(2017)
vided by National Key R&D Program of China under Grant
17. Cui, H., Yang, J., Li, S.: Nonlinear disturbance rejection control for a buck-
2018YFB0606005, National Natural Science Foundation of boost converter with load uncertainties. In: Proceedings of the 33rd Chi-
China under Grant 61933011 and 61873289, Hunan Provin- nese Control Conference. Nanjing, pp. 3788–3793 (2014)
cial Key Laboratory of Power Electronics Equipment and Grid 18. Wang, J., et al.: Finite-time disturbance observer based non-singular termi-
under Grant 2018TP1001, Project of Innovation-driven Plan nal sliding-mode control for pulse width modulation based DC–DC buck
converters with mismatched load disturbances. IET Power Electron. 9(9),
in Central South University under Grant 2019CX003, Major
1995–2002 (2016)
Project of Changzhutan Self-dependent Innovation Demon- 19. Wang, J., et al.: Extended state observer-based sliding mode control for
stration Area under Grant 2018XK2002 and Provincial Key PWM-based DC–DC buck power converter systems with mismatched dis-
R&D Program of Hunan under Grant 2019GK2211. turbances. IET Control Theory Appl. 9(4), 579–586 (2015)
20. Yin, Y., et al.: Disturbance observer based second order sliding mode con-
trol for DC-DC buck converters. In: IECON 2017 - 43rd Annual Confer-
REFERENCES
ence of the IEEE Industrial Electronics Society. Beijing, pp. 7117–7122
1. Gorji, S.A., et al.: Topologies and control schemes of bidirectional DC–DC (2017)
power converters: An overview. IEEE Access 7, 117997–118019 (2019) 21. Xu, Q., et al.: An offset-free composite model predictive control strategy
2. Jiang, T., et al.: Bidirectional LLC resonant converter for energy storage for DC/DC buck converter feeding constant power loads. IEEE Trans.
applications. In: 2013 Twenty-Eighth Annual IEEE Applied Power Elec- Power Electron. 1(1), 1–11 (2019)
tronics Conference and Exposition (APEC). Long Beach, CA, pp. 1145– 22. Komurcugil, H.: Rotating-sliding-line-based sliding -mode control for
1151 (2013) single-phase UPS inverters. IEEE Trans. Ind. Electron. 59(10), 3719–3726
3. Zeng, Z., et al.: Research of bi-directional DC/DC converter topology (2012)
based on supercapacitor energy storage system in IP transmitter. J. Eng. 23. Davila, J., Fridman, L., Levant, A.: Second-order sliding-mode observer
2019(16), 1962–1967 (2018) for mechanical systems. IEEE Trans. Autom. Control 50(11), 1785–1789
4. Daware, C.V., Sabnis, A.: Performance measurement of linear and non lin- (2005)
ear controllers for buck-boost converter in solar energy system. In: 2018 24. Seeber, R., Horn, M.: Stability proof for a well-established super-twisting
International Conference on Computation of Power, Energy, Information parameter setting. Automatica 84, 241–243 (2017)
and Communication (ICCPEIC). Chennai, pp. 225–229 (2018) 25. Zhao, Z., et al.: Terminal sliding mode control based on super-twisting
5. Linares-Flores, J., et al.: Robust passivity-based control of a buck–boost- algorithm. J. Syst. Eng. Electron. 28(1), 145–150 (2017)
converter/DC-motor system: An active disturbance rejection approach. 26. Moreno, J.A., Osorio, M.: Strict Lyapunov functions for the super-
IEEE Trans. Ind. Appl. 48(6), 2362–2371 (2012) twisting algorithm. IEEE Trans. Autom. Control 57(4), 1035–1040
6. Wu, K.C.: A comprehensive analysis of current-mode control for DCM (2012)
buck-boost converters. IEEE Trans. Ind. Electron. 51(3), 733–735 (2004) 27. Moreno, J.A., Osorio, M.: A Lyapunov approach to second-order sliding
7. Salimi, M., et al.: Indirect output voltage regulation of DC-DC buck/boost mode controllers and observers. In: 2008 47th IEEE Conference on Deci-
converter operating in continuous and discontinuous conduction modes sion and Control. Cancun, pp. 2856–2861 (2008)
using adaptive backstepping approach. IET Power Electron. 6(4), 732–741 28. Chalanga, A., Kamal, S.: Implementation of super-twisting control super-
(2013) twisting and higher order sliding mode observer based approaches. IEEE
8. Duong, M.Q., et al.: Modeling and performance evaluation of a fuzzy logic Trans. Ind. Electron. 63(6), 3677–3685 (2016)
controller for buck-boost DC/DC converters. In: 2018 IEEE Interna- 29. Moreno, J.A.: Lyapunov function for Levant’s second order differentiator.
tional Conference on Fuzzy Systems (FUZZ-IEEE). Rio de Janeiro, pp. 1– In: Proceedings of the 51st IEEE Conference on Decision and Control.
7 (2018) Maui, HI, pp. 6448–6453 (2012)
9. Zhang, X., et al.: A model predictive current controlled bidirectional three- 30. Angulo, M.T., Moreno, J.A., Fridman, L.: Robust exact uniformly con-
level DC/DC converter for hybrid energy storage system in DC micro- vergent arbitrary order differentiator. Automatica 49(8), 2489–2495
grids. IEEE Trans. Power. Electron. 34(5), 4025–4030 (2019) (2013)
10. Pan, Y., et al.: Integral sliding mode control: Performance, modification, 31. Levant, A.: Higher-order sliding modes, differentiation and output-
and improvement. IEEE Trans. Ind. Inf. 14(7), 3087–3096 (2018) feedback control. Int. J. Control 76(9/10), 924–941 (2003)
11. Ling, R., et al.: Second-order sliding-mode controlled three-level buck DC– 32. Shtessel, Y., Fridman, C., Levant, A.: Sliding Mode Control and Observa-
DC converters. IEEE Trans. Ind. Electron. 65(1), 898–906 (2018) tion. 1st ed., Springer Science & Business Media, New York (2014)
12. Guo, B., et al.: A robust second-order sliding mode control for single-
phase photovoltaic grid-connected voltage source inverter. IEEE Access
7, 53202–53212 (2019)
13. Zhang, J., et al.: Discrete-time sliding mode control with disturbance rejec-
tion. IEEE Trans. Ind. Electron. 66(10), 7967–7975 (2019)
14. Li, Q., et al.: Super-twisting algorithm based on fast terminal sliding sur- How to cite this article: Su, M., et al.: Disturbance
face for buck converter in fuel cell electric vehicle. In: 2018 IEEE Trans- observer-based sliding mode control for dynamic
portation Electrification Conference and Expo (ITEC). Long Beach, CA, performance enhancement and current-sensorless of
pp. 69–74 (2018)
buck/boost converter. IET Power Electron. 14,
15. Zhang, W., et al.: A disturbance rejection control strategy for droop-
controlled inverter based on super-twisting algorithm. IEEE Access 7, 1421–1432 (2021). https://doi.org/10.1049/pel2.12120
27037–27046 (2019)