ISA Transactions xxx (xxxx) xxx

Contents lists available at ScienceDirect

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

Research article

An Adaptive–Predictive control scheme with dynamic Hysteresis

Modulation applied to a DC–DC buck converter
∗
Nubia Ilia Ponce de León Puig a , , Dimitar Bozalakov b , Leonardo Acho c ,
Lieven Vandevelde b , José Rodellar a
a
Universitat Politécnica de Catalunya-UPC, BarcelonaTech. Escola d’Enginyeria de Barcelona Est-EEBE, Department of Mathematics, Spain
b
Ghent University, Faculty of Engineering and Architecture, Department of Electromechanical Systems and Metal Engineering, Belgium
c
Universitat Politécnica de Catalunya-UPC, Escola Superior d’Enginyeries Industrial, Aeroespacial i Audiovisual de Terrassa ESEIAAT, Spain

article info a b s t r a c t

Article history: This article proposes a recent Adaptive–Predictive (AP) control technique applied to a DC–DC buck
Received 13 January 2020 converter. This converter topology has a wide range of applications in the current electronic and
Received in revised form 27 April 2020 electrical systems that demand an efficient use of low bus voltage and specific requirements in load
Accepted 9 May 2020
current consumption. Nevertheless, this converter, and in general any DC–DC converter topology,
Available online xxxx
presents a control challenge due to its nonlinear nature. Hence, in this article, it is proposed an
Keywords: adaptive–predictive control scheme that has low implementation complexity and improves the buck
DC–DC buck converter converter performance since it provides a fast response of the output voltage. Moreover, the output
Adaptive–predictive control is adequately regulated even when the system is subjected to perturbations in the reference voltage,
Hysteresis modulation in the input voltage, in the load or in the converter parameters that may be seen as faults in the
system. On the other hand, one of the main contributions of the proposed control technique with
respect to other controllers is that the AP control scheme allows to on-line infer the parametric status
of the plant thanks to its adaptive stage. In addition, a dynamic Hysteresis Modulator (HM) is properly
inserted in the control strategy to improve the dynamic behavior of the Adaptive Mechanism (AM), and
in general, of the entire closed-loop control performance. To validate the effectiveness of the control
design, a wide range of numerical experiments are carried out by using Matlab/Simulink. Finally, the
developed control technique was implemented in a benchmark experimental platform. According to
the experimental results, the proposed predictive control is suitable for real scenarios in the power
electronics applications.
© 2020 ISA. Published by Elsevier Ltd. All rights reserved.

1. Introduction the application. For instance, in photovoltaic (PV) systems, the

DC–DC converter output voltage may be subjected to external
In recent years, the power electronics field has had an enor- perturbations due to the variations of the solar irradiation on
mous growth, mainly in the branch of regulation, conversion the PV panel. These variations may be rapid, for instance, under
and distribution of energy [1,2]. This, due to the high demand fast shading conditions [7,8]. Hence, the controller of the DC–DC
of electrical and electronic applications that require devices that converters applied to a PV system must be robust to changes
realize these tasks. Among all the variety of power electronics
in the input voltage and external perturbations [9]. In general,
devices, DC–DC converters are one of the most common and most
the main objective of DC–DC converters is to ensure stability
studied by engineers and researchers [1].
with an adequate dynamic response in order to achieve a desired
Regarding to the control engineering, the DC–DC converters
constitute an important challenge due to their switched nonlinear load output voltage by guaranteeing a good performance while
and time-varying characteristics [2–4], the fast changes in their optimizing the utility life of the electronic components [3]. The
reference voltage, their high sensitivity to the frequently chang- most common DC–DC converters are the buck converter, which is
ing loading conditions [5], their very small sampling period [6] the one treated in this paper and basically consists in decreasing
and the changes in the system parameters related to external the output voltage on the load with respect to the input voltage,
perturbations. All of these challenging tasks highly depend on the boost converter that works to increase the voltage with re-
spect to its input, and finally the buck–boost also known as Cûk
∗ Corresponding author. converter, which realizes both tasks, decreases and increases the
E-mail address: [email protected] (N.I.P. de León Puig). load voltage according to the control switch position [10].

https://doi.org/10.1016/j.isatra.2020.05.015
0019-0578/© 2020 ISA. Published by Elsevier Ltd. All rights reserved.

Please cite this article as: N.I.P. de León Puig, D. Bozalakov, L. Acho et al., An Adaptive–Predictive control scheme with dynamic Hysteresis Modulation applied to a DC–DC
buck converter. ISA Transactions (2020), https://doi.org/10.1016/j.isatra.2020.05.015.
2 N.I.P. de León Puig, D. Bozalakov, L. Acho et al. / ISA Transactions xxx (xxxx) xxx

Nowadays, and because of the buck converter advantages, of the adaptive techniques in the DC–DC converters state-of-art.
such as its small size, low weight, and high efficiency [2,11,12], Finally, in the area of power converters, some Predictive Control
it is universally used for a great amount of applications that techniques have emerged as a good alternative [6,41]. The Model
require low bus voltage and low-to-medium load current con- Predictive Control (MPC) obtains the control action by solving
sumption [7,13,14]. This converter has been employed in the real- an optimization problem with future prediction over a finite
ization of battery chargers [15,16], and battery-operated portable horizon [42]. The main advantage of this technique is that the
equipment, due to its simple structure and low-cost [17,18]. In system constraints and nonlinearities may be considered in the
the automotive field, the buck converter is widely used. One cost function [6,43]. Nevertheless, the MPC presents disadvan-
example of it is its bidirectional version for applications in dual tages, such as, the difficulty of the cost function design or the lack
battery system for hybrid electric vehicles [19,20]. Furthermore, of stability guarantees [42,43]. The Predictive Control technique
in the management of energy, the DC–DC converters have been has been successfully used in photovoltaic applications where a
adopted along with an adequate control technique, to improve Maximum Power Point Tracking is combined jointly to the DC–
the optimization of the total cost of fuel cell/battery in hybrid DC converter controller to improve the efficiency of the energy
electric vehicles [21,22]. Last but not least, in the renewable provided by the photovoltaic panel [44,45].
energies area, the buck converter also has won popularity, for Since there are still many disadvantages to be defeated in the
instance, to feed power from distributed generators into smart DC–DC converter controllers, this article proposes a recent control
grids [9] and to improve the efficiency of energy provided by pho- scheme for the buck converter based on an Adaptive– Predictive
tovoltaic panels through maximum power point tracking tech- (AP) strategy. The AP control scheme is a well established tech-
niques [23–25]. nique that has been applied in a wide variety of industrial appli-
In the last two decades, a considerable amount of control cations [46–48]. In this article, a remodeled Adaptive–Predictive
techniques have been applied to these devices, of which mainly control is proposed to improve the DC–DC buck converter perfor-
focus on the Sliding Mode Control (SMC) technique [26,27], which mance. Generally speaking, the AP control strategy requires the
is a notable control strategy that has been applied in a wide range mathematical model of the process to predict its future behavior
of engineering systems [28]. This strategy has received much that is re-planed every sample time. Moreover, the AP scheme
attention owing to its advantages of simple structure, strong utilizes an adaptive system to auto-adjust the changes in the
robustness and its immunity towards matched uncertainties [10, plant due to perturbations or faults. The strategy is established in
26,27]. However, its convergence time might be notably long and a linear model even if the process is nonlinear, as it will be seen
later in this article. This is one of the advantages of the proposed
the infinite switching frequency caused by the sign function in
method since the control scheme is simple to conceive and imple-
its controller cannot be completely avoided [28]. Some improve-
ment. Besides, unlike the traditional predictive control scheme,
ments of the conventional SMC also have been presented with
the AP control design does not use an optimization problem and
the aim of eliminating its main disadvantages, for instance, by
the prediction horizon is just realized one period in the future,
proposing a switching sliding surface or a fixed frequency and
making it a simple but effective option to reduce computational
an adaptive backstepping sliding mode control [29–31]. On the
cost and to obtain a good controller performance. Additionally,
other hand, the habitual Proportional–Integral–Derivative (PID)
an Hysteresis Modulator (HM) is inserted in the control scheme
controller has been used in DC–DC converters because of its
to improve the performance of the controller since it provides
simplicity [4]. Nevertheless, this method does not ensure ro-
persistent excitation to the Adaptive Mechanism to ensure the
bustness over a wide range of operating points since it involves
parameters convergence [48,49]. This extra stage in the control
linearization around a specific operating point [4,10]. In spite of it,
scheme guarantees to solve problems such as saturation in the
PID controller has been jointly employed with other techniques
control law and parameter drift [47], among other advantages
to improve its performance taking advantages of the combined provided by this modulation stage. Finally, it will be exposed how
strategies, for instance, with a neural network based technique, the Adaptive Mechanism provides on-line information about the
with current feed-back loop by invoking hysteresis or with a condition of the process parameters. This allows us to infer the
finite-time control strategy [32–34]. Likewise, the Pulse-Width- status of the system, which could be useful to post-process the
Modulation (PWM) control scheme is also popular in DC–DC estimated parameters with the objective of detecting faults in the
converters. This technique utilizes a signal modulation by reason system.
of the switching necessity of the actuator to achieve the control In this article a wide range of experiments is presented. Firstly,
objective [9,35]. Furthermore, optimal controllers seek to solve numerical experiments are realized in Matlab/Simulink. Further-
the control problem statement by minimizing a properly de- more, because one of the most typical application of DC–DC
signed cost function. Normally, the cost function may incorporate converters is in the solar renewable energy field, this article
the uncertainties and constraints of the system which results in a presents an application of the proposed control scheme in a
proper control performance. However, the design of an adequate PV system at a numerical experiment level. This is realized by
cost function can be a complex mathematical realization [12,36, employing a PV panel as the voltage supply to the buck converter.
37]. Other control techniques for DC–DC converters are Super- Hence, it is validated that the proposal has a good performance
Twisting (STW) algorithms [38], Fuzzy Logic controllers [39], and even when the buck converter is subjected to the common ir-
Neural Networks based techniques [32,40]. On the other hand, radiation change perturbation presented in the PV panel due
hysteretic controllers are an alternative way to achieve the con- to shading conditions. Secondly, the technique exposed here, is
trol objective of DC–DC converters thanks to their simplicity, no also tested in an experimental platform where the results were
need of compensation, instantaneous response and no limita- compared to those obtained with a typical PI controller. Hence,
tions on the switch conduction time [7]. Still, the necessity of it will be seen that the proposed technique has a good perfor-
an adequate sensing stage makes that these controllers may be mance since it has a fast convergence response under changes
costly in terms of economic saving and computational work [3, in reference voltage, in input voltage and in the load. Moreover,
33]. Additionally, the standard Adaptive Control scheme is also the controller is robust since it maintains the output voltage well
an alternative to control the DC–DC converters. By using this regulated even when there are variations in the inductance which
technique, the converter not only estimates and adapts the values is one of the principal components of the converter.
of the uncertain parameters but also reaches the steady state Hereafter, Section 2 presents the DC–DC buck converter math-
in limited time. The paper in [5] realizes a very clear summary ematical model. Afterward, Section 3 exposes the proposed

Please cite this article as: N.I.P. de León Puig, D. Bozalakov, L. Acho et al., An Adaptive–Predictive control scheme with dynamic Hysteresis Modulation applied to a DC–DC
buck converter. ISA Transactions (2020), https://doi.org/10.1016/j.isatra.2020.05.015.
 N.I.P. de León Puig, D. Bozalakov, L. Acho et al. / ISA Transactions xxx (xxxx) xxx 3

3. Adaptive–predictive control design

In this section, the stages of the Adaptive–Predictive control

scheme are described. This control methodology is based on
an approximated dynamic mathematical model that depicts the
process to be controlled. This model is required to obtain the
Predictive Model, based on a discrete-time realization, which
generates the control law. On the other hand, the process model
Fig. 1. Buck converter electronic diagram.
is also useful to design the Adaptive Mechanism, which serves to
auto-adjust the process parameters under any possible change in
the plant, provoked mainly by external perturbations or faults in
the system. Fig. 2 depicts the general structure of the Adaptive–
Adaptive–Predictive control. Numerical experiment results are
Predictive control [46–48]. Where, for the buck converter im-
depicted in Section 4. On the other hand, the results of the
plementation, y(t) is the converter output voltage, yd (k + 1) is
experimental implementation in a real DC–DC buck converter are
the desired trajectory generated by the Driver Block, â1 (k), â2 (k)
shown in Section 5; the result discussion is presented in Section 6.
and â3 (k) are the process estimated parameters and u(k) is the
Finally, conclusions are drawn in Section 7.
control law. In addition, yr (t) is the reference voltage that may
2. DC–DC buck converter modeling change with the time, depending on the application or the user
preferences and it is processed in discrete time by the Driver
The conventional electronic circuit of the buck converter is Block, as well as the process output.
shown in Fig. 1. It is mainly composed of an input voltage source
(E), a switch mechanism (S), a diode (D), an inductor (L), a 3.1. Predictive model
capacitor (C), and finally the load, which usually is a resistor (R).
Based on circuit analysis and under ideal assumptions, the buck This block generates the adaptive–predictive control law u(k).
converter dynamic model may given by [10]: Since the Predictive Model is established in discrete-time domain,
the Euler’s discretization forward-difference law is invoked [44,
Lż1 (t) = −z2 (t) + u(t)E , 47]:
z2 (t)
C ż2 (t) = z1 (t) − , (1) dx x(k + 1) − x(k)
R ≈ , (5)
dt T
where, z1 = iL is the inductor current, z2 = Vo is the output
voltage and u is the switching control input. By assuming that where, T is the sampling-time period. Then, by transforming
the circuit is in Continuous Conduction Mode (CCM), this is, the Eq. (4) into discrete-time domain through the expression in (5),
minimum instantaneous value of the inductor current does not the Predictive Model is obtained as follows,
drop to zero even in load variation conditions, the following state −y(k)[− T22 − â2 (k)
+ â3 (k)] − y(k−1)
+ â1 (k)u(k)
normalization by using the time scale transformation τ = √t , ŷ(k + 1) =
T T2
. (6)
LC 1 â2 (k)
may be introduced [4]: T2
+ T
√
z1 L z2 In the equation above, ŷ(k + 1) represents the prediction of
x1 = , x2 = . (2) the process output for the instant k + 1 done at the instant k.
E C E
Hence, the Predictive Model uses the estimated parameters by
Thus, the normalized model that will serve for control design
the Adaptive Mechanism at the instant k: â1 (k), â2 (k), â3 (k). Note
purposes is given by [10]:
that Eq. (6) is a future estimation of the process output that
ẋ1 (t) = −x2 (t) + u(t), employs all the information from the dynamic process [46,50].
x2 (t) Later on, the output prediction (6) is expressed as a desired value
ẋ2 (t) = x1 (t) − , ŷ(k + 1) = yd (k + 1). This signal is perceived as a desired trajectory
Q
projected with its own dynamic and it is provided by the Driver
y(t) = Ex2 (t), (3) Block. Thereby, the following adaptive–predictive control law is
√
where Q = R C /L and y(t) is the output equation. Hence, obtained [46,47,50]:
x1 (t) and x2 (t) are now the normalized state variables, the in-
â2 (k) â2 (k) y(k−1)
ductor current and output voltage, respectively. From the system yd (k + 1)[ T12 + T
] + y(k)[− T22 − T
+ â3 (k)] + T2
in Eq. (3) it is possible to obtain the parameterized transfer u(k) = .
â1 (k)
function, which will be later useful to design the Predictive Model
and the Adaptive Mechanism of the control scheme. Therefore, (7)
the parameterized transfer function to (3) can be stated by its
relation, in Laplace domain, between the output process Y (s) and 3.2. Adaptive mechanism
the input control U(s) as follows:
The Adaptive Mechanism (AM) design is implemented in
Y (s) a1
G(s) = = , (4) continuous-time domain, nevertheless the estimated parameters
U(s) s 2 + a2 s + a3 are processed in discrete-time domain as the Predictive Model
where a1 , a2 and a3 are the plant parameters to be estimated by requires (see Fig. 2). The Adaptive Mechanism adjusts the model
the Adaptive Mechanism, and then employed by the Predictive parameters to compensate the dynamic changes provoked, for
Model. Additionally, note that Eq. (4) is a second order transfer instance, by external perturbations or faults in the process. Thus,
function that will facilitate to conceive the control design and it the AP control scheme is robust under perturbations, since the
is suitable to obtain an outstanding control performance for the AM stage will compensate the changes in the plant provoked by
buck converter. This is one of the main important attributes of those perturbations. Due to the Adaptive Mechanism employs in-
the proposed technique, since it is possible to control nonlinear formation from the prediction error in its model, it is not required
systems through a linear control design. extra mathematical or computational work to compensate the

Please cite this article as: N.I.P. de León Puig, D. Bozalakov, L. Acho et al., An Adaptive–Predictive control scheme with dynamic Hysteresis Modulation applied to a DC–DC
buck converter. ISA Transactions (2020), https://doi.org/10.1016/j.isatra.2020.05.015.
4 N.I.P. de León Puig, D. Bozalakov, L. Acho et al. / ISA Transactions xxx (xxxx) xxx

Fig. 2. Adaptive–Predictive control block diagram.

where λ1 and λ2 are the design parameters of the filters. Thus,

by filtering the signals and by applying basic algebra concepts,
the Laplace domain process output expression is obtained:
Uf
( )
Y (s) = a1 Uf + A2 Yf 1 + A3 Yf 2 = [a1 A2 A3 ] Yf 1 = θ ⋆T φ, (10)
Yf 2

where, A2 = λ1 − a2 , A3 = λ2 − a3 and θ ⋆T forms the parameters

vector. On the other hand, Uf , Yf 1 and Yf 2 are the corresponding
filtered signals. Then, an estimation of the equation above in
continuous-time domain can be written as:
uf
( )
Fig. 3. Adaptive Mechanism scheme.
ŷ(t) = â1 uf + Â2 yf 1 + Â3 yf 2 = [â1 Â2 Â3 ] yf 1 = θ̂ T φ (t), (11)
yf 2

where, Â2 = λ1 −â2 and Â3 = λ2 −â3 . Finally, by replacing Eq. (11)
in the vector representation (8), the dynamic model that realizes
the parameter estimation is obtained:

â˙ 1 = −γ uf [â1 uf + Â2 yf 1 + Â3 yf 2 − y],

˙
Â2 = −γ yf 1 [â1 uf + Â2 yf 1 + Â3 yf 2 − y],
˙
Â3 = −γ yf 2 [â1 uf + Â2 yf 1 + Â3 yf 2 − y]. (12)

3.3. Driver block

Fig. 4. Performance of a fixed band hysteresis system [52].
The aim of this block is to generate a reference trajectory to
be used by the Predictive Model based on a reference model. This
block basically induces that the process output smoothly evolves
error between the output process y(t) and the reference value
when the reference value or the operation point of the process
yr (t), as other control techniques do it [6,51]. variable is abruptly modified. Therefore, the Driver Block makes
The Adaptive Mechanism block could be seen as a ‘‘learning the task of updating at every sample time, with information from
system’’, and its realization is by the well known Gradient Algo- the process output y(k), a desired reference trajectory yd (k + 1).
rithm [49]. Note in Fig. 3, that the Adaptive Mechanism receives In this manner, the control law is physically implementable with
information from the process output and the control signal. In the a smooth behavior. Hence, based on a fairly typical example of a
approach presented here, the objective of this block is to obtain reference model [46], in this article it is proposed the Driver Block
the parameters â1 , â2 and â3 that are the estimation of the process equation as follows:
parameters in (4) that accompany the process dynamic.
˙ yd (k + 1) = α1 yr (k) + α2 yr (k − 1) − β1 y(k), (13)
From the Gradient Algorithm expressed as θ̂ (t) = −γ φ (t)e(t),
where γ > 0 is a proposed value known as adaptation gain, where yd (k + 1) is the output signal to be employed by the
φ (t) is the regression matrix and e(t) is the output estimation Predictive Model and yr (k) is the user reference value, processed
error e(t) = ŷ(t) − y(t); a vector representation of the parameter in discrete-time domain by the Driver Block. Moreover, α1 , α2
estimation can be written as [49]: and β1 are the design stable parameters assigned to attenuate
oscillations [53].
θ̂˙ (t) = −γ φ (t)[ŷ(t) − y(t)]. (8)
3.4. The hysteresis modulator in the predictive control scheme
The regression matrix is obtained by filtering the input and
output signals of the Adaptive Mechanism (see Fig. 3). To do this, In the field of power electronic converters, the basic concept of
two stable second order filters are proposed: hysteresis is useful, for this reason some control strategies appeal
s 1 to it to accomplish its control objective [7]. This is because the
GF 1 (s) = , GF 2 (s) = , (9) actuator of the power converters is a commutator device and the
s2 + λ 1 s + λ 2 s2 + λ1 s + λ2

Please cite this article as: N.I.P. de León Puig, D. Bozalakov, L. Acho et al., An Adaptive–Predictive control scheme with dynamic Hysteresis Modulation applied to a DC–DC
buck converter. ISA Transactions (2020), https://doi.org/10.1016/j.isatra.2020.05.015.
 N.I.P. de León Puig, D. Bozalakov, L. Acho et al. / ISA Transactions xxx (xxxx) xxx 5

Fig. 5. Scheme of the proposed control strategy along with the Hysteresis Modulation block.

hysteresis systems have a kind of on/off dynamic delay that fa-

cilitates the control implementation and improves the controller
performance [52,54]. Typically, controllers based on hysteresis
consists of programming the delay band that can be fixed to a cer-
tain value or not [52]; therefore, the process output is regulated
to a given reference value. For instance, the canonical behavior of
a hysteresis controller with a fixed band is represented in Fig. 4.
Recently, the hysteresis controllers applied to DC–DC converters
have demonstrated to be useful to adjust the system response
in the event of any change of the line input voltage or load
variations, to solve the problem of switching frequency variation
or to avoid infinite switching frequency [3,54,55]. However, the
techniques that appeal just to hysteresis to control the converters
require more cost in sensors and the task of adequately adjust the
band hysteresis may result complicated [7]. Moreover, typically,
the hysteresis strategy employs a static commutation [7,52,54], it
means, there is no dynamic transition in the hysteresis loop. Fig. 6. An hysteresis loop example.
In this article, a dynamic hysteresis system is proposed as
a modulator between the Adaptive–Predictive control law and Table 1
the plant to be controlled (see Fig. 5). This block takes advan- Buck converter parameters.

tage of the hysteresis qualities mentioned before, along with the DC input voltage 50 V
Switching frequency 20 kHz
Adaptive–Predictive control scheme benefits. In this proposal, the Inductor L 0.004 H
hysteresis has an important role in the control scheme since Capacitor C 2.5 µF
it provides the persistent excitation required for the Adaptive Load resistor R 22.2 
Resistor in inductor 0.04 
Mechanism to adjust the controller dynamics [49,56]. Moreover, Diode snubber resistance 0.1 
the Hysteresis Modulation block acts as a modulator of the pre- Diode snubber capacitance 20 µF
dictive control signal. In this manner, the plant receives an ad-
equate control signal to accomplish the control objective [47].
Finally, the hysteresis provides other advantages such as avoid-
4. Numerical experiment results
ing saturation problems or adjusting the system response under
abrupt external perturbations. In this section, different numerical experiments will be il-
Thereby, the predictive control scheme along with the Hys- lustrated to evaluate the effectiveness of the proposed predic-
teresis Modulator is presented in Fig. 5. Moreover, the imple- tive control technique. These experiments were realized in Mat-
mented dynamic hysteresis equation is proposed as follows [8, lab/Simulink 2018b. The realistic buck converter parameters used
48]: for these simulations were obtained from a real DC–DC setup and
are listed in Table 1. In addition, the parameters of the control
ż(t) = αhm [−z(t) + bhm sgn(x(t) + ahm sgn(z(t)))], (14) scheme are set as follows: αhm = 5, ahm = 5, bhm = 0.1,
γ = −0.001, λ1 = 10, λ2 = 5, α1 = α2 = 100, β1 = 200.
where ahm and bhm ∈ R+ are the hysteresis loop parameters and
These parameters were tuned by the trial and error procedure
z(t) is the internal variable of the hysteresis model. On the other
by fulfilling the criteria of stability for the filters and the Driver
hand, αhm is the constant-transition rate. In Fig. 5, it is possible
Block equation. Finally, the sample time T is set to 0.05 ms. In
to observe that the input of the hysteresis model x(t) is now the Fig. 7, the Simulink model of the buck converter employed for
predictive control law u(k) and its output will be the new control the experiments is shown.
law uhm (k) employed by the Adaptive Mechanism and the process.
Fig. 6 is depicted as an example, to appreciate the behavior 4.1. Step changes in the reference voltage
of the hysteresis dynamic equation. This graphic is a x(t) vs z(t)
plot of the open loop hysteresis response with parameters set as This subsection mainly exposes how the control objective
αhm = 5, bhm = 0.1 and ahm = 5. Furthermore, the input signal of regulating the output voltage of the buck converter is ac-
x(t) is a sinusoidal one x(t) = 50 sin(0.4π ). complished even when the reference voltage is modified. This

Please cite this article as: N.I.P. de León Puig, D. Bozalakov, L. Acho et al., An Adaptive–Predictive control scheme with dynamic Hysteresis Modulation applied to a DC–DC
buck converter. ISA Transactions (2020), https://doi.org/10.1016/j.isatra.2020.05.015.
6 N.I.P. de León Puig, D. Bozalakov, L. Acho et al. / ISA Transactions xxx (xxxx) xxx

Fig. 7. Buck converter Simulink model.

Fig. 8. (a) Output voltage response with step changes in the reference voltage. (b) (c) (d) The corresponding zoom-in versions.

scenario is realistic, for instance, when the user changes the hysteretical nature of the closed-loop system and it does not com-
reference value according to a required application of the buck promise the performance of the converter controller. Moreover,
converter. In these experiments, the reference voltage is altered depending on the application, the oscillations might be attenu-
by step changes every second from 15 V, to 20 V, to 25 V. In ated by employing adequate filters, among other techniques [57].
Fig. 8 it is possible to observe that the control performance is sat-
isfactorily executed since the different values of reference voltage Finally, in order to illustrate the actuation of the control law,
are achieved in fast time and without overshoots. Note that the the PWM signal (see Fig. 7) was filtered by using a low-pass filter
oscillation condition in the output response is due to the GPWM (s) = 10/(s + 10). This filtered signal is shown in Fig. 9

Please cite this article as: N.I.P. de León Puig, D. Bozalakov, L. Acho et al., An Adaptive–Predictive control scheme with dynamic Hysteresis Modulation applied to a DC–DC
buck converter. ISA Transactions (2020), https://doi.org/10.1016/j.isatra.2020.05.015.
 N.I.P. de León Puig, D. Bozalakov, L. Acho et al. / ISA Transactions xxx (xxxx) xxx 7

4.2. Step change in the input voltage

In this numerical experiment, a step down change from 50 to

40 V in the input voltage occurs at 0.5 s. This kind of scenario
is important to consider since it is a realistic situation when the
voltage supply to the converter is subjected to variable voltage
conditions such as voltage drop or overloading, or when the
voltage supply does not have a control stage to maintain the input
voltage to the converter regulated. One example of this scenario
could be a DC–DC power converter used in a photovoltaic system.
The result of this experiment is shown in Fig. 10. From this, it is
possible to observe that the needed time to reach the reference
voltage (20 V) is approximately 5 ms. Moreover, when the input
voltage changes from 50 V to 40 V at 0.5 s, the system takes
approximately the same time to retrieve the voltage reference
value.

4.3. Step change in the load resistance

This subsection presents the numerical experiment results

Fig. 9. Filtered PWM signal.
with a step change in the converter load. Just as the previous
experiments, the variation in the load is also a realistic case in
where it is appreciable the control action every time the voltage DC–DC converters applications. For instance, when an inductive
reference is readjusted. motor linked to the DC–DC converter is broken down, the load

Fig. 10. (a) Output voltage response with a step down change in the input voltage. (b) (c) The corresponding zoom-in versions.

Please cite this article as: N.I.P. de León Puig, D. Bozalakov, L. Acho et al., An Adaptive–Predictive control scheme with dynamic Hysteresis Modulation applied to a DC–DC
buck converter. ISA Transactions (2020), https://doi.org/10.1016/j.isatra.2020.05.015.
8 N.I.P. de León Puig, D. Bozalakov, L. Acho et al. / ISA Transactions xxx (xxxx) xxx

Fig. 11. (a) Output voltage response with a step up change in the load. (b) (c) The zoom-in versions.

seen by the converter will vary from its original value. This is why, value of the inductance from 0.002 H to 0.004 H. The reference
this experiment is also useful to validate the proposed control value is 25 V. Fig. 13 exposes the output voltage response. Addi-
technique. In this simulation, the change in the load is invoked tionally, in this experiment it is also shown the estimated param-
at 0.5 s, from 22.2  to 27.2 , and the reference voltage is 25 eters behavior in Fig. 14, where it is observable the parameters
V. Additionally, it is shown the response of the process estimated reaction when the fault in the inductor occurs.
parameters â1 , â2 and â3 . This is useful to observe two important
keys of the Adaptive Mechanism: (1) the dynamic response of the 4.5. The control strategy applied to a buck converter in photovoltaic
parameters, and (2) the utility of this block of giving information system
about the parameters variation to infer when an abrupt change
occurs in the process. In this experiment, the value of the gain One of the most important current application of DC–DC con-
γ = −1000 is greater than in the previous simulations to accel- verters is in the field of renewable energies. Therefore, in this
subsection a numerical experiment by using a PV system is pre-
erate the dynamic of the adaptive system since it is slower than
sented. In this experiment, the input voltage of the DC–DC buck
the control scheme dynamic [58]. The process output response is
converter is provided by a photovoltaic panel as is shown in
presented in Fig. 11. On the other hand, the estimated parameters
Fig. 15. The implemented PV array was the SunPower SPR series
are depicted in Fig. 12 where it is possible to note the variation in
305-E, included in the libary Simscape in Simulink. This array
the estimated parameters â2 and â3 when the change in the load
consists on strings of 66 PV modules connected in parallel. Each
is invoked. string consists of 5 modules connected in series. For this exper-
iment, the condition of temperature was fixed at 28 ◦ C and the
4.4. Step change in the inductor parameter irradiance seen by the PV panel was modeled as shown in Fig. 16.
This was programmed with changes that may emulate shading
For this numerical analysis, a fault in the inductance L was in- conditions. The reference voltage was fixed at 90 V and the pa-
troduced at 0.5 s. This fault was simulated by abruptly change the rameters of both the control scheme and the buck converter, are

Please cite this article as: N.I.P. de León Puig, D. Bozalakov, L. Acho et al., An Adaptive–Predictive control scheme with dynamic Hysteresis Modulation applied to a DC–DC
buck converter. ISA Transactions (2020), https://doi.org/10.1016/j.isatra.2020.05.015.
 N.I.P. de León Puig, D. Bozalakov, L. Acho et al. / ISA Transactions xxx (xxxx) xxx 9

Fig. 12. (a) Estimated parameters when a step change load occurs at 0.5 s. (b) The corresponding zoom-in graphic.

Fig. 13. (a) Output voltage response with a step change in the inductor. (b) (c) The corresponding zoom-in graphics.

Please cite this article as: N.I.P. de León Puig, D. Bozalakov, L. Acho et al., An Adaptive–Predictive control scheme with dynamic Hysteresis Modulation applied to a DC–DC
buck converter. ISA Transactions (2020), https://doi.org/10.1016/j.isatra.2020.05.015.
10 N.I.P. de León Puig, D. Bozalakov, L. Acho et al. / ISA Transactions xxx (xxxx) xxx

Fig. 14. (a) Estimated parameters with an induced error in the inductance value. (b) The zoom-in graphic.

Fig. 15. Buck converter implementation in Simulink by employing a PV panel array as the voltage supply.

Fig. 16. Different levels of irradiance. Fig. 17. Supplied voltage from the PV panel.

the same used in the previous exposed numerical experiments. 5. Results of the control implementation in a realistic buck
Fig. 17 shows the input voltage of the converter provided by the converter setup
PV panel under the changes in the irradiance. On the other hand,
Fig. 18 shows the converter output voltage where it is worth In order to experimentally validate the proposed predictive
to notice that the reference value is achieved even under the control, a variety of experiments were carried out using a bench-
different values of the irradiance in the PV panel. mark platform in the Electrical Energy Laboratory (EELAB) in

Please cite this article as: N.I.P. de León Puig, D. Bozalakov, L. Acho et al., An Adaptive–Predictive control scheme with dynamic Hysteresis Modulation applied to a DC–DC
buck converter. ISA Transactions (2020), https://doi.org/10.1016/j.isatra.2020.05.015.
 N.I.P. de León Puig, D. Bozalakov, L. Acho et al. / ISA Transactions xxx (xxxx) xxx 11

rad/s2 and kp = 0.001 rad/s. Worth mentioning, that for the ex-
perimental implementation, the Adaptive Mechanism stage was
not programmed, instead of it, constant values were employed to
program the predictive control in Eq. (7). This is possible since,
if the dynamic of the Adaptive Mechanism is slow (due to a
small value of γ ), the estimated parameters can be conceived
as constant values and under an ideal assumption, they are the
nearest possible to the nominal process parameters. Then, for the
experiments, the estimated parameters are set â1 = 0.1, â2 = 4
and â3 = 10. On the other hand, the values for the AP control
scheme are αhm = 8, bhm = 5, α1 = α2 = 10 and β1 = 20.

5.1. Change in the reference voltage

In Fig. 20 the response of the output voltage (Ch1: yellow sig-

nal) is depicted when the reference voltage of the buck converter
has a step change from 0 V to 20 V. The time per division in
Fig. 20(a) is 500 µs, on the other hand, in Fig. 20(b) is 5 ms.
Fig. 18. Output voltage with supplied voltage from the PV panel.
From here, it is notable that the time that takes the output
voltage in achieving the stable reference value in both cases, the
PI controller and the proposal AP control, is similar. However, the
response with the proposed control technique is better than the
one with the PI controller since it does not present overshoots. A
similar conclusion can be deduced from the experiment when the
reference voltage has a step change from 20 V to 0 V in Fig. 21.
Here the time per division is 1 ms and 10 ms, respectively. Signals
blue (Ch2) and magenta (Ch3) are the measured current through
the load and in the inductor converter, respectively. Note that
in these experiments the voltage probe was scaled by two the
original measured value.

5.2. Step change in the input voltage

The results presented in this subsection were obtained by

inducing a step change in the input voltage from 50 V to 100
V and vice versa. The reference value was set at 25 V. Fig. 22
gathers the results when the input voltage has a step up change.
On the other hand, Fig. 23 recovers the results when the input
voltage varies from 100 V to 50 V. In both cases, by employing
both control techniques, the results are very similar. The system
barely notes the change and it maintains the output voltage (Ch1)
in the desired value. In these experiments the time per division is
25 ms. Additionally, the green signal (Ch4) is the input voltage to
the buck converter and the magenta signal (Ch3) is the converter
current. Finally, the blue signal (Ch2) shows the moment when
the change in the input voltage occurs. From Fig. 23, the measured
inductor current shows that there is a spike with a high di/dt.
Thus, it is worth to mention that this transient does not result
into a change of the output voltage and it is presented in both
control algorithms, it may be concluded that the origin of this
issue is in the measurement system and not due to malfunction
in the controllers.
Fig. 19. Experimental platform.

5.3. Step change in the load

Ghent University. The parameters of each component are listed In this experiment, a step change in the load is induced from
in Table 1. Furthermore, the setup is the one presented in Fig. 19, 22.2  to 27.2  and vice versa. First, when the step up change
which is a three-phase inverter described in [59,60]. However, it is invoked, the time per division in the results with the PI control
was adjusted such that only one leg is used to get a buck con- technique is 1 ms and 100 µs in the zoom version. On the other
verter configuration where the proposed control algorithm was hand, the time per division for the results with the AP proposal,
implemented. The setup is integrated with a DSP Texas Instru- is 5 ms and 2.5 ms in the zoom version. From Fig. 24, it is visible
ment processor, model TMS320F28335. This processor provides a that the time in achieving the reference voltage value is a bit
clock speed of 150 MHz and a floating-point unit. faster with the PI controller, nevertheless in contrast with it, the
The results obtained with the proposed technique are com- response of the AP controller does not present the overshoot
pared with the results obtained by employing a common PI unlike the case with the PI control. In this figure, the green signal
controller. The parameters of the PI controller are ki = 0.999 (Ch4) is used to identify when the change in the load occurs. The

Please cite this article as: N.I.P. de León Puig, D. Bozalakov, L. Acho et al., An Adaptive–Predictive control scheme with dynamic Hysteresis Modulation applied to a DC–DC
buck converter. ISA Transactions (2020), https://doi.org/10.1016/j.isatra.2020.05.015.
12 N.I.P. de León Puig, D. Bozalakov, L. Acho et al. / ISA Transactions xxx (xxxx) xxx

Fig. 20. Output voltage response (Ch1), load current (Ch2) and inductor current (Ch3), under a step up change in the reference voltage from 0 V to 20 V by employing
PI controller (a) and the proposed AP controller (b).

Fig. 21. Output voltage response (Ch1), load current (Ch2) and inductor current (Ch3), under a step down change in the reference voltage from 20 V to 0 V by
employing PI controller (a) and the proposed AP controller (b).

Fig. 22. Output voltage response (Ch1), current inductor (Ch3) and input voltage (Ch4), under a step up change in the input voltage from 50 V to 100 V by employing
PI controller (a) and the proposed AP controller (b).

set of graphics in Fig. 25 presents alike results obtained by doing since an abrupt change in the inductor may mean the converter
the step down change in the load. In these results, the time per is completely broken down, it is a good experiment to validate
division for the PI case is 2.5 ms and 250 µs in the zoom version. robustness of the control technique. This experiment was realized
Moreover, the time per division for the AP case is 5 ms and 1 ms by switching the corresponding transistor in the setup to make
in the zoom version. the adequate array with the inductors. In this specific experiment,
the AP control has a faster response (5 ms/div and 1 ms/div in
5.4. Step change in the inductor the zoom version) than the PI case (2.5 ms/div and 55 µs/div
in the zoom version). Moreover, the PI controller results contain
The set of graphics in Fig. 26 depicts the experiment results more overshooting. In these experiments, the green signal (Ch4)
when a step up change in the inductor is triggered from 2 mH represents the exact moment when the inductors are switching.
to 4 mH. Although this scenario may not be present in practice, On the other hand, Fig. 27 depicts the results by doing the step

Please cite this article as: N.I.P. de León Puig, D. Bozalakov, L. Acho et al., An Adaptive–Predictive control scheme with dynamic Hysteresis Modulation applied to a DC–DC
buck converter. ISA Transactions (2020), https://doi.org/10.1016/j.isatra.2020.05.015.
 N.I.P. de León Puig, D. Bozalakov, L. Acho et al. / ISA Transactions xxx (xxxx) xxx 13

Fig. 23. Output voltage response (Ch1), current inductor (Ch3) and input voltage (Ch4), under a step down change in the input voltage from 100 V to 50 V by
employing PI controller (a) and the proposed AP controller (b).

Fig. 24. Output voltage response (Ch1), and inductor current (Ch2) under a step up change in the load from 22.2  to 25.2  by employing PI controller (a) and
its zoom version (b), and the proposed AP controller (c) and its zoom version (d).

down change in the inductor. Note that in this case, for both cases the control strategy works as expected. If a comparison
controllers, the system barely notices the change. Here, the time between the results from numerical experiments and real im-
per division is 250 µs and 2.5 ms, respectively. plementation is made, it can be seen that the time response
is slower in the real cases. This difference is mainly because
6. Discussion of results the simulation model uses ideal conditions, while in the case
of the real implementation, the system is subjected to different
In this article, the presented results, both numerical and ex- conditions, for instance, the signal-processing time, the velocity
perimental, allow us to draw some conclusions regarding the of the controller processing unit and the optimal useful life of the
proposed control technique. Firstly, the results make inference actuators, among other characteristics always presented in real
that the Adaptive–Predictive control strategy is an adequate op- control systems. Nevertheless, even in the real implementation,
tion to control the buck converter device since its main objective the proposed technique provides acceptable results in the time
is totally accomplished. That is, regulate the output voltage to response and in the robustness of the controller. On the other
a desired value. Realistic scenarios have been emulated in nu- hand, as was presented in Section 5, the conventional PI controller
merical simulations and experimental implementation and in all is a good option to accomplish the objective, however this kind

Please cite this article as: N.I.P. de León Puig, D. Bozalakov, L. Acho et al., An Adaptive–Predictive control scheme with dynamic Hysteresis Modulation applied to a DC–DC
buck converter. ISA Transactions (2020), https://doi.org/10.1016/j.isatra.2020.05.015.
14 N.I.P. de León Puig, D. Bozalakov, L. Acho et al. / ISA Transactions xxx (xxxx) xxx

Fig. 25. Output voltage response (Ch1), and inductor current (Ch2) under a step down change in the load from 25.2  to 22.2  by employing PI controller (a) and
its zoom version (b), and the proposed AP controller (c) and its zoom version (d).

Fig. 26. Output voltage response (Ch1), inductor current (Ch2), and load current (Ch3) under a step up change in the inductor from 2 mH to 4 mH by employing
PI controller (a) and its zoom version (b), and the proposed AP controller (c) and its zoom version (d).

Please cite this article as: N.I.P. de León Puig, D. Bozalakov, L. Acho et al., An Adaptive–Predictive control scheme with dynamic Hysteresis Modulation applied to a DC–DC
buck converter. ISA Transactions (2020), https://doi.org/10.1016/j.isatra.2020.05.015.
 N.I.P. de León Puig, D. Bozalakov, L. Acho et al. / ISA Transactions xxx (xxxx) xxx 15

Fig. 27. Output voltage response (Ch1), inductor current (Ch2), and load current (Ch3) under a step down change in the inductor from 4 mH to 2 mH by employing
PI controller (a) and its zoom version (b), and the proposed AP controller (c) and its zoom version (d).

of controller has a limited behavior when the system is subjected be robust under different faulty scenarios. Moreover, the exposed
to external perturbations, for instance when the value of the results, both numerical and experimental, show that the control
inductor varies (Fig. 26). As it was shown in the experimental objective is quickly accomplished, which is a notable and always
results, the AP control technique works better than the case required property in the field of DC–DC converter applications. On
with the PI controller when the parameters of the system are the other hand, the theoretical and analytical outcomes exhibited
subjected to changes, external perturbations or faults. For this in this article about an adaptive–predictive control design are the
reason, in general, the exposed proposal can be considered as basis of a planned future work, where it is expected to implement
a better option since the AP control technique is more robust the proposed control scheme into a whole distributed energy
than the PI controller. On the other hand, the proposal allows system where the main energy is supplied by a renewable source.
to extract more information from the system thanks to the pa- Hence, the predictive control strategy will have to accomplish
rameter estimation phase. Hence, the objective in a near future is the control objective of extracting the maximum power from
to implement the stage of the Adaptive Mechanism in the setup a renewable source even when it is subjected to external per-
in order to diagnose the system in real time. Nonetheless, from turbations, efficiently regulating the extracted voltage through
numerical experiments it is possible to deduce since now, that a DC–DC converter; and properly controlling a DC–AC inverter
this phase will work as expected. Finally, from numerical and coupled to a grid network.
experimental results, it is notable that the output voltage contains
a ripple characteristic. This is an essential condition produced Declaration of competing interest
by the hysteresis modulation stage and it is typically presented
in controllers with hysteresis. Normally, the DC–DC converter The authors declare that they have no known competing finan-
objective is about regulation, and since the plant itself or the user cial interests or personal relationships that could have appeared
load may behave as signal filters, this rippling behavior is usually to influence the work reported in this paper.
accepted in most of the implementation scenarios. In comparison
with other control strategies in the state-of-the-art, the main Acknowledgments
contribution of this paper lies in the on-line process state mon-
itoring by the hystereticaly manipulated adaptive mechanism. This work has been partially funded by the Spanish Ministry
Additionally, this adaptive stage is also able to compensate possi- of Economy and Competitiveness/Fondos Europeos de Desarrollo
ble mismatched errors in the process modeling during the control Regional (MINECO/FEDER), Spain with grant number DPI2015-
design. Thus, the proposed control approach does not require 64170-R and by the scholarship for doctoral studies abroad pro-
an accurate mathematical model of the process to obtain the vided by the CONACyT, Mexico.
control law, unlike, for instance, the conventional predictive con-
trol design. Finally, the low complexity of the control technique References
makes possible to reduce mathematical and computational efforts
[1] Mohan N, Undeland TM. Power electronics: converters, applications, and
in comparison, for instance, to the conventional optimal control design. John wiley & sons; 2007.
techniques. [2] Kislovski A. Dynamic analysis of switching-mode DC/DC converters.
Springer Science & Business Media; 2012.
7. Conclusions [3] Komurcugil H. Non-singular terminal sliding-mode control of DC-DC buck
converters. Control Eng Pract 2013;21(3):321–32.
[4] Sira-Ramirez H. Nonlinear PI controller design for switchmode DC-to-DC
In this paper, a new control scheme for a DC–DC buck con- power converters. IEEE Trans Circuits Syst 1991;38(4):410–7.
verter has been completely developed and tested through nu- [5] Nizami TK, Mahanta C. An intelligent adaptive control of DC-DC buck
merical and real experiments. The control approach is a simple converters. J Franklin Inst B 2016;353(12):2588–613.
[6] Leng Z, Liu Q. A simple model predictive control for buck converter
manner to accomplish the main power control objective of these
operating in CCM. In: 2017 IEEE international symposium on predictive
electronic devices. That is, efficiently regulating the output volt- control of electrical drives and power electronics (PRECEDE). IEEE; 2017,
age of a DC–DC converter to a desired value even when the p. 19–24.
system is subjected to the commonly existing perturbations, such [7] Castilla M, de Vicuna LG, Guerrero JM, Miret J, Berbel N. Simple low-cost
as variation in the input voltage, fast changing in the reference hysteretic controller for single-phase synchronous buck converters. IEEE
Trans Power Electron 2007;22(4):1232–41.
command voltage, or a possible fault in some of the buck-process [8] Ponce de León Puig N, Acho L, Rodellar J. Design and experimental
parameters. The proposed predictive control scheme has been implementation of a hysteresis algorithm to optimize the maximum power
experimentally evaluated in a benchmark platform outcoming to point extracted from a photovoltaic system. Energies 2018;11(7):1866.

Please cite this article as: N.I.P. de León Puig, D. Bozalakov, L. Acho et al., An Adaptive–Predictive control scheme with dynamic Hysteresis Modulation applied to a DC–DC
buck converter. ISA Transactions (2020), https://doi.org/10.1016/j.isatra.2020.05.015.
16 N.I.P. de León Puig, D. Bozalakov, L. Acho et al. / ISA Transactions xxx (xxxx) xxx

[9] Wang Z, Li S, Yang J, Li Q. Current sensorless finite-time control [34] Yuan C, Huangfu Y, Ma R, Zhao B, Bai H. Nonlinear PI and finite-time
for buck converters with time-varying disturbances. Control Eng Pract control for DC-dc converter based on exact feedback linearization. In:
2018;77:127–37. IECON 2019-45th annual conference of the IEEE industrial electronics
[10] Sira-Ramírez H. On the generalized PI sliding mode control of DC-to-DC society, Vol. 1. IEEE; 2019, p. 6398–403.
power converters: a tutorial. Int J Control 2003;76(9–10):1018–33. [35] Oucheriah S, Guo L. PWM-based adaptive sliding-mode control for boost
[11] Marcillo KEL, Guingla DAP, Rocha EM, Barra W, Benavides DAV, DC-DC converters. IEEE Trans Ind Electron 2013;60(8):3291–4.
de Medeiros RLP. Parameter optimization of an interval robust controller [36] Vargas AN, Sampaio LP, Acho L, Zhang L, do Val JB. Optimal control of DC-
of a buck converter subject to parametric uncertainties. In: 2018 IEEE third DC buck converter via linear systems with inaccessible Markovian jumping
ecuador technical chapters meeting (ETCM). IEEE; 2018. modes. IEEE Trans Control Syst Technol 2016;24(5):1820–7.
[12] Leng Z, Liu J, Liu Q, Wang H. A digital optimal control method for DC- [37] Sundareswaran K, Kuruvinashetti K, Hariprasad B, Sankar P, Nayak P, Vig-
DC converters based on simple model. In: 2008 workshop on power neshkumar V. Optimization of dual input buck converter control through
electronics and intelligent transportation system. IEEE; 2008, p. 88–93. genetic algorithm. IFAC Proc Vol 2014;47(1):142–6.
[38] Chalanga A, Kamal S, Fridman LM, Bandyopadhyay B, Moreno JA.
[13] Singh G, Verma V, Urooj S, Haque A. Regulation of DC bus voltage for DC
Implementation of super-twisting control: Super-twisting and higher or-
microgrid using PSIM. In: 2018 5th IEEE Uttar Pradesh section international
der sliding-mode observer-based approaches. IEEE Trans Ind Electron
conference on electrical, electronics and computer engineering (UPCON).
2016;63(6):3677–85.
IEEE; 2018.
[39] Siano P, Citro C. Designing fuzzy logic controllers for DC-DC converters
[14] Wang H, Wu W, Li Y, Blaabjerg F. A coupled-inductor-based buck-boost AC-
using multi-objective particle swarm optimization. Electr Power Syst Res
DC converter with balanced dc output voltages. IEEE Trans Power Electron 2014;112:74–83.
2018;34(1):151–9. [40] Jafari M, Malekjamshidi Z. Design, simulation and implementation of an
[15] Li W, Zheng Y, Li W, He X, et al. A smart and simple PV charger for portable adaptive controller on base of artificial neural networks for a resonant
applications. In: 2010 twenty-fifth annual IEEE applied power electronics DC-DC converter. In: 2011 IEEE ninth international conference on power
conference and exposition (APEC). IEEE; 2010, p. 2080–4. electronics and drive systems. IEEE; 2011, p. 1043–6.
[16] Liu B, Ren R, Wang F, Costinett D, Zhang Z. A variable frequency ZVS [41] Zhou G, Xu J, Jin Y. Improved digital peak current predictive control for
control of a three-level buck without zero crossing detection for wide- switching DC-DC converters. IET Power Electron 2011;4(2):227–34.
range output voltage battery chargers. In: 2019 IEEE applied power [42] Aguilera RP, Quevedo DE. Predictive control of power converters: Designs
electronics conference and exposition (APEC). IEEE; 2019, p. 2311–6. with guaranteed performance. IEEE Trans Ind Inf 2015;11(1):53–63.
[17] Wai R-J, Duan R-Y. High-efficiency power conversion for low power fuel [43] Rawlings JB, Mayne DQ. Model predictive control: Theory and design.
cell generation system. IEEE Trans Power Electron 2005;20(4):847–56. Wisconsin: Nob Hill Pub. Madison; 2009.
[18] Gavriluta C, Citro C, Nisak K, San Segundo HB. A simple approach for fast [44] Lashab A, Sera D, Guerrero JM. A dual-discrete model predictive control-
controller prototyping for a three phase interleaved DC-DC converter. In: based MPPT for PV systems. IEEE Trans Power Electron 2019;34(10)
2012 IEEE international symposium on industrial electronics. IEEE; 2012, (2019).
p. 2015–9. [45] Shan Y, Hu J, Li Z, Guerrero JM. A model predictive control for renewable
[19] García O, Zumel P, De Castro A, Cobos A. Automotive DC-DC bidirectional energy based AC microgrids without any PID regulators. IEEE Trans Power
converter made with many interleaved buck stages. IEEE Trans Power Electron 2018;33(11):9122–6.
Electron 2006;21(3):578–86. [46] Sánchez JM, Rodellar J. ADEX Optimized adaptive controllers and systems.
[20] Erb DC, Onar OC, Khaligh A. Bi-directional charging topologies for plug-in Springer; 2015.
hybrid electric vehicles. In: 2010 twenty-fifth annual IEEE applied power [47] Ponce de León Puig NI, Acho Zuppa L, Rodellar Benedé J. Adaptive
electronics conference and exposition (APEC). IEEE; 2010, p. 2066–72. predictive control of a base-isolated hysteretic system. In: ICSTCC 2017:
[21] Hu X, Zou C, Tang X, Liu T, Hu L. Cost-optimal energy management 21st international conference on system theory, control and computing:
Sinaia, Romania: October 19-21, 2017: Proceedings book. 2017, p. 395–400.
of hybrid electric vehicles using fuel cell/battery health-aware predictive
[48] de León Puig NIP, Acho L, Rodellar J, Vidal Y, Gutíerrez-Arias JEM.
control. IEEE Trans Power Electron 2019;35(1):382–92.
Hysteretic delta modulator to prevent parameter drift in adaptive-based
[22] Hu X, Li Y, Lv C, Liu Y. Optimal energy management and sizing of
controllers. In: 2017 25th mediterranean conference on control and
a dual motor-driven electric powertrain. IEEE Trans Power Electron
automation (MED). IEEE; 2017, p. 48–53.
2018;34(8):7489–501.
[49] Sastry S, Bodson M. Adaptive control: Stability, convergence and
[23] Jiang Y, Qahouq JAA, Haskew TA. Adaptive step size with adaptive-
robustness. Courier Corporation; 1989.
perturbation-frequency digital MPPT controller for a single-sensor photo- [50] Sánchez JM, Rodellar J. Adaptive predictive control: From the concepts to
voltaic solar system. IEEE Trans Power Electron 2013;28(7):3195–205. plant optimization. Prentice Hall PTR; 1995.
[24] Li Q, Wolfs P. A review of the single phase photovoltaic module integrated [51] Wang B, Ma G, Xu D, Zhang L, Zhou J. Switching sliding-mode control
converter topologies with three different DC link configurations. IEEE Trans strategy based on multi-type restrictive condition for voltage control of
Power Electron 2008;23(3):1320–33. buck converter in auxiliary energy source. Appl Energy 2018;228:1373–84.
[25] Dehghanzadeh A, Farahani G, Vahedi H, Al-Haddad K. Model predictive [52] Nguyen-Van T, Abe R, Tanaka K. A digital hysteresis current control
control design for DC-DC converters applied to a photovoltaic system. Int for half-bridge inverters with constrained switching frequency. Energies
J Electr Power Energy Syst 2018;103:537–44. 2017;10(10):1610.
[26] Alsmadi YM, Utkin V, Haj-ahmed MA, Xu L. Sliding mode control of power [53] Ogata K, Yang Y. Modern control engineering, Vol. 4, London, 2002.
converters: DC/DC converters. Internat J Control 2018;91(11):2472–93. [54] Yao Q, Holmes DG. A simple, novel method for variable-hysteresis-
[27] Edwards C, Spurgeon S. Sliding mode control: theory and applications. Crc band current control of a three phase inverter with constant switching
Press; 1998. frequency. In: Conference record of the 1993 IEEE industry applications
[28] Utkin V. Sliding mode control of DC/DC converters. J Franklin Inst B conference twenty-eighth IAS annual meeting. IEEE; 1993, p. 1122–9.
2013;350(8):2146–65. [55] Tan S-C, Lai Y, Tse CK, Cheung MK. Adaptive feedforward and feedback
[29] Ma G, Wang B, Xu D, Zhang L. Switching control strategy based on non- control schemes for sliding mode controlled power converters. IEEE Trans
singular terminal sliding mode for buck converter in auxiliary energy Power Electron 2006;21(1):182–92.
source. Energy Procedia 2018;145:139–44. [56] Fuh C-C, Tsai H-H. Adaptive parameter identification of servo control
[30] Abeywardana DBW, Hredzak B, Agelidis VG. A fixed-frequency sliding systems with noise and high-frequency uncertainties. Mech Syst Signal
mode controller for a boost-inverter-based battery-supercapacitor hybrid Process 2007;21(3):1437–51.
energy storage system. IEEE Trans Power Electron 2017;32(1):668–80. [57] Rashid M. Power electronics: Circuits, devices and applications. Interna-
[31] Liping F, Yazhou Y, Boshnakov K. Adaptive backstepping based terminal tional ed.. Erscheinungsdatum: Prentice Hall; 2003.
sliding mode control for DC-DC converter. In: 2010 international confer- [58] Åström KJ, Wittenmark B. Adaptive control. Courier Corporation; 2013.
[59] Bozalakov DV, Mnati MJ, Laveyne J, Van den Bossche A, Vandevelde L.
ence on computer application and system modeling (ICCASM 2010), Vol.
Voltage unbalance and overvoltage mitigation by using the three-phase
9. IEEE; 2010.
damping control strategy in battery storage applications. In: 2018 7th
[32] Maruta H, Taniguchi H, Kurokawa F. A study on effects of different control
international conference on renewable energy research and applications
period of neural network based reference modified PID control for DC-
(ICRERA). IEEE; 2018, p. 753–9.
DC converters. In: 2016 15th IEEE international conference on machine
[60] Bozalakov D, Mnati MJ, Laveyne J, Desmet J, Vandevelde L. Battery
learning and applications (ICMLA). IEEE; 2016, p. 460–5.
storage integration in voltage unbalance and overvoltage mitigation control
[33] Rivetta CH, Emadi A, Williamson GA, Jayabalan R, Fahimi B. Analysis and
strategies and its impact on the power quality. Energies 2019;12(8):1501.
control of a buck DC-DC converter operating with constant power load in
sea and undersea vehicles. IEEE Trans Ind Appl 2006;42(2):559–72.

Please cite this article as: N.I.P. de León Puig, D. Bozalakov, L. Acho et al., An Adaptive–Predictive control scheme with dynamic Hysteresis Modulation applied to a DC–DC
buck converter. ISA Transactions (2020), https://doi.org/10.1016/j.isatra.2020.05.015.