See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/331243562

A model predictive control based hybrid MPPT method for boost converters

Article in International Journal of Electronics · February 2019

DOI: 10.1080/00207217.2019.1582715

CITATIONS READS

46 2,417

2 authors:

Erdal Irmak Naki Güler

Gazi University Gazi University
95 PUBLICATIONS 789 CITATIONS 54 PUBLICATIONS 446 CITATIONS

SEE PROFILE SEE PROFILE

All content following this page was uploaded by Naki Güler on 25 September 2019.

The user has requested enhancement of the downloaded file.

International Journal of Electronics

ISSN: 0020-7217 (Print) 1362-3060 (Online) Journal homepage: https://www.tandfonline.com/loi/tetn20

A model predictive control-based hybrid MPPT

method for boost converters

Erdal Irmak & Naki Güler

To cite this article: Erdal Irmak & Naki Güler (2019): A model predictive control-based
hybrid MPPT method for boost converters, International Journal of Electronics, DOI:
10.1080/00207217.2019.1582715

To link to this article: https://doi.org/10.1080/00207217.2019.1582715

Accepted author version posted online: 20

Feb 2019.
Published online: 06 Mar 2019.

Submit your article to this journal

Article views: 82

View Crossmark data

Full Terms & Conditions of access and use can be found at

https://www.tandfonline.com/action/journalInformation?journalCode=tetn20
INTERNATIONAL JOURNAL OF ELECTRONICS
https://doi.org/10.1080/00207217.2019.1582715

A model predictive control-based hybrid MPPT method for boost

converters
a b
Erdal Irmak and Naki Güler
a
Electrical and Electronics Engineering Department, Faculty of Technology, Gazi University, Ankara, Turkey;
b
Technical Sciences Vocational School, Gazi University, Ankara, Turkey

ABSTRACT ARTICLE HISTORY

In this study, Perturb and Observe Maximum Power Point Tracking Received 5 September 2018
algorithm and Model Predictive Control (MPC) algorithm are combined Accepted 3 January 2019
to improve the dynamic performance of the control structure of boost KEYWORDS
converters used in renewable energy systems. For this purpose, a digital Model predictive control
control method that does not need any comparator is developed. The (MPC); MPPT; boost
maximum power point current value is determined from the Maximum converter; photovoltaic
Power Point Tracking (MPPT) algorithm and it is used as the reference for system; current control
current control operation in the MPC algorithm. Dynamic capability and
tracking performance of the proposed control method is validated by
simulation and experimentally. The dynamic behaviour of the control
algorithm is analysed by carrying out step change tests. Results show
that the control method has excellent dynamic performance thanks to its
feature of reference tracking in only one switching period. Furthermore,
overshoots and undershoots are successfully eliminated.

1. Introduction
Photovoltaic energy sources are widely used in renewable energy applications. The significant role
of power converters in energy conversion and power flow control of photovoltaic energy sources is
one of the reasons for increasing studies in this subject lately. Because the most PV panels have low
voltage levels, boost converters are usually preferred in these systems (Hernanz, Guede,
Barambones, Zulueta, & Gamiz, 2017; Qian, Yu, Su, Sun, & Lu, 2017). In the control side, maximum
power point tracking (MPPT) algorithms have an important role in obtaining the maximum power
produced from PV panels.
Among several MPPT algorithms, Perturb and Observe (P&O) and Incremental Climbing (IC)
techniques are well-known and most preferred ones. Tracking efficiency of both algorithms is over
98% (Danandeh & Mousavi, 2018) and they not require system parameters, which can make them
easy to apply. These algorithms determine the maximum power point (MPP) by continuously
observing the changes of power and voltage levels. Furthermore, they can calculate the related
current, voltage or duty cycle values to achieve the MPPT operation. The operating point is shifted
to match the MPP condition by decreasing or increasing the voltage or current (Freitas, Tofoli,
Junior, Daher, & Antunes, 2016).
For P&O method, the switching signal can directly be generated by comparing the duty cycle with a
triangle waveform. Thus, power flow is controlled by adjusting the duty cycle (Başoğlu, 2018). Because
of the slow response times usually occurred during the use of this method, there may be delays in

CONTACT Erdal Irmak [email protected] Electrical and Electronics Engineering Department, Faculty of Technology,
Gazi University, Ankara, Turkey
© 2019 Informa UK Limited, trading as Taylor & Francis Group
2 E. IRMAK AND N. GÜLER

power flow control (Abbes, Abid, & Loukil, 2015; Metry, Bayhan, Balog, & Rub, 2016). Increasing the step
size can be considered as a solution to overcome this issue; however, this process causes some
undesired situations like overshoots and undershoots over the system responses. Detailed information
about these overshoot issues can be found in Ahmed & Salam (2016) that proposes the variable step
P&O algorithm to improve the dynamic response of the control method.
For all MPPT algorithms, the dynamic response speed is especially important when some instant
changes occur on the system. These changes may be generated due to partial shading effect
(Kandemir, Borekci, & Cetin, 2018). To increase the response time in such situations, some algo-
rithms are presented in literature such as artificial neural networks (ANN), Genetic Algorithm, fuzzy
logic, sliding mode and MPC (Khosravi, Heshmatian, Khaburi, Garcia, & Rodriguez, 2017; Kim, Moon,
& Kwak, 2018; Mahendran & Ramabadran, 2016; Öztürk & Çelik, 2014; Singh, Fulwani, & Kumar,
2017). All these studies report that these control methods need a high computational capability to
obtain a fast dynamic response. As an alternative and more recent technique, MPC approach has
led to increase in processing capabilities of digital hardware (Rodriguez & Cortes, 2012). Beside fast
dynamic response and sensitive control capability, MPC method has some other advantages such
as easily adaptation of multivariable case, easy inclusion of non-linearity, easy implementation of
the controller and suitable structure for the inclusion of modifications and extensions depending
on specific applications (Rodriguez & Cortes, 2012).
MPC algorithm is applied to most of the power converters, and dynamic responses are
discussed in boost converters for voltage control studies. Dynamic response analysis shows that
MPC technique performs voltage control processes with slight overshoots (Seo, Kim, & Choi, 2017).
According to some comparison studies (Cheng et al., 2018; Ren, Zheng, & Li, 2015), MPC technique
gives better results than conventional methods. Furthermore, MPC-based hybrid control methods
increase the sensitivity (Kim, Park, Kim, & Lee, 2014; Lashab, Sera, Guerrero, Mathe, & Bouzid, 2018).
For this purpose, many algorithms are combined with MPPT algorithm to obtain a hybrid structure
where the current or voltage values at the MPP are usually derived from the MPPT algorithm, and
then it is used as a reference for MPC (Adhikari, 2018). For example, MPP tracking capability is
tested in Priya & Parimi, (2016) that combines the FL and P&O MPPT algorithms. Similar to this
study, some other studies have shown that hybrid algorithms have much better control capabilities
than classical algorithms. But, delayed response time is emphasised as a disadvantage in Priya &
Parimi (2016). As a solution, model-based MPPT and ANN technique are combined in (Khosravi et
al. (2017) to increase the dynamic response capability; thereby the transition time has reduced up
to 45 ms. In another study, MPC and P&O MPPT techniques are adapted and dynamic performance
of the control structure is tested under variable irradiance conditions. MPP voltage that is deter-
mined by the MPPT algorithm is used as a reference in the MPC algorithm. According to the results,
the proposed algorithm has considerably reduced the dynamic response time (Metry, Shadmand,
Balog, & Abu-Rub, 2017).
In brief, the overall efficiency of a PV system depends on PV panel structure, DC–DC converter
and the control effectiveness of the MPPT technique (Mosa, Shadmand, Balog, & Abu-Rub, 2017).
To increase the PV module efficiency, advanced fabrication technology is needed which will
increase the cost of modules. Because of that, efforts are focused on other system components.
Recent studies show that DC–DC converter efficiency changes between 92% and 98% (Abu-Rub,
Malinowski, & Al-Haddad, 2014).
Considering all the issues summarised above, this study is focused on improving the dynamic
capability of the control method. For this purpose, the classical P&O MPPT algorithm is
combined with the MPC algorithm. Mathematical background, detailed analysis of proposed
control method and its hybrid structure are given in Section 2. Control capability of the
proposed model is tested and verified both in the simulation environment and experimentally
in real time, results of which are given in Sections 3 and 4, respectively. Finally, a brief
conclusion is presented in the last section.
INTERNATIONAL JOURNAL OF ELECTRONICS 3

2. Analysis of the boost converter and current control schema

In order to achieve the power flow control, this study uses a boost converter structure connected
to PV module outputs. Figure 1 illustrates the block schema of the whole system. In this circuit, the
switch, the capacitor and the diode are considered as ideal elements.
In control side, P&O algorithm is used to determine the current value at the MPP. This value is
used as the reference in the MPC algorithm that controls the input current of the converter. MPC
algorithm produces appropriate switching signals to equalise the inductor current to the reference
value. Thus, a hybrid control algorithm consisting of MPC-based MPPT structure is achieved for
power flow control of PV modules.

2.1. Continuous conduction mode (CCM) analysis

DC/DC boost converters are power electronic circuits that are typically used to increase the input
voltage. Therefore, they are especially preferred in low voltage power sources such as PV modules.
DC–DC converters are analysed according to some basic operating conditions such as possible
switching states, CCM and discontinuous conduction mode (DCM). DCM or CCM operations are
based on switching frequency, input voltage and passive components of the converter. Depending
on the system parameters of this study, control method of the converter is developed for CCM
condition.
The converter can be analysed according to two possible states of the switch, diagrams of which
are presented in Figure 2(a,b), separately. When the switch is ON, the inductor current (iL ) can be
calculated with Equation (1). Figure 2(b) illustrates the circuit model for the OFF position of the
switch. In this case, the load is supplied from both the inductor and the input source. Thus, the
input current can be expressed with Equation (2).

Figure 1. General block diagram of the system.

Figure 2. Circuit diagram of the converter. a: The switch is ON (S = 1). b: The switch is OFF (S = 0).
4 E. IRMAK AND N. GÜLER

diL 1
¼ ðVin RL iL Þ (1)
dt L

diL 1
¼ ðVin RL iL Vout Þ (2)
dt L

where;
L, RL: Inductance and resistance values of the inductor, respectively.
Vin, Vout: Input and output voltages.
iL : Inductor current
Equations (1) and (2) can be combined as in Equation (3), which gives the input current during a
switching period.

diL 1
¼ ðVin RL iL Vout ð1 dÞÞ (3)
dt L

where;

0 for OFF position of the S
d¼
1 for ON position of the S

2.2. Discrete-time model predictive current control of the converter

MPC is a powerful control technique that predicts system behaviour according to possible control
states. The optimal next control actuation is then selected based on the predicted future system
states, in order to minimise the cost function. The cost function evaluation for each sampling
period is determined by prediction horizon (N) that can be selected depending on system type and
performance requirement. When the prediction horizon is selected greater than one, predictions
are shifted one-step and the next optimisation is performed (Metry et al., 2017).
In this study, prediction horizon is selected as one (N = 1) to determine the optimal next
switching state. Thus, a cost function that includes the future states, references and future
actuations can be defined as follows (Metry et al., 2017):

g ¼ f ðxðkÞ; uðkÞ; ::::::; uðk þ N 1ÞÞ (4)

To obtain the total cost function, the discrete-time MPC method can be applied in three steps as
following (Rodriguez & Cortes, 2012):

● Modelling the converter for all possible switching states.

● Obtaining the discrete-time model for predictions in each sampling time.
● Defining the cost functions.

The first step is presented in the previous section. To obtain the discrete-time model as the second
step, some discretisation methods are used in the literature. Among them, forward-difference Euler
approximation method given in Equation (5) is widely used in first-order systems (Rodriguez &
Cortes, 2012). By combining Equations (3) and (5), Equation (6) can be obtained to predict the next
value of the inductor current.
INTERNATIONAL JOURNAL OF ELECTRONICS 5

diL iL ðk þ 1Þ iL ðkÞ
(5)
dt Ts

TS
iL ðk þ 1Þ ¼ ðVin ðkÞ RL iL ðkÞ Vout ðkÞð1 dÞÞ þ iL ðkÞ (6)
L

where;
Ts : Sampling time
k: Iteration number
iL ðk þ 1Þ: predicted current value
At each sampling time, the optimisation problem given in Equation (6) is solved again by using
a new set of measured data to determine the new state of the switch.
The MPC method cannot determine the optimal key state only based on predictions. The
minimisation of the error between the reference and the predicted value is also required, which
is carried out using the cost function mentioned as the third step above. This is the reason why the
cost function is one of the key points of MPC-based applications.
In this study, the cost function given in Equation (7) is used. In order to determine this function
in the MPC method, the error between the reference value and the predicted value is calculated as
seen from the equation.

gi ðk þ 1Þ ¼ ½iref iL ðk þ 1Þ2 (7)

where;
iref : Reference current
The fixed-step control method that predicts the next value in each sampling period needs a
second cost function to decrease the switching frequency (Danandeh & Mousavi, 2018). Since
digital control methods do not use comparators to produce the switching signal, they may have
variable switching frequencies. If the switching frequency approximately equals to the sampling
frequency, the control error will increase. To decrease the switching frequency, the cost function in
Equation (8) is used and a weighting factor (λ) is utilised to limit the effects of it on the complete
cost function. In this study, weighting factor is determined as 0.1 by using cost function classifica-
tion method proposed in Rodriguez & Cortes (2012).

gsw ðk þ 1Þ ¼ λ ½dðkÞ dðk 1Þ2 (8)

For simultaneously controlling both the current and the switching frequency, two cost functions
are needed as mentioned above. One of the significant advantages of MPC method is its capability
of controlling multiple parameters in a single function. Thus, a complete cost function is created as
given in Equation (9) which is the combination of Equations (7) and (8). Since the use of squared
error technique in multiple terms cost functions serves better results as reported in Rodriguez &
Cortes (2012), the same approach is preferred in this study to obtain more accurate results. This is
the reason why the squares of Equations (7) and (8) are used.

gðk þ 1Þ ¼ gi ðk þ 1Þ þ gsw ðk þ 1Þ (9)

Total cost function is calculated in each sampling time for all possible switching states and then
minimised as seen in Equation (10). By using Equation (11), the d(k) value that gives the smallest
value of the cost function is calculated and the switching state is selected according to this process.

dðkÞ ¼ ½1 0arg mind g (10)

6 E. IRMAK AND N. GÜLER

min gðk þ 1Þ ¼ minðgi ðk þ 1Þ þ gsw ðk þ 1ÞÞ (11)

dðkÞ

For detailed analysis of this process, a sample control result when the converter operates with 3 A
reference value is given in Figure 3. In this figure, the cost function results are given separately for
ON (S = 1) and OFF (S = 0) positions.
As obviously seen from Figure 3, the control method decides the switch position by considering
the smaller cost function. For example, if the calculated cost function for ON position of the switch
is less than its value for OFF position, the control method selects ON state for the switch.
Furthermore, the proposed control structure does not need any average calculation method
because the cost function includes all system parameters.

2.3. Model predictive based MPP tracking control

MPPT algorithms determine the MPP by observing power parameters. Typical characteristic of a
PV array is given in Figure 4. As it is shown in Figure 4(a), the point where dP/dV = 0 can be
expressed as the MPP. Similarly, the MPP can also be obtained from IV characteristics as given in
Figure 4(b).
In order to determine the MPP, P&O algorithm is a popular approach that can be applied by
changing the output parameter. As mentioned earlier, this parameter can be the duty cycle of the
switching signal, the current value or the voltage value at MPP. Among them, the current value is
selected as the output parameter in this study and it is used as a reference in the MPC algorithm.
A complete flowchart of the proposed control structure is given in Figure 5. As seen from the
figure, the control process starts with measuring power parameters. Changes on the reference
value are determined by comparing the previous and the next power parameters. After this
procedure, the MPC algorithm starts with defining some values such as passive component
parameters, possible switching state matrix and sampling time. During a sampling step, prediction
algorithm (Equation 6) and cost function (Equation 11) are calculated for all possible switching
states. The switch position is determined according to the smaller cost function. Thus, MPC
algorithm performs the current control based on the MPPT reference.

Figure 3. Generation of switching signals according to 3 A reference current.

INTERNATIONAL JOURNAL OF ELECTRONICS 7

Figure 4. Characteristic curves of PV array.

Figure 5. Flowchart of the proposed control algorithm.

Table 1. Simulation parameters.

Parameter Value
PV Power (1000 W/m2) 470 W
Capacitors (C and Cin) 1000 µF
Inductance (L) 5 mH
Resistance of inductor (RL) 0.2Ω
Load resistance (RLOAD) 30Ω
Sampling time of MPC (Ts) 15 µs
Sampling time of MPPT 100 µs

3. Simulation results
To verify the effectiveness of the proposed control algorithm, the designed system previously given
in Figure 1 is simulated in MATLAB/Simulink platform. Simulation parameters are presented in
Table 1. Two series connected panels are used as the main energy source. PV array output is
connected to the boost converter for power flow control.
8 E. IRMAK AND N. GÜLER

Figure 6. Simulation results.

A test scenario is created with variable irradiance and constant temperature conditions as seen
in Figure 6 where the simulation starts with 400 W irradiance. The MPPT algorithm calculates the
reference value after increasing PV voltage. As clearly seen from the figure, the MPPT algorithm
changes the reference in line with the irradiance. On the other hand, the MPC algorithm provides
the current control operation simultaneously with the MPPT process.
To test the dynamic capability of the proposed algorithm, step changes are created over the
irradiance (points A and B in Figure 6) and detailed graphs for transition times are given in Figure 7.
At the first point (A), the irradiance is suddenly decreased as 25%. Accordingly, the MPPT algorithm
updates the reference current to an appropriate value as seen from Figure 7(a). Based on the new

Figure 7. Simulation results against instant transitions. a: Decreasing irradiance. b: Increasing irradiance.
INTERNATIONAL JOURNAL OF ELECTRONICS 9

reference, the MPC algorithm makes the switch position OFF. Thus, the average value of the
inductor current is decreased to the MPP current. Similar test is repeated at point B where the
irradiance is increased instantly.
According to Figure 7(b), the new MPP is determined by increasing the reference current. When
the reference increases, the MPC algorithm changes the switch state from OFF to ON. In both
situations, transitions are realised in a very short time thanks to the fast dynamic response of the
MPC algorithm.
Power graphs obtained during the control process are given in Figure 8. As seen, the proposed
control algorithm successfully tracks the maximum power curve of the PV array. Tracking capability
of the proposed method is over 99%. Moreover, the dynamic behaviour of the method is also
successful as obviously seen in zoomed parts in Figure 8 where detailed views for the transition
times (A and B) are presented.

4. Experimental studies
In order to test the performance of the proposed control method in real operational conditions, a
prototype boost converter is designed. A complete view of the experimental set is given in Figure 9.
Two series connected PV panels are used. The power value of each panel is 235 W. The output of panels
is connected to a boost converter in the laboratory.
The circuit parameters of the experimental setup are given in Table 2. Because the system
behaviour is dependent on both converter design and control software, Table 2 also includes the
sampling times of control software. The sampling time is usually selected according to the
execution time of the control software. In this study, the execution time of the proposed control
algorithm has been measured as 11 µs via the ControlDesk software. As reported in Rodriguez &
Cortes (2012), the sampling time should be greater than the execution time to ensure the
operation of the control software without any issue. Therefore, the sampling time is selected as
15 µs during the experimental studies.
In the system, dSPACE ds1104 is used as the controller device, and ControlDesk software is used
to display the measured parameters both graphically and numerically in real time. Voltage and
current sensors are used to measure the power parameters. The system evaluates the

Figure 8. Power tracking results.

10 E. IRMAK AND N. GÜLER

Figure 9. Experimental set.

Table 2. Specifications of the system elements.

Parameter Value
Capacitors (C and Cin) 1000 µF
Inductance (L) 0.5 mH
Resistance of Inductor (RL) 0.2Ω
IGBT (2MBI100N-060) 100 A–600 V
Fast recovery diode (dsei60-06a) 60 A–600 V
Load resistance (RLOAD) 41Ω
Sampling time of MPC (Ts) 15 µs
Sampling time of MPPT 100 µs

measurement results according to the control algorithm, and an appropriate switching signal is
applied from the digital IO channel of dSPACE.
As a typical characteristic of discrete-time MPC technique, the proposed control algorithm uses
variable switching frequencies. In this study, it is observed during the experimental tests that the
control system automatically varies the switching frequency from 3.125 to 16.6 kHz in accordance
with the operational conditions. When the converter is being operated with a 2 A reference value,
the inductor, IGBT and diode currents are given in Figure 10 for a detailed analysis. As obviously
seen, the inductor is charged during ON state of the switch. For OFF state, the inductor current is

Figure 10. Diode, inductor and IGBT currents.

INTERNATIONAL JOURNAL OF ELECTRONICS 11

decreased since it is in discharging mode. At the times when the ON–OFF transitions have
occurred, no distorting ripple is observed as seen from the figure.
By using a similar scenario created in simulation studies, capability of the proposed MPC-
based MPPT control technique is experimentally investigated in real-time conditions. Results
obtained under variable irradiances are given in Figure 11. In order to analyse the system
response against the transitional events separately, some crucial periods are indicated in
Figure 11 as follows:

● Periods A, C and E: The constant irradiance

● Period B: Increasing irradiance
● Period D: decreasing irradiance

4.1. Period A (constant irradiance)

As seen in Figure 11(a), the reference current (iref) is determined about 3.9 A by the MPPT
algorithm. During the first 17 s, the PV current is fixed to the reference by the MPC algorithm
with 2% error. To analyse the system parameters more accurately, Figure 12 illustrates the results in
detail. As shown in Figure 12(a), the average value of the inductor current is 3.85 A. PV voltage and
converter output voltage are 49.2 and 86.5 V, respectively. Converter output current is about 2.1 A.
Input and output powers of the converter are 192 and 182 W, respectively. Thus, these are assumed
as the initial conditions for the experimental test scenario.

4.2. Period B (increasing irradiance)

Irradiance is increased between 17 and 30 s. Consequently, the MPPT algorithm increases the
reference in this period. Although the reference is changed, the current control operation is
successfully provided by the MPC algorithm. Furthermore, small ripples on the PV voltage shown
in Figure 11(d) indicate that the current value at the MPP has been successfully determined. Since
the experimental studies are performed under constant resistive load conditions, voltage and
current of the load vary depending on input current. At the end of this period, the PV power is
increased from 192 to 300 W.

4.3. Period C (constant irradiance)

As seen from Figure 11(a,b), the irradiance is almost constant between 30 and 57 s and the reference
current is about 6 A during this period. In such a situation according to Figure 12(b), the MPC
regulates the PV current to follow the reference with the least error. As defined in Figure 4, it is an
inherent characteristic of MPPT that the panel voltage variation is not tangible when the MPP is
changing. This issue is verified in both Figures 11(d) and 12 likewise.

4.4. Period D (decreasing irradiance)

As shown in Figure 11(a), the MPPT algorithm reduces the reference current because of decrease
on the irradiance. Nevertheless, the MPC algorithm successfully provides current tracking. At the
end of this control process, the panel power is decreased from 300 to 190 W. Considering the PV
current result, it is seen that the MPC algorithm has such a sensitiveness to show a fast response
against a small change that occurs in the current reference at 68 s.
12 E. IRMAK AND N. GÜLER

Figure 11. Experimental results. a: Reference current. b: PV current. c: Load current. d: PV voltage. e: Converter output voltage.
f: PV power. g: Load power.
INTERNATIONAL JOURNAL OF ELECTRONICS 13

Figure 12. Experimental results under constant irradiance. a: Low irradiance (Period A). b: High irradiance (Period C).

4.5. Period E (constant irradiance)

For the last part of the control process, the obtained results are similar to Period A and the current
is tracking the reference successfully.
Concluding, both the steady state and the transient state studies show that MPPT algorithm
determines the MPP current and the PV current is matched to this value by the MPC algorithm.

4.6. Step change analysis

Besides the MPPT-based control, dynamic response of the MPC algorithm is also tested with step change
analysis that is realised using a 500 W DC power supply. To observe the system response against the
transient events, the reference current is changed through the Controldesk software. Results are given in
Figure 13(a) for the step up change from 2 to 4 A. As seen from this figure, the switch is kept in ON
position (180 µs) by the MPC algorithm till the inductor current reaches the maximum ripple value of the
new reference. After this transition time, the average value of inductor current is kept on 4 A.
In Figure 13(b), the reference current is decreased from 4 to 2 A to analyse the step-down
response. As seen, the control algorithm keeps the switch in the OFF position (100 µs) until the
inductor current reaches to the minimum ripple value of the reference. Thus, in both step change
tests, the MPC algorithm updates the switch status in accordance with the reference variations.

Figure 13. Step change results. a: Step up result. b: Step down result.
14 E. IRMAK AND N. GÜLER

Table 3. Comparison table.

(Valdez-Resendiz Presented
(Khosravi et al., 2017) (Singh et al., 2017) (Errouissi et al., 2016) et al., 2017) study
Control method ANN-based MPPT & Sliding mode Continuous time- Closed-loop Discrete time-
MPC MPC MPC
Dynamic response 40 ms 5 ms 20 ms 400 ms 100–180 µs
Overshoot and <5% ~0.0% <2% >50% ~0.0%
undershoot
Sampling time 10 µs 10 µs 80 µs Not specified 15 µs
Comparator usage No No Yes Yes No
Max efficiency Not specified Not specified 94% 94.14% 98%
Switching Not specified Not specified 12.5 kHz 50 kHz 3.125–16.6 kHz
frequency

The developed discrete time-MPC control methodology is compared with other techniques as given
in Table 3. For this purpose, some of the well-known control methods such as sliding mode control,
continuous time MPC, ANN-based MPPT & MPC and only closed-loop control are considered. Some
recent studies (Errouissi, Al-Durra, & Muyeen, 2016; Khosravi et al., 2017; Singh et al., 2017; Valdez-
Resendiz et al., 2017) using these techniques are compared with the presented study. As seen from
Table 3, dynamic response time varies depending on the control method. Among the compared
studies except from this study, the lowest dynamic response time is obtained in Singh et al. (2017), in
which the step change analysis has been performed by doubling and then reducing the reference
current. The same approach has been also applied during the experimental tests of the control system
proposed in this paper. As clearly seen in Figure 13 that illustrates the results of step change analysis,
dynamic responses for step up and step down are 180 and 100 µs, respectively.
Considering the studies compared in Table 3, the fast dynamic response of the presented system
originates from the fact that instead of average values MPC uses instant values measured in each
sampling time and the switch state is determined in each sampling period as well. In some systems
where the average value is used, as in (Valdez-Resendiz et al., 2017), the calculations are continued along
one switching period at least. This situation delays the control operations as well as causes overshoots
and undershoots. Therefore, using instant values instead of average value and determining the switch
state position in each sampling time play an important role in obtaining the fast dynamic response and
eliminating overshoots and undershoots. Thanks to these features, the control system not allows to
instant changes on the panel voltage and thereby the ripples on the voltage is decreased considerably.
In addition to providing a better control system, the efficiency of converters has great impor-
tance in PV systems. As seen from Figure 14 that illustrates the efficiency graph of the proposed
system, minimum efficiency is about 92% that measured when the PV power increases to upper

Figure 14. Efficiency graph.

INTERNATIONAL JOURNAL OF ELECTRONICS 15

than 340 W. The maximum efficiency is 98% that is acquired at 70 W. Thus, the proposed prototype
model offers the same efficiency with similar ones those have proposed for similar power levels in
the literature (Abu-Rub et al., 2014).

5. Conclusion
This paper presents an MPC-based dynamic MPPT algorithm for photovoltaic systems. For this
purpose, mathematical background of boost converters and MPC technique is studied to apply the
control method. Dynamic performance of the proposed method is considerably improved by using
the discrete-time MPC technique that requires neither the average calculation nor the comparator.
The control method proposed uses instantaneous values instead of average values. In order to
determine the switch state, instant values are measured at each sampling period in real time and
then the prediction algorithm calculates the current value for the next sampling time. Thus, the
calculation time and the dynamic response time are shortened considerably that results to
eliminate overshoots and undershoots. In addition, the use of a digital control technique in the
proposed method is another important point to overcome such issues. Non-digital control meth-
ods have to wait for the next carrier signal to change the switch status, whereas the digital control
methods can make this at each sampling time. Accordingly, the presented control method allows
the switch position to be changed when the predicted value reaches the reference. Therefore, the
issues of overshoot and undershoots are significantly eliminated.
Performance of both the control algorithm and the power structure is firstly tested and verified
with simulations. Step change tests verify that the proposed control method achieves the reference
tracking process in a very short time, which is less than one switching period. Furthermore, overshoots
and undershoots do not occur during the transient states. These results show that the proposed
control method has a remarkable performance in terms of dynamic response time and robustness.
Besides the dynamic performance, the system response is tested against both the constant and
variable irradiance conditions. It is observed that the control method successfully tracks the MPP in
any case. In addition to all these advantages, the efficiency analysis is investigated and it is proved
that the proposed prototype model operates up to 98% efficiency.

Disclosure statement
No potential conflict of interest was reported by the authors.

Funding
This work was supported by the Gazi University Scientific Research Projects Unit [07/2016-19].

ORCID
Erdal Irmak http://orcid.org/0000-0002-4712-6861
Naki Güler http://orcid.org/0000-0003-3882-733X

References
Abbes, H., Abid, H., & Loukil, K. (2015). An improved MPPT incremental conductance algorithm using T-S fuzzy system
for photovoltaic panel. International Journal of Renewable Energy Research, 5(1), 160–167.
Abu-Rub, H., Malinowski, M., & Al-Haddad, K. (2014). Power electronics for renewable energy systems. In:
Transportation and industrial applications. United Kingdom: John Wiley & Sons Ltd.
Adhikari, N. (2018). Design of solar photovoltaic energy generation system for off-grid applications. International
Journal of Renewable Energy Technology, 9(1/2), 198–207.
16 E. IRMAK AND N. GÜLER

Ahmed, J., & Salam, Z. (2016). A modified P&O maximum power point tracking method with reduced steady state
oscillation and improved tracking efficiency. IEEE Transactions on Sustainable Energy, 7(4), 1506–1515.
Başoğlu, M. E. (2018). An enhanced scanning-based MPPT approach for DMPPT systems. International Journal of
Electronics, 105(12), 2066–2081.
Cheng, L., Acuna, P., Aguilera, R. P., Jiang, J., Wei, S., Fletcher, J. E., & Lu, D. D. C. (2018). Model predictive control for
DC–DC boost converters with reduced-prediction horizon and constant switching frequency. IEEE Transactions on
Power Electronics, 33(10), 9064–9075.
Danandeh, M. A., & Mousavi, S. M. G. (2018). Comparative and comprehensive review of maximum power point
tracking methods for PV cells. Renewable and Sustainable Energy Reviews, 82(3), 2743–2767.
Errouissi, R., Al-Durra, A., & Muyeen, S. M. (2016). A robust continuous-time MPC of a DC–DC boost converter
interfaced with a grid-connected photovoltaic system. IEEE Journal of Photovoltaics, 6(6), 1619–1629.
Freitas, A. A. A., Tofoli, F. L., Junior, E. M. S., Daher, S., & Antunes, F. L. M. (2016). Analysis of high voltage step-up
nonisolated DC–DC boost converters. International Journal of Electronics, 103(5), 898–912.
Hernanz, J. R., Guede, J. M. L., Barambones, O., Zulueta, E., & Gamiz, U. F. (2017). Novel control algorithm for MPPT with
boost converters in photovoltaic systems. International Journal of Hydrogen Energy, 42(28), 17831–17855.
Kandemir, E., Borekci, S., & Cetin, N. S. (2018). Comparative analysis of reduced-rule compressed fuzzy logic control
and incremental conductance MPPT methods. Journal of Electronic Materials, 47(8), 4463–4474.
Khosravi, M., Heshmatian, S., Khaburi, D. A., Garcia, C., & Rodriguez, J. (2017, April). A novel hybrid model-based MPPT
algorithm based on artificial neural networks for photovoltaic applications. IEEE Southern Power Electronics
Conference (SPEC), Puerto Varas, Chile.
Kim, J. C., Moon, S. K., & Kwak, S. (2018). Direct model-based predictive control scheme without cost function for
voltage source inverters with reduced common-mode voltage. International Journal of Electronics, 105(4), 629–644.
Kim, S. K., Park, C. R., Kim, J. S., & Lee, Y. I. (2014). A stabilizing model predictive controller for voltage regulation of a
DC/DC boost converter. IEEE Transactions on Control Systems Technology, 22(5), 2016–2023.
Lashab, A., Sera, D., Guerrero, J. M., Mathe, L., & Bouzid, A. (2018). Discrete model-predictive-control-based maximum
power point tracking for PV systems: Overview and evaluation. IEEE Transactions on Power Electronics, 33(8), 7273–
7287.
Mahendran, V., & Ramabadran, R. (2016). Fuzzy-PI-based centralised control of semi-isolated FP-SEPIC/ZETA BDC in a
PV/battery hybrid system. International Journal of Electronics, 103(11), 1909–1927.
Metry, M., Bayhan, S., Balog, R. S., & Rub, H. A. (2016, February). Model predictive control for PV maximum power point
tracking of single-phase submultilevel inverter. IEEE Power and Energy Conference at Illinois (pp. 1–8). Urbana, IL.
Metry, M., Shadmand, M. B., Balog, R. S., & Abu-Rub, H. (2017). MPPT of photovoltaic systems using sensorless current-
based model predictive control. IEEE Transactions on Industry Applications, 53(2), 1157–1167.
Mosa, M., Shadmand, M. B., Balog, R. S., & Abu-Rub, H. (2017). Efficient maximum power point tracking using model
predictive control for photovoltaic systems under dynamic weather condition. IET Renewable Power Generation, 11
(11), 1401–1409.
Öztürk, N., & Çelik, E. (2014). An educational tool for the genetic algorithm-based fuzzy logic controller of a permanent
magnet synchronous motor drive. International Journal of Electrical Engineering Education, 51(3), 218–223.
Priya, T. H., & Parimi, A. M. (2016). Design of adaptive perturb and observe-fuzzy MPPT controller for high voltage gain
multilevel boost converter. IEEE 7th Power India International Conference (PIICON), Bikaner.
Qian, Q., Yu, J., Su, C., Sun, W., & Lu, S. (2017). A LLC resonant converter with dual resonant frequency for high light
load efficiency. International Journal of Electronics, 104(12), 2033–2047.
Ren, H. P., Zheng, M. M., & Li, J. (2015 October). A simplified mixed logical dynamic model and model predictive
control of boost converter with current reference compensator. IEEE 24th International Symposium on Industrial
Electronics (pp. 61–65), Buzios, Brazil.
Rodriguez, J., & Cortes, P. (2012). Predictive control of power converters and electrical drives. United Kingdom: John
Wiley & Sons.
Seo, S. W., Kim, Y., & Choi, H. H. (2017). Model predictive controller design for boost DC–DC converter using T–S fuzzy
cost function. International Journal of Electronics, 104(11), 1838–1853.
Singh, S., Fulwani, D., & Kumar, V. (2017). Emulating DC constant power load: A robust sliding mode control approach.
International Journal of Electronics, 104(9), 1447–1464.
Valdez-Resendiz, J. E., Sanchez, V. M., Rosas-Caro, J. C., Mayo-Maldonado, J. C., Sierra, J. M., & Barbosa, R. (2017).
Continuous input-current buck-boost DC-DC converter for PEM fuel cell applications. International Journal of
Hydrogen Energy, 42(51), 30389–30399.

View publication stats