A Fast and Efficient MPPT Technique For PV

Systems Using Divide and Conquer

Elias SALEH Nabil KARAMI
Saint-Joseph University Higher Colleges of Technology
ESIB, Faculty of Engineering Faculty of Engineering Technology and Science
2023 Advances in Science and Engineering Technology International Conferences (ASET) | 978-1-6654-5474-2/23/$31.00 ©2023 IEEE | DOI: 10.1109/ASET56582.2023.10180727

Beirut, Lebanon Dubai, United Arab Emirates

Email: [email protected] Email: [email protected]

Abstract—With the growing use of PV generators worldwide, are based on tracking an estimated MPP; they rely on a math-
researchers are focusing on extracting the maximum available ematical approximated model. On the other hand, heuristic
power from solar panels using maximum power point track- methods are practical methods relying on nonlinear control,
ing (MPPT) algorithms. Many MPPT methods are found in
literature. In this paper, a novel MPPT technique is proposed, such as fuzzy logic control tracking [3], [8]. Alternatively,
using a Divide and Conquer search algorithm. The proposed search algorithms seek the actual MPP rather than an estimate.
MPPT method is simulated through Matlab-Simulink under They find the MPPT by varying the PV module’s voltage
different critical conditions. A whole PV system is designed (Boost while measuring and comparing the obtained corresponding
converter, PID voltage control loop and a Divide and Conquer powers. Some examples of search algorithms include P&O
algorithm). The obtained results are compared to those of the
popular Perturb and Observe (P&O) tracking method. methods, incremental conductance method [3], [8], golden
Keywords—PV, MPPT, Divide and Conquer search, Boost section search [9], dividing rectangles algorithm [10] and
converter, PID voltage control the Fibonacci search methods [7]. The search algorithms
do not require any of the PV panel’s characteristics, which
I. I NTRODUCTION makes them very flexible in practice and are generally fast in
Due to rising demand on energy worldwide, depletion of response. In spite of that, their major challenge is steady state
fossil fuels, problems associated with the structure of the clas- oscillations around the MPP, as they run continuously.
sical energy production and distribution, rising prices of fuel in The objective of this paper is to find the MPP using
conventional energy, global warming and climate changes, the the divide and conquer search algorithm. The idea is to
renewable energy capacity has reached an estimated 2,350,755 use the PV module’s voltage as the searching interval. The
megawatts (MW) worldwide in the end of 2018 [1]. Amongst searching interval will be divided into smaller intervals at each
all the renewable energies solar photovoltaic (PV) owns an iteration in order to track the MPP. This method consists of
important share. In fact, the capacity of PV has reached minimal searching, providing a very fast-tracking time, with
480,357 MW in 2018 [1]. no oscillations around the MPP once found. The proposed
The PV module power output varies depending on the method does not require irradiance or temperature sensors, yet
amount of falling solar irradiance, the temperature of the it is highly efficient and deals with critically changing weather
solar cells and the connected load. Owing to semiconductors’ and load conditions. The system will only need current and
nonlinear nature, the current voltage (I-V) relationship of a voltage sensors mounted on the PV module. We will use a
PV cell is nonlinear. As a result, the power curve of a solar ratio of 1/3 for the dividing strategy, as equal divisions of the
module is not linear. This power curve has a unique point searched range provide the best convergence time. In addition,
corresponding to maximum current and voltage, therefore cre- the standard searching space of (0-VOC) will be shrunk from
ating a point where the power is maximal. Hence, researchers the left side, as the MPP is always after 0.5VOC. The obtained
developed multiple methods for finding the maximum power simulation results will be compared to an optimized P&O, as
point (MPP). it remains for today, one of the most practical and simple
Numerous maximum power point tracking (MPPT) algo- methods. The interface between the PV panel and the load is
rithms and designs are available in literature such as Perturb done through a DC/DC power converter (Figure 1).
and Observe (P&O) [2], P&O with adaptive perturbation
step [3], incremental conductance [2], constant voltage [4], II. PV M ODELING
look-up table [5], fuzzy logic control, differentiation, neural
network [3]. Each tracking method has its own specifications, A photovoltaic cell (PV cell) is formed of two doped
limitations and applications. semiconductors, one is P-type and the other is N-type. On
The main MPPT methods can be classified as estimation contact, they form a PN junction. Interconnecting these junc-
methods, heuristic methods and search algorithms [6]. For tions in series and parallel creates a PV panel or module. The
instance, system modelling and curve fitting techniques [7] equivalent circuit of a PV cell is displayed in Figure 2.

Authorized licensed use limited to: UNIVERSIDADE FEDERAL DE OURO PRETO. Downloaded on January 30,2025 at 04:00:11 UTC from IEEE Xplore. Restrictions apply.
 As the output power varies depending on the irradi-
ance and the temperature of the solar cells, Figure 3
shows the current-voltage (I-V) and power-voltage (P-V)
curves of the module in terms of multiple irradiance values
(1000, 800, 600, 400, 200 W/m2 ) at ambient temperature
(25◦C).

TABLE I
S OLAR CELL S IMULINK MODEL

Fig. 1. Scheme of the proposed MPPT

Fig. 2. Practical PV cell circuit model

Thus, the PV current is given by:

IP V = IP h − Id − ISh (1)

Where,
G
IP h = [Isc,ref + µIsc (T − Tref )] (2)
Gref
Vd
Id = Is (e AvT
− 1) (3)
Fig. 3. (I-V) and (P-V) characteristics under changing irradiance
VP V + I P V RS
ISh = (4)
RSh III. B OOST C ONVERTER D ESIGN
2 ◦ A DC/DC boost converter consists of a filter capacitor at
where, G (W/m ) is the solar irradiance, T ( K) is the
temperature of the cell, Gref (W/m2 ) is the illumination the input Cf (optional), an inductor L, a controllable switch
reference, Tref (◦ K) is the reference temperature, Isc,ref (A) (MOSFET or IGBT), a diode, and a DC filter at the output
is the PV module short circuit current, µIsc (A/◦ K) is the CDC (Figure 4).
short circuit current temperature coefficient, Is (A) is the cell
Vout T 1
reverse saturation current, Vd (V ) is the voltage across the = = (5)
diode, A is the ideality factor, vT (V ) is the thermal voltage Vin tof f 1−D
equivalent, VP V (V ) is the PV voltage, IP V (A) is the PV where, Vout is the average output voltage, Vin is the input
current, RS (Ω) is the PV series resistor, RSh (Ω) is the PV voltage, T is the switching period, D is the duty cycle, tof f
shunt resistor. is the OFF time during one switching period.
Table I shows the parameters and the specifications of the The gate of the MOSFET is controlled by a pulse width
chosen PV module, extracted from the datasheet. modulated signal (PWM). Depending on the application, the

Authorized licensed use limited to: UNIVERSIDADE FEDERAL DE OURO PRETO. Downloaded on January 30,2025 at 04:00:11 UTC from IEEE Xplore. Restrictions apply.
 TABLE II
PARAMETERS OF THE BOOST CONVERTER

Fig. 4. DC/DC boost converter TABLE III

S PECIFICATIONS OF THE MOSFET

switching frequency can range between 10 kHz and 50 kHz.

We chose a switching frequency of 20 kHz which is a good
compromise between the size of the converter (size of the
capacitors) and the switching losses The maximum power
point tracker uses the boost converter to adjust the PV voltage
VP V to reach the MPP at VM P P . intervals. Hence, at each iteration, [X, Y ] will be divided
Concerning inductor selection, the higher the inductor value, into three equal sub-intervals using a 1/3 ratio. Consequently,
the higher is the maximum output current because of the during each iteration, we will use two checkpoints V1 and V2
reduced ripple current [11]. The inductor size can be estimated inside [X, Y ], the power is to be measured at these points
using the below equation: (i.e. P1 = P (V1 ) and P2 = P (V2 )) and compared:
Vin × (Vout − Vin ) 1) If (P1 < P2 ), then the extremum is inside [V 1, Y ].
L= (6) The searching space is updated accordingly. It is shrunk
∆IL × fs × Vout
from the left side in this case (see Figure 5), where:
where, L (H) is the inductance, Vin (V ) is the typical voltage [X, Y ] = [0, VOC ] initially.
input, Vout (V ) is the desired voltage output, fs (Hz) is the 2) If (P1 > P2 ), then the extremum is inside [X, V2 ]. The
switching frequency of the converter, ∆IL (A) is the estimated searching space is updated accordingly. Conversely, it is
ripple current. A good estimation for the ripple current is 20% shrunk from the right side.
to 40% of the output current.
We chose a Schottky diode in order to minimize losses.
Consulting datasheets, we used in the simulation a forward
voltage Vf = 0.45 V with a resistance Ron = 0.001 Ω.
Regarding capacitor selection, on one hand, the input ca-
pacitor is chosen to stabilize the input voltage due to current
peaks requirements when switching. On the other hand, the
output capacitor’s minimum estimated value is giver by [11]:
Iout (max) × D
CDC (min) = (7)
fs × ∆Vout
where, CDC (min) (F ) is the minimum capacitance,
Iout (max) (A) is the maximum output current of the appli-
cation, D is the duty cycle between 0 and 1, ∆Vout (V ) is Fig. 5. D&C concept
the desired output voltage ripple, fs (Hz) is the switching
frequency of the converter. In order to achieve fast convergence, the searching range
Taking all the above into consideration, along with the is divided using a ratio of 1/3 during each iteration; resulting
system’s response time, Table II displays the parameters of the in equal search range eliminations at the left and right sides.
designed boost converter. Table III shows the specifications of Therefore, V1 and V2 are computed using the below equations:
the MOSFET used in the simulation.
V1 = X + 0.33(Y − X) (8)
IV. P ROPOSED MPPT
A. Concept V2 = Y − 0.33(Y − X) (9)
The proposed MPPT method consists of finding the MPP
B. Flowchart
by using a divide and conquer (D&C) search algorithm. Thus,
let [X, Y ] be our searching space, the algorithm will search Applying the above-mentioned search method yields very
for the MPP inside [X, Y ], dividing it into equal smaller small steady state oscillations around the MPP. The proposed

Authorized licensed use limited to: UNIVERSIDADE FEDERAL DE OURO PRETO. Downloaded on January 30,2025 at 04:00:11 UTC from IEEE Xplore. Restrictions apply.
 flowchart solves this issue by comparing the difference be-
tween the measured power values P1 and P2 to a fixed error
as E:

|P1 − P2 | < E (10)

For the best efficiency/response time in the designed MPPT,

we took E = 0.00015. When the condition of Equation (10) is
satisfied, the searching process is stopped. The MPP is reached
at VM P P given by Equation (11):

V1 + V 2
VM P P = (11)
2
Moreover, the algorithm is optimized by using a smaller
search range. In fact, by looking at the PV module charac-
teristics in Figures 3 it is graphically evident that VM P P is
always after 23 VOC . Hence, the initial searching range in the
flowchart is:

2
[X, Y ] = [ V OC , VOC ] (12)
3
Furthermore, for each of the calculated checkpoints V1 and
V2 , the algorithm needs to wait 1 ms; this delay is required
for the electrical system to respond to the reference voltage
(V1 and V2 ) that are imposed by the tracker. A voltage control
loop will be designed for this purpose in the next section. The
flowchart is presented in Figure 6. The input of the algorithm
is the measured power PP V and its output is the reference
voltage Vref . Fig. 6. Flowchart of the proposed D&C method
Once the MPP is reached, the power is measured and
stored in memory as Pmax . The searching process is stopped.
However, PP V will remain monitored and it will always 
 P = 0.6 × Kcr = 0.6 × 83 = 50
be compared with Pmax . Subsequently, when the difference I = 0.5 × Pcr = 0.5 × 0.002396 = 0.0012 (14)
between Pmax and PP V gets relatively important (2%), the 
D = 0.125 × Pcr = 0.125 × 0.002396 = 0.0003
algorithm is restarted and the searching range is re-initialized,
as this power difference is interpreted as an irradiance change.
This approach eliminates the need for an irradiance sensor.

C. PID voltage control loop

In order for the system to keep working under variable
loads, voltage control using a feedback loop is necessary.
Voltage control requires a PI or PID controller, which forces
the system to keep operating at the given Vref [9]. A PID
controller has three parameters KP , KI and KD . We will in-
corporate a PID controller in the proposed system, as they are
simple and have excellent performance in many applications.
The chosen PID model in Matlab is given by:

 
1 1 z−1
U (z) = P 1 + I Ts +D (13)
z−1 Ts z
Fig. 7. Response of the PID voltage control loop at Vref = 18 V
The second method of Ziegler-Nichols is used for tuning
the PID parameters. In accordance with simulation, Kcr = 83 Figure 7 displays the response of the system with and
and Pcr = 2.396 ms. Thus, the PID parameters are concluded without PID control. Figure 8 shows the whole architecture
as: of the proposed MPPT.

Authorized licensed use limited to: UNIVERSIDADE FEDERAL DE OURO PRETO. Downloaded on January 30,2025 at 04:00:11 UTC from IEEE Xplore. Restrictions apply.
 Fig. 8. Scheme of the proposed PV system in Simulink

V. S IMULATION RESULTS
A. Simulation at STC
We start by showing the results of the P&O and D&C Fig. 10. D&C at STC: PP V , VP V and IP V
algorithms under standard testing conditions (STC).

TABLE IV
P&O AND D&C COMPARISON AT STC

Fig. 9. P&O at STC: PP V , VP V and IP V

Figures 9 and 10 present the performance of the P&O

and the proposed D&C algorithm. The obtained results are
compared in Table IV.
The dynamics used for tracking the MPP in the D&C
algorithm are shown below in Figure 11: where we see the
checkpoints V1 and V2 shrinking at each iteration, as the
searching space gets narrowed.
B. Simulation under critical conditions
We simulate both algorithms under critically changing irra-
diances while the load varies. The load begins as R = 15 Ω,
becomes R = 65 Ω at t = 0.5 s and gets switched back to
R = 15 Ω at t = 1.2 s. Regarding the irradiance, the values Fig. 11. D&C dynamics: V1 , V2
considered are 1000 – 500 – 300 – 700 and 800 W/m2 .

Authorized licensed use limited to: UNIVERSIDADE FEDERAL DE OURO PRETO. Downloaded on January 30,2025 at 04:00:11 UTC from IEEE Xplore. Restrictions apply.
 TABLE V
C OMPARISON BETWEEN D&C AND G OLDEN SECTION METHOD [9]

R EFERENCES
[1] R. C. Statistics, “International renewable energy agency (irena): Abu
dhabi,” United Arab Emirates, p. 60, 2019.
[2] M. A. Husain, A. Tariq, S. Hameed, M. S. B. Arif, and A. Jain,
“Comparative assessment of maximum power point tracking procedures
for photovoltaic systems,” Green Energy & Environment, vol. 2, no. 1,
pp. 5–17, 2017.
[3] N. Karami, N. Moubayed, and R. Outbib, “General review and classifi-
cation of different mppt techniques,” Renewable and Sustainable Energy
Reviews, vol. 68, pp. 1–18, 2017.
Fig. 12. P&O under critical conditions: PP V , VP V , IP V [4] R. B. Koad and A. F. Zobaa, “Comparison study of five maximum power
point tracking techniques for photovoltaic energy systems,” 2014.
[5] A. Ilyas, M. R. Khan, and M. Ayyub, “Lookup table based modeling
and simulation of solar photovoltaic system,” in 2015 Annual IEEE India
Conference (INDICON). IEEE, 2015, pp. 1–6.
[6] A. M. Farayola, A. N. Hasan, and A. Ali, “Curve fitting polynomial
technique compared to anfis technique for maximum power point
tracking,” in 2017 8th International Renewable Energy Congress (IREC).
IEEE, 2017, pp. 1–6.
[7] J.-H. Zhang, X.-Y. Wei, L. Hu, and J.-G. Ma, “A mppt method based on
improved fibonacci search photovoltaic array,” Tehnički vjesnik, vol. 26,
no. 1, pp. 163–170, 2019.
[8] R. Sankar, S. Velladurai, R. Rajarajan, and J. A. Thulasi, “Ii. pv system
description: Maximum power extraction in pv system using fuzzy logic
and dual mppt control,” in 2017 International Conference on Energy,
Communication, Data Analytics and Soft Computing (ICECDS). IEEE,
2017, pp. 3764–3769.
[9] A. Kheldoun, R. Bradai, R. Boukenoui, and A. Mellit, “A new golden
section method-based maximum power point tracking algorithm for
photovoltaic systems,” Energy Conversion and Management, vol. 111,
pp. 125–136, 2016.
[10] T. L. Nguyen and K.-S. Low, “A global maximum power point tracking
scheme employing direct search algorithm for photovoltaic systems,”
IEEE transactions on Industrial Electronics, vol. 57, no. 10, pp. 3456–
3467, 2010.
[11] B. Hauke, “Application report: Basic calculation of a boost converter’s
power stage,” Dallas: Texas Instruments, 2014.
Fig. 13. D&C under critical conditions: PP V , VP V , IP V

VI. C ONCLUSION
The simulation results for both methods were shown (P&O
and the proposed D&C), under various changing conditions.
The proposed system performed better in terms of speed
of tracking, static error, efficiency, steady state oscillations
around the MPPT and the voltage ripple across the supplied
load was considerably small.
Lastly, comparing the D&C method with the golden section
method simulated in [9], the D&C has a slightly faster
response time and efficiency (see Table V). However, it is
worth mentioning that the switching frequency used in [9]
is relatively high, which leads to using smaller capacitor and
inductor in the DC converter, resulting in a faster response. We
used a lower switching frequency for practical implementation.

Authorized licensed use limited to: UNIVERSIDADE FEDERAL DE OURO PRETO. Downloaded on January 30,2025 at 04:00:11 UTC from IEEE Xplore. Restrictions apply.