You are on page 1of 11

IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 53, NO.

2, MARCH/APRIL 2017 1157

MPPT of Photovoltaic Systems Using Sensorless


Current-Based Model Predictive Control
Morcos Metry, Student Member, IEEE, Mohammad B. Shadmand, Member, IEEE,
Robert S. Balog, Senior Member, IEEE, and Haitham Abu-Rub, Senior Member, IEEE

Abstract—Variability in the solar irradiance level and ambient energy shining on a PV device that is converted into usable elec-
temperature of photovoltaic (PV) systems necessitates the use of tricity” [3]. The energy generated from PV systems is highly
maximum power point tracking (MPPT) of PV systems to ensure dependent on environmental factors such as solar irradiation,
continuous harvesting of maximum power. This paper presents a
sensorless current (SC) MPPT algorithm using model predictive cloud coverage, wavelength, reflection, and ambient tempera-
control (MPC). The main contribution of this paper is the use of ture. These factors can negatively affect the PV cell conversion
model-based predictive control principle to eliminate the current efficiency, making it an obstacle for widespread use. Dynami-
sensor that is usually required for well-known MPPT techniques cally changing environmental conditions necessitate maximum
such as perturb and observe (P&O). By predicting the PV system power point tracking (MPPT) to ensure that maximum available
states in horizon of time, the proposed method becomes an elegant,
embedded controller that allows faster response and lower power energy is harvested from the solar module [4].
ripple in steady state than the conventional P&O technique under The MPPT controller is commonly implemented in the dc/dc
rapidly changing atmospheric conditions. This becomes possible converter input stage of the solar inverter [5]. An MPPT con-
without requiring expensive sensing and communications equip- troller and power converter can efficiently track and convert the
ment and networks for direct measurement of solar irradiation maximum available solar energy to increase the overall conver-
changes. The performance of the proposed SC-MPC-MPPT with
reduced load sensitivity is evaluated on the basis of industrial Euro- sion efficiency of the PV, without relying on an improvement in
pean Efficiency Test, EN 50530, that assesses the performance of PV the solar cell itself.
systems under dynamic environmental conditions. The proposed Many MPPT methods have been suggested recently; the rela-
control technique is implemented experimentally using dSPACE tive merits of these various approaches are discussed in [6] and
DS1007 platform to verify the simulation results. [7]. Several efforts have been made throughout the literature to
Index Terms—Maximum power point tracking (MPPT), tackle the challenges associated with conventional MPPT tech-
photovoltaic systems, predictive control, solar energy, solar power
generation. niques, such as perturb and observe (P&O) and incremental con-
ductance. Some approaches have used simple algorithms that
I. INTRODUCTION focus on the conventional techniques themselves. For example,
OLAR photovoltaic (PV) systems are one of the fastest some authors proposed the use of a parameter-based estimation
S growing renewable energy generation systems with annual
growth of more than 40% from 2010 to 2014 [1]. Global aver-
approach to improve the performance of the P&O MPPT as in
[8]. Authors also proposed the use of variable step perturbation
age costs for utility scale solar PV declined by two-thirds from on P&O MPPT to improve the step response and reduce the rip-
2010 to 2015, and is expected to decline by additional quarter ple of the PV system [9]. Other authors have proposed the use
from 2015 to 2020 [2]. The U.S. Department of Energy defines of multisampling P&O to improve the effectiveness of the PV
conversion efficiency of a PV cell as “the percentage of the solar system to fast irradiance and temperature changes [10]. A differ-
ent approach common in the literature is to couple conventional
Manuscript received March 1, 2016; revised May 30, 2016 and August 15, MPPT techniques with other advanced algorithms and control
2016; accepted September 27, 2016. Date of publication October 31, 2016; date techniques to improve upon the conversion efficiency and the
of current version March 17, 2017. Paper 2016-IPCC-0145.R2, presented at
the 2015 IEEE Energy Conversion Congress and Exposition, Montreal, QC, robustness of the PV system. For example, some authors pro-
Canada, Sep. 20–24, and approved for publication in the IEEE TRANSACTIONS posed the use of a 1-D Newton Raphson method to shorten the
ON INDUSTRY APPLICATIONS by the Industrial Power Converter Committee of
calculation time required to achieve MPP [11]. Other authors
the IEEE Industry Applications Society. This publication was made possible by
NPRP Grant 7-299-2-124 from the Qatar National Research Fund (a member of proposed the use of fuzzy logic controllers for improving track-
Qatar Foundation). The statements made herein are solely the responsibility of ing accuracy and power transfer of PV systems [12]. Efforts
the authors. The work of M. Metry was supported by Qatar Research Leadership have also been made using model predictive control (MPC) to
Program under Qatar Foundation Research and Development.
M. Metry, M. B. Shadmand, and R. S. Balog are with the Renewable Energy improve tracking accuracy and reduce conversion settling time
and Advanced Power, Electronics Research Laboratory,Texas A&M University, by using the predictive nature of MPC [13]. Further efforts sug-
College Station, TX 77843 USA (e-mail: morcos.m.metry@ieee.org; mohamad- gested that a modulated MPC improves the tracking efficiency
shadmand@gmail.com; robert.balog@ieee.org).
H. Abu-Rub is with Texas A&M University at Qatar, Doha, Qatar (e-mail: as in [14]. Different efforts combine conventional MPPT tech-
haitham.abu-rub@qatar.tamu.edu). niques together such as in [15], where P&O is combined with
Color versions of one or more of the figures in this paper are available online fractional short-circuit MPPT to improve the power harvest and
at http://ieeexplore.ieee.org.
Digital Object Identifier 10.1109/TIA.2016.2623283 reduce oscillation.

0093-9994 © 2016 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See http://www.ieee.org/publications standards/publications/rights/index.html for more information.
1158 IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 53, NO. 2, MARCH/APRIL 2017

The aforementioned MPPT algorithms, whether conventional


or improved, require sensing the PV voltage and current, which
necessitates both voltage and current sensors. The presence of
both sensors increases the cost of the system and makes it bulkier
[16], especially when working on cascaded systems. Current
sensors in particular are more prone to magnetic interference
[17], hence, hindering the measurement accuracy. Moreover,
having less components in the system increases its reliability
[18]. This paper investigates an MPPT technique with high ef-
fectiveness using model-based predictive control with an elim-
inated current sensor. Since the essence of sensorless current
control is to use a surrogate signal for the input current without
actually measuring it, the concept fits well with MPC in that
both techniques rely on models of the converter to perform the
control actions.
Several approaches have been proposed in the literature to
achieve a sensorless current mode (SCM) for a variety of control
objectives. SCM is used for voltage control of a dc/dc converter
in discontinuous conduction mode using charge balance princi-
ple and current error compensation as in [19]. SCM is specifi-
cally applied to achieve sensorless current MPPT (SC-MPPT). Fig. 1. Flyback converter with snubber circuit for PV application.
For example, an approach to SCM is to use an observer-based
model as a surrogate to the current signal as in [17]. Another in Fig. 1, on the already developed MPC-MPPT method [28]–
approach is to use capacitor current ripple to estimate the current [31]. This paper is organized as follows: Section II introduces
as in [20] and [21]. A third approach is to use transcendental the principle of MPC. Section III derives the SC-MPC-MPPT
PV cell relations to determine the MPP using a single voltage algorithm. The detail of the European Efficiency, EN 50530,
sensor as in [22]. Using power estimation based on the switch- implemented on this system is explained in Section IV. Finally,
ing information of a hysteresis voltage regulator MPP can be results and discussions are provided in Section V.
tracked, in a PV system, without a current sensor as in [23].
Overall efficiency of PV systems is commonly evaluated II. PRINCIPLE OF MPC
based on three main factors: the conversion efficiency of the
PV module which is approximately 10%–15% [24] for residen- MPC applications in power electronics can be found in litera-
tial applications, the efficiency of the dc/dc conversion stage ture from as early as the 1980s for high-power systems with low
which is between 92% and 98% [25], and the control effective- switching frequency [25]. Higher switching frequency devices
ness of the MPPT technique which is commonly around 98% were not viable at that time due to the immense calculation time
[9] in conventional MPPT methods. In order to increase the required for the control algorithm. However, powerful micropro-
efficiency of a PV module, advanced fabrication is needed, in- cessors, which were later developed, have fueled considerable
creasing the cost of the PV module. Improving dc/dc converter interest in the application of MPC in power electronics over the
design and topologies may increase the overall energy capture last decade [32], [33].
from the PV module but will increase the overall cost of the Finite set MPC is a powerful control technique that predicts
PV system too. Improving the efficiency and the performance the future behavior of the system. This relies on an arbitrary
of the third factor, MPPT method, is less expensive compared number of sampling steps into the time horizon selected based
to the two aforementioned factors, and it can be implemented on a set of possible control actuation. The optimal next control
in already existing PV power plants. Thus, the main motiva- actuation is then selected based on the predicted future system
tion of this work is to improve the performance of the MPPT states, in order to minimize a desired cost function. This class
method for the PV system while maintaining and even reducing of controller is ideally suited for the optimal operation control
the overall installed PV system cost by eliminating the current problem studied in this paper. One step (N = 1) prediction is
sensors. To better evaluate the performance of the proposed sen- used in this paper, meaning MPC predicts the system behavior
sorless current MPC MPPT (SC-MPC-MPPT) on the specified at the next sampling interval, x̃(k + N ), N = 1.
criteria, an industry approved test, prEN 50530, is used [26]. However, based on the application and performance require-
The proposed SC-MPC-MPPT has demonstrated an improved ment, more steps into the time horizon can be considered. The
overall efficiency over conventional P&O MPPT as evidenced predicted system control response is used to determine the op-
from experimental results. timal next switching state by minimizing a cost function. The
The main characteristic of the MPC technique is to predict discrete time model of the control variables used for prediction
the error one step ahead in the horizon of time. This property can be presented as a state-space model
can be used to obtain a better estimated current in the SC- x̃(k + 1) = Ax(k) + Bu(k) (1)
MPC-MPPT algorithm [27]. The main contribution of this pa-
per is to introduce a sensorless current technique, as illustrated y(k) = Cx(k). (2)
METRY et al.: MPPT OF PHOTOVOLTAIC SYSTEMS USING SENSORLESS CURRENT-BASED MODEL PREDICTIVE CONTROL 1159

At each sampling time, the optimization problem is solved


again by using a new set of measured data to obtain a new
sequence of optimal actuation. The general scheme of MPC
for power electronics converters is illustrated in Fig. 3. In this
block diagram, measured variables, X(k), are used in the model
to estimate predictions, X̃(k + 1), of the controlled variables
for all the possible switching states σ ∈ {1 : m} named plant
in Fig. 3. These predictions are then evaluated using a cost
function which compares them to the reference values, X ∗ (k +
1), by considering the design constraints. Finally, the optimal
actuation, σ, is selected and applied to the converter. The general
form of the cost function, g, subject to minimization can be
formulated as follows:
 
 σ ∈{1:m } 
min g = λ1 X̃1 (k + 1) − X1∗ (k + 1)
 
 σ ∈{1:m } 
+λ2 X̃2 (k + 1) − X2∗ (k + 1)
 
 
+ · · · + λn X̃nσ ∈{1:m } (k + 1) − Xn∗ (k + 1)

subject to x(k + 1) = Ax(k) + Bu(k)


y(k) = Cx(k)

Fig. 2. Moving in horizon of time principle (N = 3). |y(k)| ≤ yb oundary (5)


where λ is the weight factor for each objective. This technique
A cost function that takes into consideration the future states, can be applied to any power converter topology.
references, and future actuations can then be defined as follows:
III. PROPOSED SC-MPPT USING MPC
g = f (x (k) , u (k) , . . . , u (k + N − 1)) . (3)
The relation between the magnetizing inductor current iLm ,
The defined cost function g is minimized for a predefined step and the PV voltage vPV is used to develop an observer for the
on the time horizon N; a sequence of N optimal actuations, as the PV-side current iPV . The flyback converter in Fig. 4 is analyzed
controller output u (k), will be determined where the controller in continuous conduction mode during two states: switch “ON”
only applies the first element of the sequence. Thus, the first (σ = 1) and “OFF” (σ = 0)
control signal u (k) is sent to the process while the next control
dvPV (t)
signals calculated are rejected [34]. This is because the output iσLM
=1
(t) = iPV (t) − CIN Switch on
is already known at the next sampling state. In this paper, the dt
 
elements of the vector u (k) represent the determined switching dvC (t) vC (t)
iσLM
=0
(t) = C + n Switch off. (6)
states (plants) that minimize g as follows: dt R
 
u(k) = 1 0 · · · 0 arg min g. (4) Using node equations to obtain a relation between iPV and
u
iLm
Thus, the controller output u(k) is the switching signal (either dvPV (t)
switch “ON (σ = 1)” or “OFF (σ = 0)”) that will be applied to iPV (t) = CIN + iLm (t). (7)
dt
the converter.
At the next sampling (k + 1), the system states are calculated By using (6) and (7), the average iPV is given by
(predicted) using the system model, the horizon is shifted by  
dvC (t) vC (t) dvPV (t)
one step, and another optimization is applied. As illustrated iPV (t) = C + n + CIN . (8)
dt R dt
in Fig. 2, for a horizon length of N = 3, the horizon taken
into consideration, the minimization of g slides forward as k The discrete-time version of (8) in steady state is found using
increases [35]. An open-loop optimal control scheme makes the Euler forward method for discretization:
use of the optimum open-loop initial decision at each stage, then nC
incorporates feedback in the observation of the obtained actual iPV (k) = (vc (k + 1) − vc (k))
TS
state [36]. The MPC scheme shown in Fig. 3 determines the
CIN n
optimal actuation using the system model in open loop, and uses + (vPV (k + 1) − vPV (k)) + vc (k) (9)
the feedback measurements to adjust the system accordingly. TS R
Thus, MPC amounts to an open-loop-optimal feedback control where D is the duty ratio, and Ts is the sampling period of
methodology. the MPC microprocessor. Therefore, (9) is used as an observer
1160 IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 53, NO. 2, MARCH/APRIL 2017

Fig. 3. General MPC schematic for power electronics converters.

for the PV and the output capacitor voltages. Model relations for
the load value can be inferred using the same flyback converter
model (see Fig. 4). The relation between PV current and output
current is given by
 
1−D 1
io = iPV . (12)
D n
The resistive load observer model is given by
Fig. 4. Analysis of the flyback converter to obtain PV current surrogate equa-
vc vc (k)
tion.
γ = R(k) = =
1 (13)
io iPV (k) 1−DD n
model for PV current to eliminate the current sensor. The im-
which can be used to improve the robustness of the system to
plementation of this algorithm is shown in Fig. 5.
load parameter mismatch of the system and is applied to the
By deriving the set of discrete-time equations, the behavior
predictive model of the system in (10) and (11).
of the control variable can be predicted at next sampling time
After determination of the next step reference voltage using
k + 1. The proposed methodology is based on the fact that the
the procedure shown in Fig. 5, the cost function subject to
slope of the PV array power curve is zero at the predicted MPP,
minimization can be obtained as follows:
positive on the left and negative on the right of the predicted  
 σ ∈{0,1} 
MPP. min g = ṽPV (k + 1) − vref (k)
The predicted PV voltage for the flyback converter in Fig. 4     
at next sampling time (k + 1) is given by (10) when the switch σ=1 TS 1−D
subject to ṽPV (k+ 1)= vc (k)− vc (k)
is ON and (11) when the switch is OFF: RC nD
      
TS 1−D DTS TS
σ =1
ṽPV (k + 1) = vc (k) − vc (k) (10) σ =0
ṽPV (k + 1) = iPV (k) − vc (k) + vc (k)
RC nD nC RC
   
DTS TS 1−D
σ =0
ṽPV (k + 1) = iPV (k) − vc (k) + vc (k) . (14)
nC RC nD
 
1−D The final switching state is the state that minimizes (14); the
. (11)
nD complete procedure of the controller is summarized in Fig. 5.
While model-based relations offer great robustness [37], a
drawback of such system, in general, is its dependence on the IV. STANDARD EN 50530 TEST
system model parameters. Generally, load resistance (R) is vari- Performance evaluations of MPPT techniques depend largely
able and sudden unpredicted perturbation in the load can render on the test being conducted. Different literature suggests differ-
the full system unstable. Hence, a simple, yet effective solu- ent test types like step changes. While such tests may prove the
tion is proposed to provide better monitoring on the load using effectiveness of the system, they fail to follow a standardized
an observer-based approach, without PV current measurement. acceptance [38]. Therefore, using a globally accepted test for
The proposed algorithm uses only the already existing sensors MPPT is essential to evaluate a system’s performance.
METRY et al.: MPPT OF PHOTOVOLTAIC SYSTEMS USING SENSORLESS CURRENT-BASED MODEL PREDICTIVE CONTROL 1161

Fig. 6. Test sequence principle, medium-to-high solar irradiance level (black


solid line) and low-to-medium solar irradiance level (blue-dashed line).

TABLE I
SYSTEM MODEL PARAMETER TABLE

System model parameter table

Average switching frequency (F) 5 kHz


Sampling time (T S ) 10 μs
Load value (R) 10 Ω
Output capacitor (C) 470 μF
Open-circuit voltage (V O C ) 64.2 V
Short-circuit current (I S C ) 5.96 A
Voltage at MPP (V M P ) 54.7 V
Current at MPP (I M P ) 5.58 A

Fig. 7. I–V and P–V characteristics of the PV array.

Fig. 5. SC-MPC-MPPT procedure to determine reference voltage, and [38]. The principle of the test sequence is illustrated in Fig. 6
determination of switching state using cost function minimization. parametrically. The slope of each ramp is named ζn which is
incrementally increasing by a factor of ε, and this sequence
To overcome the inconsistency in performance tests, an in- is repeated n times during the period under the test. The test is
ternational working group was set up in late 2006 to develop a comprised of three components as in Fig. 8: 1) low-to-medium
standardized test that takes into account both MPPT accuracy irradiation (150−500 W/m2 ); 2) medium-to-high irradiation
and conversion efficiency [38]. The test was accepted as a stan- (300−1000 W/m2 ); and 3) startup and shutdown irradiation
dard in the European Union by the end of 2009 and published (2−100 W/m2 ). Slopes for 1 vary from 0.5 to 50 W/m2 /s,
as The Standard EN 50530 Test [38]. while slopes for 2 vary from 10 to 100 W/m2 /s.
The dynamic EN 50530 standard tests are run under rapidly According to the standard, the test looks into both MPPT
changing weather conditions. It combines rising and falling accuracy and conversion efficiency of grid-connected PV sys-
ramp profiles with different slopes to represent irradiation levels tems [38]. Since the target of this efficiency test is the control
1162 IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 53, NO. 2, MARCH/APRIL 2017

Fig. 8. SC-MPC-MPPT simulation waveforms. (a) PV voltage. (b) PV current.

2
Fig. 9. Two components of the EN 50530 standard test: low-to-medium irradiation (150 − 500 W/m ) and medium-to-high irradiation (300 − 1000 W/m 2 ).

effectiveness of the MPPT technique, not the conversion stage SC-MPC-MPPT was found to be 8 μs. dSPACE was used in
(inverter of the PV system) the test becomes valid for the MPPT this paper for expedited prototyping; however, readily available
effectiveness for off-grid PV systems [39]. In this paper, the test and low-priced microprocessors, such as the Altera DEO-Nano
is applied to the proposed SC-MPC-MPPT technique to validate FPGA, are capable of handling the controller’s execution time
its performance. By which case, the inverter efficiency calcu- for the real application. In a fixed step MPC, unlike controllers
lation is not made and segment (C) of the test is disregarded with pulse-width modulator, the switching signals are directly
[39]. manipulated; thus, the “switching frequency” can vary from
In order to assess the dynamic performance of the MPPT one fixed sampling interval to the next. The sampling frequency
accuracy, the output power calculated from the measured volt- should be much higher than the switching frequency in order to
age and measured current is compared to the ideal maximum get good performance controller, such as 20 times higher accord-
power point [40]. Then, the control effectiveness of the proposed ing to the guidelines for accurate modeling of power electronics
SC-MPC-MPPT is determined and compared to fully observed [41]. For well-behaved MPC systems, we can compute an “av-
MPC-MPPT as well as the conventional P&O technique. erage switching frequency” which may offer some insight into
the operation of other aspects of the system. In this paper, the
sampling frequency is 100 kHz, which results in an average
V. RESULTS AND DISCUSSION switching frequency of 5 kHz.
The SUNPOWER SPR-305-WHT is used as PV module Simulation results corresponding to a ramp and step response
model. The PV module characteristics under standard test con- for the SC-MPC-MPPT voltage and current are shown in Fig. 8.
dition (solar irradiance = 1 kW/m2 , cell temperature = 25 °C) The irradiance profile of the EN 50530, Fig. 9, is then applied
are tabulated in Table I. Two modules are connected in parallel on the SC-MPC-MPPT controlled flyback converter. The EN
with the string I–V and P–V characteristics illustrated in Fig. 7. 50530 test performance results for duty cycle, power, voltage,
The control algorithm is implemented in MATLAB/Simulink; and current are as shown in Figs. 10, 11, 12, and 13, respectively.
the sampling time TS is 10 μs which corresponds to a sampling Due to highly effective MPPT, the real expected power from the
frequency of 100 kHz. This sampling time is chosen based PV is not added to the curve since they are overlapping and it is
on the capability of the dSPACE DS1007 platform processor not clear how much the error is at each test segment.
which is used for experimental results in this paper. Based on Using dSPACE DS1007 platform, the proposed controller is
dSPACE implementation, the execution time for the proposed implemented experimentally and three tests are performed to
METRY et al.: MPPT OF PHOTOVOLTAIC SYSTEMS USING SENSORLESS CURRENT-BASED MODEL PREDICTIVE CONTROL 1163

Fig. 10 Duty cycle of the flyback converter under dynamic EN 50530 standard test.

Fig. 11. PV power under dynamic EN 50530 standard test.

Fig. 12. PV voltage under dynamic EN 50530 standard test.

verify the simulation results: transient change in solar irradi- Fig. 14 shows the performance of the proposed SC-MPC-MPPT
ance level, real sky condition, step change response to solar for this experiment; as shown, the controller accurately tracks
irradiance level, and performance evaluation under steady-state the MPP with average efficacy of 99.4%.
condition. In the first experimental verification, the solar irra- The second experiment, Fig. 15, verifies the stability and fast
diance level was initially 750 W/m2 , then gradually decreased dynamic response of the SC-MPC-MPPT to step change in solar
to 500 W/m2 like the ramp rate in standard EN 50530 test. irradiance level from 500 to 750 W/m2 . Finally, the steady-state
1164 IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 53, NO. 2, MARCH/APRIL 2017

Fig. 13. PV current under dynamic EN 50530 standard test.

Fig. 14. PV current and voltage under gradually change in solar irradiance. Fig. 16. PV current and voltage ripple at 750 W/m2 solar irradiance level.

Fig. 15. PV current and voltage under step change in solar irradiance level.

performance at 750 W/m2 of the proposed technique is shown


in Fig. 16. As shown, the oscillation around MPP is negligible Fig. 17. Control effectiveness of MPC-MPPT versus SC-MPC-MPPT com-
pared to the conventional P&O MPPT (before and after optimizing the pertur-
with PV voltage and current ripple of 3.556% and 2.353%, bation size) over a range of irradiance values.
respectively.
METRY et al.: MPPT OF PHOTOVOLTAIC SYSTEMS USING SENSORLESS CURRENT-BASED MODEL PREDICTIVE CONTROL 1165

Fig. 18. Control effectiveness of the conventional P&O technique at different


perturbation size for solar irradiance levels from 200 to 1000W/m2 .

Fig. 20. Control effectiveness of SC-MPC-MPPT under load parameter mis-


match with and without the load model in the system.

which an MPPT technique, such as P&O, uses to increment


and decrement voltage to observe the MPP power based on the
characteristics of the P–V curve. Perturbation size was varied
from 0.01 to 0.15 at increments of 0.01. Then, a contour plot
of the control effectiveness was plotted over the range of irra-
diance from 200 to 1000 W/m2 , as shown for P&O MPPT in
Fig. 18 and SC-MPC-MPPT in Fig. 19. The range of the per-
turbation sizes, used in the quantitative analytical comparison,
was chosen to generate the greatest control effectiveness values.
As can be seen from Fig. 19, a perturbation size value of 0.05
for SC-MPC-MPPT, used to generate Fig. 17, yields the largest
high control effectiveness range. However, for P&O MPPT, the
Fig. 19. Control effectiveness of the proposed SC-MPC-MPPT technique at
different perturbation size for solar irradiance levels of 200–1000 W/m2 . perturbation size value of 0.06 was found to generate a slightly
higher control effectiveness than 0.05. Therefore, an optimized
P&O MPPT is plotted in Fig. 17 to demonstrate this effect.
To have a better comparison reference, the performance of the As mentioned earlier in this paper, a characteristic of MPC
proposed SC-MPC-MPPT, using the observer model for the PV is the use of system models for selecting optimal actuations.
current, is compared to conventional P&O MPPT and previously Therefore, evaluating the effect of model parameter mismatch
published MPC-MPPT [29], [30], considering the same pertur- on control effectiveness is of interest. The robustness of the pro-
bation size for all three methods initially. The control effective- posed SC-MPC-MPPT technique with and without [using the
ness, which quantifies the harvested PV power as a percentage load observer model in (13)] the load model in the proposed
from the expected power, is measured in this comparison. Under system is analyzed. The control effectiveness of the proposed
a wide range of irradiance from 200 to 1000 W/m2 , Fig. 17 com- SC-MPC-MPPT under load parameter variation from its nom-
pares control effectiveness performance between conventional inal value is shown in Fig. 20. In Fig. 20, the plot with the
P&O MPPT, MPC-MPPT, and the proposed SC-MPC-MPPT. circular marker represents the control effectiveness with the
Fig. 17 demonstrates that MPC-MPPT and SC-MPC-MPPT load model in the system. The plot with the diamond marker
performance exceed that of the conventional P&O MPPT as represents the control effectiveness without the load model in
explained in Section II. By the same token, the SC-MPC-MPPT the system, instead the load observer model from (13) is used
using the observer model for the PV current exhibits similar per- to reduce the load sensitivity of the controller. As it is shown,
formance, when compared to previously proposed MPC-MPPT. the control effectiveness of the proposed SC-MPC-MPPT with
Thus, while eliminating the current sensor, the controller perfor- a load mismatch of ±10% exhibits a high performance, almost
mance remained the same. To allow a fair comparison between similar to the nominal value (with control effectiveness of more
the proposed SC-MPC-MPPT and the conventional P&O MPPT, than 99.9%) which demonstrates approximately 100% model
a thorough analysis was made on the perturbation sizes asso- parameter mismatch rejection. As the load mismatch increased
ciated with these two controllers. Perturbation size is a value more than ±10%, the average parameter mismatch rejection
1166 IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 53, NO. 2, MARCH/APRIL 2017

achieved was 99.6% for the SC-MPC-MPPT without the load [9] J. Ahmed and Z. Salam, “A modified P&O maximum power point track-
model while the average parameter mismatch rejection achieved ing method with reduced steady state oscillation and improved tracking
efficiency,” IEEE Trans. Sustain. Energy, vol. 7, no. 4, pp. 1506–1515,
for the SC-MPC-MPPT with the load model is 98.9%. As the Oct. 2016.
parameter mismatch value of the load is away from the nominal [10] G. Escobar, S. Pettersson, C. N. M. Ho, R. E. Quintal-Palomo, and
value (0% mismatch), the control effectiveness falls down. As I. E. Llanez-Caballero, “Multi-sampling maximum power point tracker
(MS-MPPT) to compensate irradiation and temperature changes,” in
it is depicted, the lowest control effectiveness of the controller Proc. Annu. Conf. IEEE Ind. Electron. Soc., Dallas, TX, USA, 2014,
with the load model is 98.6% for –30% mismatch in load while it pp. 4974–4980.
is 99.2% for the controller without the load model. Thus, by just [11] M. Uoya and H. Koizumi, “A calculation method of photovoltaic array’s
operating point for MPPT evaluation based on one-dimensional Newton–
using the already existing sensors for SC-MPC-MPPT and using Raphson method,” IEEE Trans. Ind. Appl., vol. 51, no. 1, pp. 567–575,
the observer model of the load in (13), a more robust MPC to Jan./Feb. 2015.
load parameter mismatch is achieved. However, the performance [12] A. El Khateb, N. Abd Rahim, J. Selvaraj, and M. Nasir Uddin,
“Fuzzy-logic-controller-based SEPIC converter for maximum power point
of the SC-MPC-MPPT without the load model is comparable, tracking,” IEEE Trans. Ind. Appl., vol. 50, no. 4, pp. 2349–2358,
even slightly better than the load-dependent model. The use of Jul./Aug. 2014.
SC-MPC-MPPT without the load model makes it possible to use [13] P. E. Kakosimos, A. G. Kladas, and S. N. Manias, “Fast photovoltaic-
system voltage- or current-oriented MPPT employing a predictive digital
the proposed technique in this paper for more general systems current-controlled converter,” IEEE Trans. Ind. Electron., vol. 60, no. 12,
where the output of the dc/dc converter is connected to a dc bus pp. 5673–5685, Dec. 2013.
or an inverter. [14] L. Tarisciotti, P. Zanchetta, A. Watson, J. C. Clare, M. Degano, and S.
Bifaretti, “Modulated model predictive control for a three-phase active
rectifier,” IEEE Trans. Ind. Appl., vol. 51, no. 2, pp. 1610–1620, Mar./Apr.
VI. CONCLUSION 2015.
[15] H. A. Sher, A. F. Murtaza, A. Noman, K. E. Addoweesh, K. Al-Haddad,
This paper presented an improved MPPT technique by pre- and M. Chiaberge, “A new sensorless hybrid MPPT algorithm based
on fractional short-circuit current measurement and P&O MPPT,” IEEE
dicting the error at next sampling time before applying the Trans. Sustain. Energy, vol. 6, no. 4, pp. 1426–1434, Oct. 2015.
switching signal using MPC. Then, it proposed a sensorless [16] G. Farivar, B. Hredzak, and V. G. Agelidis, “A DC-side sensorless cas-
current algorithm to eliminate the additional cost of the cur- caded H-bridge multilevel converter-based photovoltaic system,” IEEE
Trans. Ind. Electron., vol. 63, no. 7, pp. 4233–4241, Jul. 2016.
rent sensor. The proposed solution presented a comparable per- [17] P. Midya, P. T. Krein, and M. F. Greuel, “Sensorless current mode control—
formance response to the MPC-MPPT under rapidly changing An observer-based technique for dc–dc converters,” IEEE Trans. Power
atmospheric conditions, using EN 50530 standard test. It also Electron., vol. 16, no. 4, pp. 522–526, Jul. 2001.
[18] Y. Yang, V. Sularea, K. Ma, and F. Blaabjerg, “Advanced Design tools for
showed a better performance when compared to conventional the reliability of power electronics—Case studies on a photovoltaic (PV)
P&O MPPT. Moreover, an observer model is developed for the system,” in Proc. Annu. Conf. IEEE Ind. Electron. Soc., Yokohama, Japan,
resistive load using existing sensors to improve the robustness of Nov. 9–12, 2015, pp. 2828–2833.
[19] R. Min, Q. Tong, Q. Zhan, X. Zou, K. Yu, and Z. Liu, “Digital sensorless
the controller to disturbances in the load and increase the system current mode control based on charge balance principle and dual cur-
reliability. The dSPACE DS1007 platform was used for imple- rent error compensation for dc–dc converters in DCM,” IEEE Trans. Ind.
menting the control technique experimentally and verification Electron., vol. 63, no. 1, pp. 155–165, Jan. 2016.
[20] N. Kasa, T. Iida, and L. Chen, “Flyback inverter controlled by sensorless
of simulation results. current MPPT for photovoltaic power system,” IEEE Trans. Ind. Electron.,
vol. 52, no. 4, pp. 1145–1152, Aug. 2005.
REFERENCES [21] G. M. Dousoky, E. M. Ahmed, and M. Shoyama, “Current-sensorless
MPPT with dc–dc boost converter for photovoltaic battery chargers,”
[1] International Energy Agency. 2014. World Energy Investment Outlook in Proc. IEEE Energy Convers. Congr. Expo., Raleigh, NC, USA,
| Special Report. [Online]. Available: https://www.iea.org/publications/ Sep. 20–24, 2012, pp. 1607–1614.
freepublications/publication/WEIO2014.pdf, accessed on Jan. 31, 2016. [22] E. Dallago, D. G. Finarelli, U. P. Gianazza, A. L. Barnabei, and A. Liberale,
[2] International Energy Agency. 2016. Tracking Clean Energy Progress “Theoretical and experimental analysis of an MPP detection algorithm
2016: Energy Technology Perspectives 2016 Excerpt, IEA Input to the employing a single-voltage sensor only and a noisy signal,” IEEE Trans.
Clean Energy Ministerial. [Online]. Available: https://www.iea.org/ Power Electron., vol. 28, no. 11, pp. 5088–5097, Nov. 2013.
publications/freepublications/publication/TrackingCleanEnergyProgress [23] M. Bond and J.-D. Park, “Current-sensorless power estimation and MPPT
2016.pdf, accessed on Jun. 28, 2016. implementation for thermoelectric generators,” IEEE Trans. Ind. Elec-
[3] US Department of Energy. (2013, Aug. 20). Solar Performance tron., vol. 62, no. 9, pp. 5539–5548, Sep. 2015.
and Efficiency. [Online]. Available: http://energy.gov/eere/energybasics/ [24] M. A. Green, K. Emery, Y. Hishikawa, W. Warta, and E. D. Dunlop,
articles/solar-performance-and-efficiency, accessed on Aug. 7, 2016. “Solar cell efficiency tables (Version 45),” Progress Photovolt.: Res. Appl.,
[4] A. Elrayyah, Y. Sozer, and M. Elbuluk, “Microgrid-connected PV-based vol. 23, pp. 1–9, 2015.
sources: A novel autonomous control method for maintaining maximum [25] H. Abu-Rub, M. Malinowski, and K. Al-Haddad, Power Electronics for
power,” IEEE Ind. Appl. Mag., vol. 21, no. 2, pp. 19–29, Mar./Apr. 2015. Renewable Energy Systems, Transportation and Industrial Applications.
[5] M. Das and V. Agarwal, “Novel high-performance stand-alone solar PV New York, NY, USA: Wiley, 2014.
system with high-gain high-efficiency dc–dc converter power stages,” [26] T. Andrejasic, M. Jankovec, and M. Topic, “Comparison of direct max-
IEEE Trans. Ind. Appl., vol. 51, no. 6, pp. 4718–4728, Nov./Dec. 2015. imum power point tracking algorithms using EN 50530 dynamic test
[6] T. Esram and P. L. Chapman, “Comparison of photovoltaic array maximum procedure,” IET Renewable Power Gener., vol. 5, pp. 281–286, 2011.
power point tracking techniques,” IEEE Trans. Energy Convers., vol. 22, [27] M. Metry, M. B. Shadmand, L. Yushan, R. S. Balog, and H. Abu Rub,
no. 2, pp. 439–449, Jun. 2007. “Maximum power point tracking of photovoltaic systems using sensorless
[7] A. Reza Reisi, M. Hassan Moradi, and S. Jamasb, “Classification and current-based model predictive control,” in Proc. IEEE Energy Convers.
comparison of maximum power point tracking techniques for photovoltaic Congr. Expo., Montreal, QC, Canada, Sep. 20–24, 2015, pp. 6635–6641.
system: A review,” Renewable Sustain. Energy Rev., vol. 19, pp. 433–443, [28] M. B. Shadmand, M. Mosa, R. S. Balog, and H. Abu-Rub, “An im-
Mar. 2013. proved MPPT technique of high gain dc–dc converter by model pre-
[8] J. Teng, W. Huang, T. Hsu, and C. Wang, “Novel and fast maximum power dictive control for photovoltaic applications,” in Proc. IEEE Appl.
point tracking for photovoltaic generation,” IEEE Trans. Ind. Electron., Power Electron. Conf. Expo., Fortworth, TX, USA, Mar. 16–20, 2014,
vol. 63, no. 8, pp. 4955–4966, Aug. 2016. pp. 1048–2334.
METRY et al.: MPPT OF PHOTOVOLTAIC SYSTEMS USING SENSORLESS CURRENT-BASED MODEL PREDICTIVE CONTROL 1167

[29] M. B. Shadmand, R. S. Balog, and H. Abu-Rub, “Model predictive control Mohammad B. Shadmand (S’09–M’15) received
of PV sources in a smart dc distribution system: Maximum power point the B.S. degree from Qatar University, Doha, Qatar,
tracking and droop control,” IEEE Trans. Energy Convers., vol. 29, no. 4, in 2010, and the M.S. and Ph.D. degrees from Texas
pp. 913–921, Dec. 2014. A&M University, College Station, TX, USA, in 2012
[30] M. B. Shadmand, R. S. Balog, and H. Abu-Rub, “Maximum power point and 2015, respectively, all in electrical engineering.
tracking using model predictive control of a flyback converter for photo- He was a Research Associate in the Renewable
voltaic applications,” in Proc. IEEE Power Energy Conf. Illinois, Urbana, Energy and Advanced Power Electronics Research
IL, USA, Feb. 28–Mar. 1, 2014, pp. 1–5. Laboratory, Texas A&M University, from 2010 to
[31] M. Metry, M. B. Shadmand, R. S. Balog, and H. A. Rub, “Sensitivity 2015, where he is currently a Texas A&M Engineer-
analysis to model parameter errors of MPPT by model predictive control ing Experiment Station Research Engineer. He was
for photovoltaic applications,” in Proc. IEEE 1st Workshop Smart Grid a Visiting Researcher with Smart Grid Center, Texas
Renewable Energy, Doha, Qatar, Mar. 22–23, 2015. A&M University at Qatar, 2014. He was an Instructor in the Department of Elec-
[32] H. Jiefeng, Z. Jianguo, and D. G. Dorrell, “Model predictive control of trical and Computer Engineering, Texas A&M University, in Fall 2015, where
grid-connected inverters for PV systems with flexible power regulation he has been again an Instructor since 2016. He was a Postdoctoral Research
and switching frequency reduction,” IEEE Trans. Ind. Appl., vol. 51, Associate in the Department of Electrical Engineering, Texas A&M University
no. 1, pp. 587–594, Jan./Feb. 2015. at Qatar, in 2016. He has published more than 45 journal and conference pa-
[33] S. Kouro, M. A. Perez, J. Rodriguez, A. M. Llor, and H. A. Young, “Model pers. His current research interests include advanced model predictive control,
predictive control—MPC’s role in the evolution of power electronics,” maximum power point tracking of photovoltaic systems, photovoltaic balance
IEEE Ind. Electron. Mag., vol. 9, no. 4, pp. 8–21, Dec. 2015. of system, matrix converter, and control of smart microgrid systems.
[34] C. Bordons and C. Montero, “Basic principles of MPC for power convert- Dr. Shadmand awarded the second place in the IEEE Industrial Application
ers: Bridging the gap between theory and practice,” IEEE Ind. Electron. Society Graduate Thesis Contest for his M.S. thesis in 2013. He received the
Mag., vol. 9, no. 3, pp. 31–43, Sep. 2015. IEEE Standard Education Award for the project “fixed-step model predictive
[35] J. Rodriguez and P. Cortes, Predictive Control of Power Converters and control of grid-tied photovoltaic inverter” in 2014.
Electrical Drives, vol. 37. New York, NY, USA: Wiley, 2012.
[36] J. B. Rawlings and D. Q. Mayne, Model Predictive Control: Theory and
Design. Madison, WI, USA: Nob Hill Publishing, 2009. Robert S. Balog (S’92–M’96–SM’07) received the
[37] H. A. Young, M. A. Perez, and J. Rodriguez, “Analysis of finite-control- B.S. degree from Rutgers–The State University of
set model predictive current control with model parameter mismatch New Jersey, New Brunswick, NJ, USA, in 1996, and
in a three-phase inverter,” IEEE Trans. Ind. Electron., vol. 63, no. 5, the M.S. and Ph.D. degrees from the University of
pp. 3100–3107, May 2016. Illinois at Urbana-Champaign, Urbana, IL, USA, in
[38] R. Bründlinger, N. Henze, H. Häberlin, B. Burger, A. Bergmann, and 2003 and 2006, respectively, all in electrical engi-
F. Baumgartner, “prEN 50530—The new European standard for perfor- neering.
mance characterisation of PV inverters,” in Proc. 24th Eur. Photovolt. Sol. He was an Engineer with Lutron Electronics,
Energy Conf., Germany, Sep. 21–25, 2009, pp. 3105–3109. Coopersburg, PA, USA, from 1996 to 1999, a Re-
[39] K. Ishaquea, Z. Salamb, and G. Laussc, “The performance of perturb searcher with the U.S. Army Corp of Engineers,
and observe and incremental conductance maximum power point track- Engineering Research and Development Center,
ing method under dynamic weather conditions,” Appl. Energy, vol. 119, Construction Engineering Research Lab, Champaign, IL, from 2005 to 2006,
pp. 228–236, Apr. 15, 2014. a Senior Engineer at SolarBridge Technologies (now SunPower, Inc.), Cham-
[40] M. Metry, M. B. Shadmand, R. S. Balog, and H. Abu Rub, “High effi- paign, from 2006 to 2009, and then joined Texas A&M University, College
ciency MPPT by model predictive control considering load disturbances Station, TX, USA, where he is currently an Associate Professor in the Depart-
for photovoltaic applications under dynamic weather condition,” in Proc. ment of Electrical and Computer Engineering and the Director of the Renew-
Annu. Conf. IEEE Ind. Electron. Soc., Yokohama, Japan, Nov. 9–12, 2015, able Energy and Advanced Power Electronics Research Laboratory. He holds
pp. 4092–4095. a joint faculty appointment with the Department of Electrical Engineering at
[41] H. F. Blanchette, T. Ould-Bachir, and J. P. David, “A state-space model- Texas A&M University in Qatar. He holds 14 issued U.S. patent with additional
ing approach for the FPGA-based real-time simulation of high switching patents pending. His current research interests include power converters for so-
frequency power converters,” IEEE Trans. Ind. Electron., vol. 59, no. 12, lar energy, particularly microinverters for ac photovoltaic modules and highly
pp. 4555–4567, Dec. 2012. reliable electrical power and energy systems including dc microgrids.
Dr. Balog is a Registered Professional Engineer in Illinois. He received the
IEEE Joseph J. Suozzi INTELEC Fellowship in power electronics in 2001. He
is a member of Eta Kappa Nu, Sigma Xi, the National Society of Professional
Engineers, the American Solar Energy Society, and the Solar Electric Power
Association. He is the 2011 recipient of the Rutgers School of Engineering
Distinguished Engineer Award.

Haitham Abu-Rub (M’99–SM’07) received the


M.Sc. degree in electrical engineering from Gdynia
Marine Academy, Gdynia, Poland, in 1990, and the
Morcos Metry (S’13) received the Bachelor of Sci- Ph.D. degree in electrical engineering from Gdansk
ence degree (with hons.) in electrical and computer University of Technology, Gdansk, Poland, in 1995.
engineering from Texas A&M University at Qatar, For eight years, he was with Birzeit University,
Doha, Qatar, and is currently working toward the Birzeit, Palestine, where he was first an Assistant Pro-
Ph.D. degree in the Renewable Energy and Advanced fessor and then an Associate Professor and was the
Power Electronics Research Laboratory, Department Chairman of the Electrical Engineering Department
of Electrical and Computer Engineering, Texas A&M for four years. He is currently a Full Professor in the
University, College Station, TX, USA. Department of Electrical and Computer Engineering,
He is with the Research and Development Depart- Texas A&M University at Qatar, Doha, Qatar. He is coauthor of five books, five
ment of Qatar Foundation (QF R&D) as a Research book chapters and around 300 conference and journal paper. His main research
Associate. His research interests are in photovoltaic interests include the electrical machine drives and power electronics.
integration with the grid using advanced control methods. Dr. Abu-Rub received many international prestigious awards, such as the
Mr. Metry received the Graduate Teaching Fellowship from Texas A&M American Fulbright Scholarship (at Texas A&M University, College Station,
University, to co-teach the Fundamentals of Power Electronics to more than TX, USA), the German Alexander von Humboldt Fellowship (at Wuppertal
150 undergraduate and graduate students. He also served as the President of University, Wuppertal, Germany), the German DAAD Scholarship (at Ruhr
the Eta-Kappa-Nu honor society, the Lambda Mu student chapter, and as the University Bochum, Bochum, Germany), and the British Royal Society Schol-
Program Chair of the IEEE student organization in 2013 and 2014. arship (at the University of Southampton, Southampton, U.K.).

You might also like