8th International Conference on Renewable Energy Research and Applications Brasov, ROMANIA, Nov.

3-6, 2019

MPPT Based Model Predictive Control of

Grid Connected Inverter for PV Systems
Naki Güler1 Erdal Irmak2
1 2
Technical Sciences Vocational School, Electrical and Electronics Engineering Department
Gazi University, Ankara, Turkey Gazi University, Ankara, Turkey
[email protected] [email protected]

Abstract—This paper presents a Maximum Power Point Fast dynamic response, robustness and insensitive to system
Tracking(MPPT) based Model Predictive Control (MPC) parameters are certain advantages of the sliding mode control
approach to obtain high accuracy and fast dynamic response. The algorithm. Because of these advantages, the SMC method is
tracking capability of the base algorithm is improved by the preferred in many study to control of power electronic
combination of two-methods. The proposed control approach converters. In [12], SMC and MPPT algorithms are used for
tested on a three-phase grid-connected inverter that fed by PV power transfer from PV to grid. PV panel group is connected a
panel group. Switching signals of the inverter are generated by the DC/DC converter and the converter is controlled by a variable
MPC algorithm. Reference current of the MPC algorithm step incremental conductance MPPT method. Power flow from
determined by Perturb and Observe MPPT method. Thus, power
DC/DC converter to grid is provided by an inverter that
flow is controlled by the MPC algorithm based on MPPT. Power
flow, MPPT efficiency and THD analyzes are examined in a
controlled with the SMC algorithm. The paper points out that,
simulation that performed by using MATLAB/Simulink the fast dynamic response is obtained by using SMC algorithm.
environment. Especially, the effectiveness of the proposed MPC method is another control method that has similar
approach has been tested under varying irradiation and cloudy advantages to SMC. The MPC algorithm has been applied to
conditions. Besides the MPPT analyzes, current tracking most parts of power electronic converters in the past decade.
capability of the MPC algorithm is examined under dynamic Besides the only voltage or current control, the MPC algorithm
transition conditions. Results show that, MPPT efficiency of the is combined with MPPT to obtain better tracking performance
proposed control approach is 98%. [2, 13]. Especially, the algorithm offers a better performance
under varying irradiation conditions. In [13], both DC/DC and
Keywords—model predictive control; maximum power point DC/AC power converters are controlled by the MPC method.
tracking; three-phase inverter; current control Thus, tracking performance is improved under cloudy weather
conditions. The predictive based algorithm is used for improving
I. INTRODUCTION the MPPT efficiency in [14]. A grid-connected Z-Source
The decreasing of fossil fuel reserves and the reduction of inverter is controlled by the algorithm. Comparative results with
photovoltaic cell costs have increased the using of PV systems. P&O method and proposed algorithm shown the algorithm
Even if the PV panels have a low cost, popularities of MPPT offers better performance. Only one stage power converter is
methods are continuing as part of control the power electronic used in the study.
converters [1]. In general, secondary control algorithms [1-5] are
One stage power converter usage is another highlight of the
preferred to increase of sensitivity and dynamic capability of
paper. In general, the first power stage is used to increase the
control algorithms. MPC [2] and sliding mode control (SMC)
[6] techniques are popular methods in terms of dynamic input voltage. As in [12], [13] and [15], two-stage power layer
responses. Although these methods need high computational usage increases the total cost of the system. Instead of that,
single-stage power converter usage suggested to reduce the cost
power, the challenge has eliminated by technological
improvements on microcontrollers. of system in [16]. But, in that case, series-connected panel group
is needed to increase the DC bus voltage. Even if different
Different solutions are proposed to improve the effectiveness solutions are proposed to increase of DC bus voltage [14, 17-
of MPPT algorithm. In [7-10], Adaptive Neuro Fuzzy Inference 19], one-stage power layer usage with series-connected PV
and P&O MPPT methods are combined to obtain better tracking panels is continuing to be the most preferred model [15, 20].
performance. As a result of the study, the combined algorithm is
With the above motivations, a grid-connected PV system is
offered better results than the classical method. Mechanical
proposed in the paper. The system includes a three-phase
tracker usage is another method for improving the tracking
inverter that controlled by MPPT-MPC algorithm. MPP is
efficiency [11]. This system type can be implemented for low
determined by the MPPT algorithm and it generates a reference
power PVs, but mechanical observer usage is not a realistic
current for the inverter. MPC algorithm generates the switching
approach for high power systems. Because of the challenges on
signals by depending on the reference current. Thus, power flow
mechanical observer, advanced control algorithms are used to
has been provided by all the power and control structures.
obtain a better dynamic response and sensitive control.

ICRERA 2019 978-1-7281-3587-8/19/$31.00 ©2019 IEEE 982

8th International Conference on Renewable Energy Research and Applications Brasov, ROMANIA, Nov. 3-6, 2019

II. SYSTEM MODEL & CONTROL METHOD

General circuit schema and control techniques are given in
Fig. 1. PV panel group is directly connected to grid through a
three-phase inverter. Inductance and capacitance of the LC filter
can be calculated by Eq. (1) and Eq. (2), respectively.
.
= (1)
.
= (2)

PV power is used as a total system power for calculations.

PV panel group contains 24 series and 4 parallel arrays and total
input power is 20kW. Filter parameters are calculated by using
the power value per phase. The inductance and capacitance Fig 2. Space vectors of three-phase two-level inverter
values are calculated as 6.9mH and 5μF, respectively.
A. Model Predictive Current Control Method
The MPC algorithm works by minimizing the control
parameter error between the reference and the next sampling
interval. MPC can be summarized in two steps as prediction and
minimizing. Discretization methods are using to predict the
control parameter. Exact discretization and forward Euler are
widely used methods. Exact discretization method has high
accuracy. But, the method increases the computational burden.
Because of that, the forward Euler approximation method (6) is
preferred in low order systems.
( ) ( )
≈ (6)

Eq. (7) can be obtained by substituting the Euler method to

the derivative in Eq. (4).
Fig 1. General schema of the system
( ) ( )
= ( − ( ) − ( )) (7)
The classic P&O MPPT algorithm that detailed discussed in
[2] is used to determine the reference current for the inverter. The prediction equation (Eq. 8) can be obtained by arranging
Because of that, this study focuses on model predictive current the Eq. (7).
control method which is used to control the inverter current
(iabc). General voltage equation of the inverter is given in Eq. (3). ( ) = ( ) − ( ) − ( ) + ( ) (8)
= + + (3)
Current values are predicted for all possible voltage states
(1….n) by using Eq. (8). As mentioned before, the minimization
Output current can be expressed by replacing of inductor of error is the second step of the MPC algorithm. Most part of
voltages in Eq. (4): control algorithms uses difference between reference and
( ) measured value. Different from the other methods, the MPC
= ( ( ) − ( ) − ( )) (4) algorithm uses the error between reference and predicted value.
Thanks to this feature, the algorithm generates an action for the
Where, Vgrid(k) is measured grid voltage in the current sampling next step. Therefore, MPC has a better dynamic capability. To
interval. All the measured parameters are converted to α-β calculate the error term, cost functions are using in the MPC
coordinate by using Clarke transformation. Vinv is inverter algorithms. The cost function of the control algorithm is
voltage vector that used for predictions. Space vectors of the represented in Eq. (9). iα and iβ refer to real and imaginary value
voltage are shown in Fig. 2. As shown in the figure, three-phase of currents, respectively.
inverters have six active and two “0” states. It shows that the = ∗
− + ∗
− (9)
( ) ( )
inverter has 8 possible states. Phase states (SA, SB, SC) are
defined as an array that contains all voltage states. Inverter Similar to the prediction equation, the cost function is
voltage states can be expressed with Eq. (5) in α-β frame [19]. evaluated for all possible voltage vectors in each sampling
Thus, all the control parameters in Eq. (4) are defined in α-β interval. Thus, error values are obtained for all possible
frame. In Eq. (5), n is state number {1,2……..8} switching positions. The minimization step selects the optimum
√ √
cost function that has the minimum error. Switching signals are
( ) = ( ) + − + ( ) + − − ( ) (5) generated according to the optimum cost function.

ICRERA 2019 978-1-7281-3587-8/19/$31.00 ©2019 IEEE 983

8th International Conference on Renewable Energy Research and Applications Brasov, ROMANIA, Nov. 3-6, 2019

III. SIMULATION RESULTS After the dynamic performance analysis, power flow control
System model and proposed control structures in Fig. 1 are performance of the proposed control algorithm has been tested
verified with simulation studies. Simulation parameters are under varying irradiation conditions. The irradiation curve is
given in Table I. Firstly, the dynamic performance of the model given in Fig. 4. As seen from the figure, the curve also includes
predictive current control algorithm has been tested with step two cloud effects. PV power results are given in Fig. 5. As seen
changes on the reference. As seen from the results in Fig. 3, the from Fig. 5(a) and Fig. 5 (b), the MPPT algorithm is tracking the
reference current is defined as 15A at the beginning of the MPP with 98% efficiency. Even if the efficiency decrease in
simulation. Peak value of the inverter current successfully cloud conditions, the efficiency increased to 98% after these
regulated to the reference value by the MPC algorithm. Then, times.
the reference current changed from 15A to 30A. As seen from
the detailed results, inverter current is regulated to the new
reference in less than 1ms. Step down test is performed by
decreasing the reference from 30A to 15A. The results show
that, reference tracking is achieved in 0.1ms. As seen from both
step-change results, the peak values of inverter currents are
successfully fixed to the reference by the MPC algorithm.

TABLE I. SYSTEM PARAMETERS

Parameter Value Fig 4. Irradiation curve

U (Grid voltage) 380V

f (Grid frequency) 50Hz
P (Nominal power) 20kW
Lf (Filter inductance per phase) 6.9mH
RLf (Resistance of inductor) 0.7Ω
Cf (Filter capacitance per phase) 5μF
C (DC link capacitance) 330μF
PV Panel (24 series and 4 parallel) 213WMPP and 29VMPP (a)
TS (Sampling time) 20μs

(b)

(c)

(d)
Fig 3. Step change results Fig 5. PV results, a: Power, b: MPPT Efficiency, c: PV voltage, d: PV current

ICRERA 2019 978-1-7281-3587-8/19/$31.00 ©2019 IEEE 984

8th International Conference on Renewable Energy Research and Applications Brasov, ROMANIA, Nov. 3-6, 2019

Voltage and current results of the PV panel group are given Detailed results of cloudy conditions (“A” and “B” in Fig. 6)
in Fig. 5 (c) and Fig. 5(d), respectively. While the current is are given in Fig. 7. The reference current decreased from 35A to
increasing up to 28A, PV voltage approximately constant. It's 22A in the cloudy condition “A”. As seen in Fig. 7(a), inverter
caused by natures of MPPT control in PVs and it gives currents are successfully regulated to the reference by the MPC
information about the control performance. algorithm. Similar control actions are seen also for “B”
condition in Fig. 7(b). Current tracking results show the
Besides the performance test of the MPPT algorithm, the combined control algorithms successfully track the MPP not
inverter side is examined in terms of power, efficiency and only in varying irradiation but also in cloudy conditions.
harmonic distortions. As shown in Fig. (6), the inverter currents
are successfully fixed to the reference value (Iref*) by the MPC
algorithm. Because of the MPPT algorithm changes the
reference current value, the output currents of the inverter are
also changed depending on the irradiation curve in Fig. (4). Fig.
6(b) show that, harmonics are lower than 5% for bigger power
values than 5kW. The THD value is lower than the specified
limit in standards [21].

(a)

(a)

(b)
Fig 8. Inverter efficiency results, a: Power results, b: Overall efficiency of the
system

Transferred power to grid at the output of the inverter and

ideal MPP curves have been compared to analyze the overall
efficiency of the proposed system. The power losses between
inverter output power and ideal maximum power curve can be
(b) seen in Fig 8(a). Especially the power losses are increased
Fig 6. Inverter results, a: Currents, b: THD
around the nominal power. It caused by filter inductance. Even
if the losses increased, Fig. 8(b) shows that the minimum system
efficiency is approximately 90% for bigger power values than
5kW. Inverter efficiency can be calculated as 92% by using Fig.
5(b) and Fig. 8(b).
IV. CONCLUSION
The paper proposes a combined control algorithm for PV
systems. The proposed control approach consists of MPPT and
MPC algorithms. Thanks to the fast and sensitive control
capability of MPC, the combined algorithm offered effective
(a)
tracking under both varying irradiation and cloudy conditions.
Besides the combined algorithm performance tests, it is seen
from the step-change tests that the MPC algorithm has a superior
current control capability. Efficiency analysis has been
performed for both control and power structures. The control
efficiency of the MPPT method is 98% and inverter efficiency
is 92%. The overall system efficiency for both control
algorithms and inverter is approximately 90%.

(b)
Fig 7. Detailed current tracking results in cloudy conditions, a: “A”, b: “B”

ICRERA 2019 978-1-7281-3587-8/19/$31.00 ©2019 IEEE 985

8th International Conference on Renewable Energy Research and Applications Brasov, ROMANIA, Nov. 3-6, 2019

REFERENCES
[1] M. Metry, M. B. Shadmand, R. S. Balog and H. Abu-Rub, “MPPT of
photovoltaic systems using sensorless current-based model predictive [12] N. Altin, S. Ozdemir, H. Komurcugil, I. Sefa and S. Biricik, “Two-stage
control,” IEEE Transactions on Industry Applications, vol. 53, no. 2, pp. grid-connected inverter for PV systems,” 2018 IEEE 12th International
1157-1167, March-April 2017. Conference on Compatibility, Power Electronics and Power Engineering
(CPE-POWERENG 2018), Doha, 2018, pp. 1-6.
[2] Erdal Irmak and Naki Güler, “A model predictive control-based hybrid
MPPT method for boost converters,” International Journal of Electronics, [13] M. B. Shadmand, X. Li, R. S. Balog and H. A. Rub, “Model predictive
in press, 2019, DOI:10.1080/00207217.2019.1582715. control of grid-tied photovoltaic systems: Maximum power point tracking
[3] E. Kabalci, G. Gokkus and A. Gorgun, “Design and implementation of a and decoupled power control,” 2015 First Workshop on Smart Grid and
Renewable Energy (SGRE), Doha, 2015, pp. 1-6.
PI-MPPT based Buck-Boost converter,” 2015 7th International
Conference on Electronics, Computers and Artificial Intelligence (ECAI), [14] S. Sajadian, R. Ahmadi and H. Zargarzadeh, “Extremum seeking-based
Bucharest, 2015, pp. SG-23-SG-28. model predictive MPPT for grid-tied z-source inverter for photovoltaic
systems,” IEEE Journal of Emerging and Selected Topics in Power
[4] Ihssane Chtouki, Patrice Wira, Malika Zazi, Bruno Collicchio,Sami
Electronics, vol. 7, no. 1, pp. 216-227, March 2019.
Meddou, “Design, implementation and comparison of several neural
perturb and observe MPPT methods for photovoltaic systems”, [15] I. Colak, E. Kabalci and G. Bal, “Parallel DC-AC conversion system
International Journal of Renewable Energy Research-IJRER, 2019, vol. based on separate solar farms with MPPT control,” 8th International
9, no. 2, pp. 757-770. Conference on Power Electronics - ECCE Asia, Jeju, 2011, pp. 1469-
1475.
[5] N. Genc, H. Uzmus and I. Iskender, “Dynamic behavior of dc-dc boost
converter controlled with cascade PI-ASC,” 8th International Conference [16] Saban Ozdemir, Necmi Altin and Ibrahim Sefa “Single stage three level
on Electronics, Computers and Artificial Intelligence (ECAI), Ploiesti, grid interactive MPPT inverter for PV systems,” Energy Conversion and
2016, pp. 1-4. Management, Volume 80, 2014, pp. 561-572.
[6] S. Biricik, H. Komurcugil, N. D. Tuyen and M. Basu, “Protection of [17] M. Kamalirad, H. Iman-Eini, B. Farhangi and S. Bacha, “A Reliable three-
sensitive loads using sliding mode controlled three-phase dvr with phase transformerless grid-connected PV inverter with inductive DC
adaptive notch filter,” IEEE Transactions on Industrial Electronics, vol. link,” IEEE Journal of Photovoltaics, vol. 8, no. 5, pp. 1305-1312, Sept.
66, no. 7, pp. 5465-5475, July 2019. 2018.
[7] K. Amara et al., "Improved Performance of a PV Solar Panel with [18] Naci Genc, Yavuz Koc, “Experimental verification of an improved soft-
Adaptive Neuro Fuzzy Inference System ANFIS based MPPT," 2018 7th switching cascade boost converter,” Electric Power Systems Research,
International Conference on Renewable Energy Research and Vol. 149 , pp 1-9, 2017.
Applications (ICRERA), Paris, 2018, pp. 1098-1101. [19] E. Irmak and N. Güler, “Model predictive control of grid-tied three level
[8] D. Haji and N. Genc, "Fuzzy and P&O Based MPPT Controllers under neutral point clamped inverter integrated with a double layer multi-input
Different Conditions," 2018 7th International Conference on Renewable single output DC/DC converter,” 2018 IEEE 12th International
Energy Research and Applications (ICRERA), Paris, 2018, pp. 649-655. Conference on Compatibility, Power Electronics and Power Engineering
(CPE-POWERENG 2018), Doha, 2018, pp. 1-6.
[9] Neeraj Priyadarshi, Amarjeet Kumar Sharma, Faarooque Azam, “A
Hybrid Firefly-Asymmetrical Fuzzy Logic Controller based MPPT for [20] A. Sangwongwanich and F. Blaabjerg, “Mitigation of interharmonics in
PV-Wind-Fuel Grid Integration”, International Journal of Renewable PV systems with maximum power point tracking modification,” IEEE
Energy Research-IJRER, 2017, vol. 7, no. 4, pp. 1546-1560. Transactions on Power Electronics, vol. 34, no. 9, pp. 8279-8282, Sept.
2019.
[10] Mahdi Heidari, “Improving Efficiency of Photovoltaic System by Using
Neural Network MPPT and Predictive Control of Converter”, [21] IEEE Recommended Practice and Requirements for Harmonic Control in
International Journal of Renewable Energy Research-IJRER, 2016, vol. Electric Power Systems, in IEEE Std 519-2014 (Revision of IEEE Std
6, no. 4, pp. 1524-1529. 519-1992), June 11 2014, pp.1-29.
[11] A. Dolara, S. Leva, G. Magistrati, M. Mussetta, E. Ogliari and R. V.
Arvind, “A novel MPPT algorithm for photovoltie systems under
dynamic partial shading — Recurrent scan and track method,” 2016 IEEE
International Conference on Renewable Energy Research and
Applications (ICRERA), Birmingham, 2016, pp. 1122-1127.

ICRERA 2019 978-1-7281-3587-8/19/$31.00 ©2019 IEEE 986