Академический Документы
Профессиональный Документы
Культура Документы
Solar Energy
journal homepage: www.elsevier.com/locate/solener
A R T I C L E I N F O A B S T R A C T
Keywords: This paper focuses on the use of model predictive control (MPC) to control a DC/DC boost converter in order to
PV system regulate the PV power. When integrated with the grid, the PV system must deliver maximum power most of the
MPPT time; however, if a voltage sag occurs, new grid codes demand that the control system should limit the PV power
MPC generated to avoid over current conditions and, consequently, a grid disconnection. Maximum and reduced
CHIL
power modes are implemented following the MPC strategy to achieve high–performance and stable operation in
the system. First, the system is modeled in Matlab/Simulink and PLECS to understand its operation and to
evaluate the effectiveness of the proposed algorithm. Secondly, experimental results are verified using the
control hardware–in–the–loop (CHIL) approach on the Real Time Digital Simulator (RTDS).
1. Introduction power operating point than the MPP to lessen the negative impact of
grid overloading. Therefore, PV controllers must offer a fast and reliable
During normal conditions, the maximum power point tracking transition between the normal and fault operation mode to satisfy
(MPPT) technique are required on PV installations to maximize solar current standards.
harvesting and increase the profit provided by feed-in tariff govern- Pulse width modulation (PWM) techniques are commonly used to
mental programs. Different methods for MPP tracking on PV systems achieve the operation modes previously described. The proper selection
have been proposed in the literature and can be classified into direct of the switching frequency is an important step in the system design
and indirect methods. Indirect methods require prior knowledge of the because it can affect both the size of the converter—by increasing the
PV array characteristics under different irradiance/temperature condi- size of the energy store elements like inductances—as well as the
tions. In contrast, direct methods use voltage and current measurements overall efficiency of the system (Rodriguez-Rodrıguez et al., 2017).
from the PV array to achieve the optimal operation point. The selection Moreover, the classical control techniques with PI regulators assume
of a specific technique depends mostly on the implementation com- that the converter is a linear system, and the transient performance as
plexity, number of sensors required, and convergence speed that is re- well as the robustness of the system are limited by the selection of the
quired for a certain application like solar vehicles (Ko and Chao, 2012). compensator bandwidth (Rivera et al., 2013).
In the literature, the Hill Climbing/Perturb and Observe (P&O) Several articles have been published in literature in order to obtain
(Balasankar et al., 2017; Rezk and Eltamaly, 2015) and Incremental a RPM from PV arrays. They are implemented for both single- and two-
Conductance (López et al., 2016) techniques are the most widely used stage PV systems (Kiran et al., 2015; Urtasun et al., 2013;
because they are easy to implement, and they provide good efficiency Sangwongwanich et al., 2016). All utilize a P&O-based approach
that goes from 96 to 99% (Ezinwanne et al., 2017; Bendib et al., 2015). technique to move PV power PPV operation from MPP to a lower point.
Due to the increase penetration of PV systems into the grid, current This is done by means of a perturbation step change on the PWM duty
grid codes demand ancillary services to improve overall power system cycle. The drawbacks on this method are the oscillations around the
reliability (Nanou and Papathanassiou, 2014). This is the case when a new PV operation point and the speed of the process. Although, similar
fault occurs, the PV system must deliver voltage support by injecting P&O modified methods used for MPPT could be implemented to reduce
reactive power at the point of connection (Yang et al., 2014; Hudson the mentioned disadvantages (Dileep and Singh, 2015; Harrag and
and Heilscher, 2012). In this scenario the system should operate in a Messalti, 2015; Sheik Mohammed et al., 2016), the tracking speed re-
reduced power mode (RPM) to decrease the active power injection, and mains as the major weakness for a fast controller response during
in this way, avoid over current condition and eventual disconnection of transient events like grid faults.
the PV plant. Therefore, in a RPM, the PV controller must select a lower Model predictive control (MPC) offers an alternative to overcome
⁎
Corresponding author.
E-mail address: fd13c@my.fsu.edu (F. Diaz Franco).
http://dx.doi.org/10.1016/j.solener.2017.10.005
Received 8 May 2017; Received in revised form 14 September 2017; Accepted 2 October 2017
Available online 15 October 2017
0038-092X/ © 2017 Elsevier Ltd. All rights reserved.
F. Diaz Franco et al. Solar Energy 158 (2017) 679–686
In this section, existent grid code requirement for PV systems and its
relationship with PPV control is presented next. Additionally, the
mathematical model using FCS-MPC is elaborated and the algorithm to
accomplish MPPT and RPM is presented next.
680
F. Diaz Franco et al. Solar Energy 158 (2017) 679–686
based on the converter power limitation; and (iii) the PV operation As shown in (5), the prediction on PPV requires PV kth measured values
point moves from MPP to the required PPV∗ which is located either to for power, current, and voltages at every side of the boost converter.
the left or right of the PV curve (points a and b in Fig. 2). Also, to obtain the capacitor power PCPV, the derivative in (5) can be
Due the fast nature of voltage sags, the conversion from MPPT to discretized using the backward Euler method (Narimani et al., 2015).
RPM must be done in a similar speed range. Moreover, failing to reduce Eq. (5) must be solved for both switching state conditions to later
PPV at the required speed could lead to a protection trip of the system. minimize the cost function defined in (1). One way to improve the
To satisfy the control dynamics during fault scenarios, a FCS-MPC- execution speed for implementation is by solving (5) in terms of SBoost as
based controller is proposed and the modeling and algorithm im-
1 ⎧ L ∗
plementation is explained next. SBoost (k + 1) = 1 + [PPV (k + 1)−PPV (k ) + PCPV ]
Vdc (k ) ⎨
⎩ Ts VPV (k )
2.2. Modeling and control
−VPV (k )⎫,
⎬
⎭ (6)
In a two–stage PV system as shown in Fig. 3, the DC/DC boost ∗
where PCPV is the CPV power loss and (k + 1) has taken the place of
PPV
converter side, also known as the generator–side, regulates the PV
the predicted power. Eq. (6) provides an indication of the optimal
output power. Therefore, the mathematical model of this section is
switching state to achieve the required reference. Using the notation in
developed next. The control objectives are: (1) to achieve MPPT for the
(6), the cost function defined in (1) is modified to
normal operation and (2) to track PPV as it would be required during
fault scenarios. Using MPC-based control, the generator–side cost gB (k ) = [SBoost (k + 1)−SBoost States]2 , (7)
function can be expressed using the quadratic error form as (Sultana
et al., 2017) where SBoost_States represents a vector of the states of the boost converter
∗
(0 and 1).
gB (k ) = [PPV (k + 1)−PPV (k + 1)]2 , (1) The advantage of FCS-MPC offers over PWM techniques is that the
where ∗
(k + 1) represents the power reference for the next iteration
PPV switching state that delivers minimum error is chosen and applied to
that during fault conditions is proportional to the sag depth, and the power converter. Therefore, it can track the reference signal faster.
PPV (k + 1) corresponds to the predicted PPV delivered. As shown in (6), using PV measurements the system can track a re-
To predict the future behavior of PPV, the differential equation that quired reference by calculating the minimum error of the cost function
describe the boost converter dynamics must be obtained. The inductor defined in (7).
current can be found solving the KVL equations when the switch is on It has been shown the mathematical modeling of the generator–side
∗ ∗
and off, and it is given by (Diaz-Franco et al., 2016) in order to track PPV . Next, the selection of PPV is discussed for normal
operation (MPPT) and RPM.
diL 1
= [VPV + (SBoost −1) Vdc ],
dt L (2)
2.3. PPV reference calculation
where SBoost represents the switching state of the boost converter that
can be either 1 or 0. Using the first order approximation for the deri- In the RPM, PPV ∗
is calculated based on the voltage sag depth, and
vative in (2), and shifting the output current into one future sample the PV system capacity as it was shown in Fig. 2. However, in normal
∗
(Yaramasu et al., 2015; Sultana et al., 2017), the predicted current in conditions, the PPV must track the optimal operation point at all times
discrete time notation is given by to guarantee maximum power delivery. Using the approach described
Ts in Section 2.2., conventional methods such as P&O and IncCond (Alik
iL (k + 1) = [VPV (k ) + (SBoost −1) Vdc (k )] + iL (k ), and Jusoh, 2017; Gad et al., 2017) can be adapted to track MPP using
L (3)
the FCS–MPC principle. Such methods originally vary VPV or the duty
where Ts corresponds to the sampling time. If the sampling time is small cycle by Δ steps to climb to the MPP; therefore, if such methods were
enough, it can be assumed that VPV (k ) ≅ VPV (k + 1) . Therefore, one can implemented following the FCS–MPC approach, the variation term
multiply both sides of (3) by VPV (k ) to solve in terms of PV power must be modified to update instead PPV ∗
in (6).
Ts VPV (k ) The approach followed in this paper is different since it considers
PPV (k + 1) = [VPV (k ) + (SBoost −1) Vdc (k )] + PPV (k ). the natural behavior of the boost converter and the optimal switching
L (4)
state provided in (6). To explain the methodology, consider first the
Eq. (4) supposes no power loss on CPV. This valid if CPV is chosen large data presented on Table 1 and Fig. 4. In the steady state condition, one
enough and the sampling time given by Ts is small. If the conditions can assume that PCPV ≈ 0 ; therefore, the term [PPV∗
(k + 1)−PPV (k )], now
previously stated were not satisfied, (4) can be modified to include CPV referred as ΔP in (6), can be either positive or negative. If ΔP > 10 , the
dynamics as
Ts VPV (k ) Table 1
PPV (k + 1) = [VPV (k ) + (SBoost −1) Vdc (k )] + PPV (k ) Simulation parameters.
L
dV
−CPV VPV PV . Variable Description Value
dt (5)
PV Parameters
VOC PV Open Circuit Voltage (V) 32.9
ISC PV Short Circuit Current (A) 8.21
Number of PV Modules in Series—Parallel 6–4
VMPP PV String Array MPP Voltage (V) 169.1
IMPP PV String Array MPP Current (A) 28.4
PPV Power PV array (kW) 4.8
DC/DC Converter and Controller Parameters
L Boost Filter Inductance (mH) 10
CPV PV Capacitance (μF) 512
Cdc DC-Link Capacitance (μF) 380
TS Sampling Time (μs) 25
Vdc Vdc bus voltage reference (V) 400
Fig. 3. Circuit diagram of the PV system.
681
F. Diaz Franco et al. Solar Energy 158 (2017) 679–686
On the other hand, the left-side of Fig. 5 shows the normal operation
of the PV system, this is MPPT enabled. The Update MPP block is set to
true by an irradiation change detection—given by δ—or using a
time–based approach. If the MPP reference needs to be updated, a
∗
perturbation step is applied making PPV = 1.05PN which forces the
power switch to conduct as explained in Fig. 4. If during this time a
drop in PPV is detected, it means that the current PV operation point is
∗
located at the left side of the MPP. In such case, PPV is changed to, for
∗
instance PPV = 0 , which forces the system to correct the direction and
climb to the MPP. Finally, the Find Peak step can be easily implemented
by storing previous PPV points as it approaches the MPP. Once the PPV
∗
starts to decline, the PPV can be obtained as the maximum value from
the stored values.
The generator–side controller can be summarized in the following
∗
Fig. 4. Operation principle to select PPV . steps: (a) measurement of the PV electrical signals (b) identification of
∗
the converter operation mode provides PPV , if normal, the reference is
minimization function will choose to close the switch for all values of calculated following the steps on left–side of Fig. 5; (c) SBoost is calcu-
Vpv. On the other hand, if is ΔP < −25, the result will be to leave it open. lated using (6); (d) the optimal switching state is computed by means of
Therefore, the response of the boost converter using FCS-MPC can be the minimization function given by (7); and finally, (e) the optimal
used to control the PV operation point path on the PPV vs VPV curve as state is applied to the boost converter.
shown in Fig. 4 so that the MPP is always on the track.
A flow chart shown in Fig. 5 summarizes the process for the cal-
∗ 2.4. Simulation results
culation of PPV during the normal operation of the PV system, and
through grid voltage sag conditions. Both functions are explained next:
The performance of the proposed FCS-MPC for PPV control is eval-
The right–side of Fig. 5 corresponds to the RPM operation. During
voltage sag conditions, the PV power reference must be careful selected uated using MATLAB/Simulink with the modeling of the PV system in
PLECS blockset. The analytical five–parameter PV model has been used
to avoid overcurrent and eventually disconnection from the grid.
∗
Therefore, PPV is chosen to comply grid code requirements presented in for the implementation of irradiance variations as outlined in (Djoudi
Gherbi et al., 2017; Chouder et al., 2012). The system parameters for
Fig. 1, and to satisfy the inverter power rating as shown in Fig. 2.
simulation are listed in Table 1. The PV system parameters were chosen
based on the power rating of a testbed available in the lab, additional
information about the testbed can be found in (Hasanzadeh et al., 2014;
Vu et al., 2017); whereas the control parameter Ts is selected small
enough to satisfy the condition described in Section 2.2.
A first test is performed to show the performance of the proposed
∗
controller to follow a designed PPV profile and the results are shown in
Fig. 6. In this test, both solar irradiation and temperature are main-
tained constant to 1 kW m−2 and 25 °C, respectively. The profile chosen
tests the ability of the controller to respond to fast transients like those
expected during voltage sag conditions. It is observed that the system
has good tracking performance, even when a large variation in the
∗
Fig. 5. PPV calculation based on FCS-MPC principle. Fig. 6. Simulation of the boost converter for PV power tracking in the RPM.
682
F. Diaz Franco et al. Solar Energy 158 (2017) 679–686
Fig. 7. MPPT response under irradiation change for different updating time. (a) 50 ms, (b) 10 ms.
power reference is required as shown at time 0.55 s. Moreover, from the MPP, the impact on the harvested power of similar VPV fluctuations is
∗
VPV and IPV curves, it is observed that the PPV is always achieved on the less noticeable.
right–side of the MPP where the system is more sensitive to VPV var-
iations. 3. CHIL experimentation methodology
A second test is performed to verify the MPPT method under solar
irradiation change conditions and 25 °C constant temperature, the results In order to verify the efficacy of the proposed methodology, the
are presented in Figs. 7 and 8. To better illustrate the proposed method, real–time implementation of the PPV controller is performed using RTDS
50 ms is initially chosen has the MPP update time and the results are Technologies product. CHIL is a combination of computer simulation in
shown in Figs. 7a and 8. The figures indicate how the algorithm perturbs real–time and a controller under test (CUT) that is connected to the
∗
the system to update PPV and tracks this value for the next 50 ms. The real–time simulator. The methodology offers a more realistic environ-
enhanced performance is obtained by decreasing the update time so that ment than software simulation because it can help to identify limita-
the tracking of the MPP looks like a continuous–time reference signal. The tions of the CUT, and verify CUT performance for an electrical system
results of decreasing the updating time parameter can be seen in Fig. 7b whose design has not yet been prototyped (Edrington et al., 2015).
where a 10 ms update time was chosen. It can be clearly seen the ad- The generator–side of the system depicted in Fig. 3 and the para-
vantage of the proposed methodology compared to other techniques meters given in Table 1 are modeled on RSCAD, which is a built–in
where the MPP is reached by varying the duty cycle by Δ steps. graphical user interface software that manages the real–time simulator.
The nature of FCS-MPC method forces a change on the switching Details on modeling a PV system on RTDS/RSCAD are not addressed in
state when the current converter state does not deliver minimum error; this document; however, interested readers are encouraged to read
therefore, it is expected that the switching frequency not to be constant. (Islam et al., 2015) that describes in detail RTDS/RSCAD modeling
Table 2 shows the average switching frequency Fsw for the simulated principle, more specifically, large- and small–time step interfacing ap-
cases. It can be seen that although all cases use the same sampling time, proach.
there is a large variation on the Fsw for all cases. This is due to the PV The proposed control is implemented using a Texas Instrument
operating point on the PPV vs VPV curve. The results presented in Fig. 6 C2000 F28377S Launchpad evaluation board that runs at speed of
showed that the operating point is located at the right–side of the MPP; 200 MHz. The F28377S has 14 analog–to–digital converter ADC chan-
therefore, small variations on VPV could lead to larger fluctuations on nels available used to interface with the real–time simulator, one di-
IPV. On the other hand, as the PV operating point is located closer to the gital–to–analog DAC channel used for debugging purposes, two Control
Area Network (CAN) ports used to connect it to a local network where
measurement devices and other equipment are connected. The con-
troller program is developed in Matlab Simulink using Embedded Coder
with support for Texas Instruments C2000 family.
The experiment setup is shown in Fig. 9. The CUT is connected to
Table 2
Average switching frequency for the simulated cases.
683
F. Diaz Franco et al. Solar Energy 158 (2017) 679–686
RTDS using a GTDI card that receives the switch control action from the
CUT, and a GTIO card that provides the analog signals to be measured
by the digital signal processor. All these signals are multiplexed to a
Yokogawa scope for data recording and analysis. Finally, both cards
transfer the data to and from the real–time simulator located in a dif-
Fig. 9. Experiment setup.
ferent lab using fiber optic cables.
684
F. Diaz Franco et al. Solar Energy 158 (2017) 679–686
5. Conclusions
Acknowledgment
Fig. 12. Experimental results for MPPT operation scope image. Authors thank Florida State University and the Center for Advanced
Power Systems (CAPS) for the hardware and software provided to
conduct this research, as well as the Administrative Department of
Science, Technology and Innovation of Colombia (COLCIENCIAS) and
4. Results and discussion Fulbright Colombia for F.D.F scholarship.
The first CHIL test is conducted to validate the PPV tracking response References
of the system. The recorded scope signals and the scaled physical
magnitudes are shown in Fig. 10 and Fig. 11 respectively. From the Technical requirements for the connection to and parallel operation with low-voltage
results, the performance is similar to the results shown in Fig. 6, where distribution networks, 2011. VDE-AR-N 4105.
Hasanzadeh, A., Edrington, C.S., Stroupe, N., Bevis, T., 2014. Real-time emulation of a
the controller presents a fast-transient response, and it always operates high-speed microturbine permanent-magnet synchronous generator using multiplat-
the PV array at the point in which it delivers the required power. form hardware-in-the-loop realization. IEEE Trans. Ind. Electron. 61, 3109–3118.
Moreover, it can be seen clearly in Fig. 10 zoom2 area how the Sangwongwanich, A., Yang, Y., Blaabjerg, F., 2016. High-performance constant power
generation in grid-connected pv systems. IEEE Trans. Power Electron. 31,
FCS–MPC maintains an optimal switching state of the converter when a 1822–1825.
fast response is required. This improves the settling time and reduces Urtasun, A., Sanchis, P., Marroyo, L., 2013. Limiting the power generated by a photo-
the power loses produced by a constant PWM switching. Additional voltaic system. 10th International Multi-Conferences on Systems, Signals & Devices
2013 (SSD13), 1–6.
analysis showed that the Fsw is 6.1 kHz, which is below the results Aguilera, R.P., Lezana, P., Quevedo, D.E., 2013. Finite-control-set model predictive
shown in Table 2. This decrement on the switching frequency for ex- control with improved steady-state performance. IEEE Trans. Ind. Inform. 9,
perimental results is in concordance with previous results published by 658–667.
Alik, R., Jusoh, A., 2017. Modified Perturb and Observe (P&O) with checking algorithm
other authors (Choi and Lee, 2015; Panten et al., 2016).
under various solar irradiation. Sol. Energy 148, 128–139.
A second experiment was performed to verify the proposed MPPT Balasankar, R., Arasu, G.T., Christy Mano Raj, J.S., 2017. A global MPPT technique in-
operation, and the results are shown in Figs. 12 and 13. After a tuning voking partitioned estimation and strategic deployment of P&O to tackle partial
process, 1 ms update time was chosen to improve the performance of the shading conditions. Sol. Energy 143, 73–85.
Bendib, B., Belmili, H., Krim, F., 2015. A survey of the most used MPPT methods: con-
MPPT method. The solar irradiation rate for decreasing and increasing ventional and advanced algorithms applied for photovoltaic systems. Renew. Sustain.
were set on RSCAD to be −100 and 200 W s−1 respectively. For the ir- Energy Rev. 45, 637–648.
radiation profile implemented, it is observed that the proposed algorithm Bordons, C., Montero, C., 2015. Basic principles of MPC for power converters: bridging
the gap between theory and practice. IEEE Ind. Electron. Mag. 9, 31–43.
converges quickly to the MPP as the atmospheric conditions vary. Edrington, C.S., Steurer, M., Langston, J., El-Mezyani, T., Schoder, K., 2015. Role of
power hardware in the loop in modeling and simulation for experimentation in power
and energy systems. Proc. IEEE 103, 2401–2409.
Choi, D., Lee, K., 2015. Dynamic performance improvement of AC/DC converter using
model predictive direct power control with finite control set. IEEE Trans. Ind.
Electron. 62, 757–767.
Chouder, A., Silvestre, S., Sadaoui, N., Rahmani, L., 2012. Modeling and simulation of a
grid connected PV system based on the evaluation of main PV module parameters.
Simul. Model. Pract. Theor. 20, 46–58.
Dileep, G., Singh, S.N., 2015. Maximum power point tracking of solar photovoltaic system
using modified perturbation and observation method. Renew. Sustain. Energy Rev.
50, 109–129.
Djoudi Gherbi, A., Hadj Arab, A., Salhi, H., 2017. Improvement and validation of PV
motor-pump model for PV pumping system performance analysis. Sol. Energy 144,
310–320.
Ezinwanne, O., Zhongwen, F., Zhijun, L., 2017. Energy performance and cost comparison
of MPPT techniques for photovoltaics and other applications. Energy Procedia 107,
297–303.
Diaz-Franco, F., Vu, T., El Mezyani, T., Edrington, C.S., 2016. Low-voltage ride-through
for PV systems using model predictive control approach. 2016 North American Power
Symposium (NAPS), 1–6.
Gad, H.H., Haikal, A.Y., Arafat Ali, H., 2017. New design of the PV panel control system
using FPGA-based MPSoC. Sol. Energy 146, 243–256.
Harrag, A., Messalti, S., 2015. Variable step size modified P&O MPPT algorithm using GA-
based hybrid offline/online PID controller. Renew. Sustain. Energy Rev. 49,
1247–1260.
Hudson, R., Heilscher, G., 2012. PV grid integration – system management issues and
utility concerns. Energy Procedia 25, 82–92.
IEEE Std 1547a, 2014. IEEE Standard for Interconnecting Distributed Resources with
Electric Power Systems - Amendment 1. IEEE Std 1547a–2014 (Amendment to IEEE
Std 1547–2003), 1–16.
Islam, G., Muyeen, S.M., Al-Durra, A., Hasanien, H.M., 2015. RTDS implementation of an
Fig. 13. Experimental results for MPPT operation. improved sliding mode based inverter controller for PV system. ISA Trans.
685
F. Diaz Franco et al. Solar Energy 158 (2017) 679–686
Ko, S., Chao, R., 2012. Photovoltaic dynamic MPPT on a moving vehicle. Sol. Energy 86, L.E., Anaya-Ruiz, G.A., 2017. The Proportional-Values Modulation (PVM), a tech-
1750–1760. nique for improving efficiency and power density of bidirectional DAB converters.
López, J., Seleme Jr., S.I., Donoso, P.F., Morais, L.M.F., Cortizo, P.C., Severo, M.A., 2016. Electr. Power Syst. Res. 144, 280–289.
Digital control strategy for a buck converter operating as a battery charger for stand- Sheik Mohammed, S., Devaraj, D., Imthias Ahamed, T.P., 2016. A novel hybrid maximum
alone photovoltaic systems. Sol. Energy 140, 171–187. power point tracking technique using perturb & observe algorithm and learning
Panten, N., Hoffmann, N., Fuchs, F.W., 2016. Finite control set model predictive current automata for solar PV system. Energy 112, 1096–1106.
control for grid-connected voltage-source converters with LCL Filters: a study based Sultana, W.R., Sahoo, S.K., Sukchai, S., Yamuna, S., Venkatesh, D., 2017. A review on
on different state feedbacks. IEEE Trans. Power Electron. 31, 5189–5200. state of art development of model predictive control for renewable energy applica-
Nanou, S.I., Papathanassiou, S.A., 2014. Modeling of a PV system with grid code com- tions. Renew. Sustain. Energy Rev. 76, 391–406.
patibility. Electr. Power Syst. Res. 116, 301–310. Vu, T.V., Perkins, D., Diaz, F., Gonsoulin, D., Edrington, C.S., El-Mezyani, T., 2017.
Narimani, M., Wu, B., Yaramasu, V., Zargari, N.R., 2015. Finite Control-Set Model Robust adaptive droop control for DC microgrids. Electr. Power Syst. Res. 146,
Predictive Control (FCS-MPC) of Nested Neutral Point-Clamped (NNPC) Converter. 95–106.
IEEE Trans. Power Electron. 30, 7262–7269. Xia, C., Liu, T., Shi, T., Song, Z., 2014. A simplified finite-control-set model-predictive
Kiran, P.B.S., Manjunath, K., Sarkar, V., 2015. Limited power control of a single-stage grid control for power converters. IEEE Trans. Industr. Inf. 10, 991–1002.
connected photovoltaic system. 2015 Annual IEEE India Conference (INDICON), 1–6. Yang, F., Yang, L., Ma, X., 2014. An advanced control strategy of PV system for low-
Rezk, H., Eltamaly, A.M., 2015. A comprehensive comparison of different MPPT techni- voltage ride-through capability enhancement. Sol. Energy 109, 24–35.
ques for photovoltaic systems. Sol. Energy 112, 1–11. Yaramasu, V., Rivera, M., Narimani, M., Wu, B., Rodriguez, J., 2015. High performance
Rivera, M., Rodriguez, J., Yaramasu, V., Wu, B., 2013. A simple current control strategy operation for a four-leg NPC inverter with two-sample-ahead predictive control
for two-level four-leg inverters: The model predictive approach. Power Engineering, strategy. Electr. Power Syst. Res. 123, 31–39.
Energy and Electrical Drives (POWERENG), 2013 Fourth International Conference Young, H.A., Perez, M.A., Rodriguez, J., Abu-Rub, H., 2014. Assessing finite-control-set
on, 46–51. model predictive control: a comparison with a linear current controller in two-level
Rodriguez-Rodrıguez, J.R., Moreno-Goytia, E.L., Venegas-Rebollar, V., Ugalde-Caballero, voltage source inverters. IEEE Ind. Electron. Mag. 8, 44–52.
686