Comparison of control strategies for grid-connected photovoltaic systems during unbalanced voltage dips Edmar F. Cota1,3, Lucas S. Xavier1, Allan F. Cupertino1,2,3 and Heverton A. Pereira1,3 1 Gerência de Especialistas em Sistemas Elétricos de Potência Universidade Federal de Viçosa Av. P. H. Rolfs s/nº, 36570-000 Viçosa, MG, Brazil edmar.elt@gmail.com, lsantx@gmail.com 3 2 Departamento de Engenharia de Materiais Centro Federal de Educação Tecnológica de Minas Gerais Av. Amazonas 5253, 30421-169 Belo Horizonte, MG, Brazil allan.cupertino@yahoo.com.br Graduate Program in Electrical Engineering Federal University of Minas Gerais Av. Antônio Carlos 6627, 31270-901 Belo Horizonte, MG, Brazil hevertonaugusto@yahoo.com.br Abstract— Unbalanced voltage dips are common disturbances in power systems. Therefore, grid-connected photovoltaic systems need to be able to work during these disturbances. Conventional control strategies do not consider the negative sequence current control, which is an important limitation. In this context, this paper compares through simulations four strategies of negative sequence current control: Notch filter based strategy, measured signals decoupling based strategy, reference and error signals decoupling based strategy and resonant controllers based strategy. All simulations are performed in Matlab/Simulink environment. Representation of point of common coupling voltage (PCC) uses experimental data of an unbalanced voltage sag. Comparisons during unbalanced voltage sags and the impact of voltage harmonics on techniques are analyzed. Obtained results shows resonant controllers based strategy has a superior behavior, in terms of robustness and simplicity. I. INTRODUCTION Dispersed generations (DGs) have put significant challenges to voltage regulation in distribution systems. Among these renewable sources, photovoltaic (PV) technology is one of the main candidates to play an important role in the future of electricity generation [1]. A system with significant PV penetration can be seriously affected with the variation of generation caused by cloud movement that may result in voltage variation condition [2]. Serious voltage drops may occur due to the intermittent loss of PV generation during cloudy days. Interconnection of large scale PV into the power system demands the continuous revision of grid codes and power quality standards in order to fulfill restrictive constraints. Furthermore, most host utilities also require that PV must tolerate system disturbances. The increasing installation of single-phase solar systems contribute to the system unbalance [3]. Ideally, a network voltage is balanced in normal conditions. But, the most usual The Brazilian agencies CAPES, CNPq and FAPEMIG support this work. grid faults are two and single-phase faults, that although less severe than a three-phase fault, cause voltage unbalance, which leads to voltage oscillation due to appearance of negative sequence voltages and currents [4], [5], [6], [7]. Under these conditions, the dc-link voltage harmonics and odd ac input currents harmonics will appear [5], [8]. The appearance of a negative sequence current component deteriorates the control performance [8]. Even for well-balanced three-phase source voltages, different parameters of the power lines or unbalanced load conditions will also cause unbalanced voltages to the gridconnected converters. Under unbalanced input voltages, desirable features, such as unity power factor, constant dc bus voltage, and sinusoidal input currents cannot be guaranteed using those methods developed under balanced voltage conditions [6]. In such a case, voltage ripples in the dc bus arise, which may affect negatively the inverters connected through the grid. The dc link voltage oscillation provokes low-order harmonics in the current injected by the inverter into the grid and should be minimized [4]. Other consequences of unbalances are increase of electric losses, rise of temperature in electrical machines and transformers, appearance of torque pulsations, noise and ripple in rectifiers, and in cases of zero sequence currents can affect the protection system [7], [1], [9]. Nowadays, there are grid operators that allow providing unbalanced currents so that the power at the PCC remains constant. This is very useful in weak grids because it allows connecting all loads to a balanced PCC voltage, which is especially important for sensitive loads that could disconnect if the AC voltage does not meet restrictive voltage standards [8]. Thus, PV systems operating as a high-reliability current source must contributed to balance the grid voltage [1]. In this work, some technics to improve the current injected by three phases PV system are analyzed and the main contributions of each one are discussed during unbalanced voltage sag. II. MODELING OF THE SYSTEM Figure 1 presents the studied system. Photovoltaic panels are connected to the grid through a three phase inverter modulated by the space-vector pulse-width modulation (SVPWM) technique. In order to reduce the harmonics generated by converter switching, LCL filter topology is used. The design of LCL filter follows the methodology proposed by [10]. The PV panel electrical model is showed in Figure 2. The series resistance ( ) and parallel resistance ( ) represent the voltage drop when charge migrates from the electrical contacts and the reverse leakage current of diode, respectively. According to [11], the value of these resistances can be obtained through the parameters informed by manufactures. More details about this model can be found in references [11] and [12]. ⎧ = ⎨ ⎩ =− − + − =− − + − = − − , + − , , , (2) In equations (1) and (2), the symbols + and – represents the positive and negative sequence components. The indexes and represent the terms of direct and quadrature axis. is grid current, is inverter voltage and is PCC voltage. = + , = + and is the grid fundamental frequency. The DC bus dynamic is given by: = (3) − Where C is capacitance of DC bus, is the solar panel current and is the current that flows to the inverter. The term is a disturbance and is disregarded. Furthermore, considering a balanced grid voltage and negleging the converter losses, equation (2) is obtained: ∗ = 3 2 , (4) A maximum power point tracker (MPPT) algorithm calculates the reference of the DC bus voltage in order to extract the maximum power of the solar panel. There are many structures of MPPT in literature. This work implement the incremental conductance algorithm proposed in [14]. Next section presents the control strategies in details. Figure 2 – Photovoltaic panel electrical model. In this work, all control strategies are implemented in synchronous reference frame. Therefore, a PLL (Phase-Locked Loop) structure is necessary. Due to its robustness during unbalanced voltages and harmonics, the Dual Second Order Generalized Integrator based PLL (DSOGI-PLL) is used. Details about this structure are found in [13]. The following points are considered in order to obtain the inverter dynamic model: • Only the fundamental considered; • The capacitor of LCL filter is considered an open circuit at fundamental frequency; • The point of common couple (PCC) voltage is unbalanced. However, the dynamic of the zero component is neglected. dynamic + (1) Figure 1 – Topology of a grid-connected photovoltaic system. frequency − is In this situation, removing the coupling of symmetric components, it is possible to obtain the following system of equations in synchronous reference frame [14]: III. CONTROL STRATEGIES A. Conventional control strategy (CONV) Firstly, conventional control strategy is presented. This structure neglects the negative sequence components of the PCC voltage and is based on a cascade control with aninner loop, faster, controlling the currents of the inverter and an outer loop, slower, controlling the dc bus voltage. The crossed couplings of dynamic equations presented in (1) are removed through feed-forward actions. Figure 3 presents the block diagram of conventional control strategy. All compensators are proportional-integral actions designed through pole assignment method. The gains of current compensators are [15]: , =2 (5) , =2 (6) Where is the crossover frequency of current loop. This value is limited up to 10 times smaller than the switching frequency [15]. The block diagram model of the DC bus voltage control is presented in Figure 4. Considering DC bus voltage loop 10 times slower than current loop and faster than MPPT algorithm, it is possible to obtain the following gains of the DC bus compensator [15]: current control loops. The block diagram of this structure is presented in Figure 6. Influence of the filter in current dynamic response and increase in complexity of the control algorithm are drawbacks of this structure. Figure 3 – Control block diagram for conventional control strategy. Figure 4 – DC bus voltage control model. = 2 = Where = ∗ 4 ( + ) Figure 5 – Control block diagram for NOTCH, MSD and RESD strategies. (7) (8) . This control structure does not compensate the negative sequence components during unbalanced voltage dips. Next control strategies will compensate the negative sequence components in order to improve the performance of this technique. However, the controller gains are maintained for all strategies. B. Notch filter based strategy (NOTCH) The largest limitation of the conventional control strategy is the current controller, which cannot compensate the oscillations of 120 Hz generated by interactions between positive and negative sequence. In this situation, during unbalanced voltage dips the inverter will inject unbalanced currents into the grid. One possible solution is controlling separately the positive and negative current as shown in Figure 5. The negative sequence structure is similar to the positive sequence. In this situation, the same gain controllers are used in this structure. However, a structures able to detect the variables , , and separately are necessary in order to implement the control algorithm shown in Figure 5. In order to remove the oscillations in the variables and and detect the positive and negative currents, reference [16] proposes use of notch filters. These filters remove the 120 Hz components and detects the current signals that will used in Figure 6 – Positive and negative sequence detector based on notch filters (NOTCH). C. Measured signals decoupling based method (MSD) This strategy is based on the block diagram of Figure 5 and detects the positive and negative sequence based on a decoupling of the measured signals. The 120 Hz oscillations in positive sequence current are generated by the value of negative sequence component referred to the positive synchronous reference frame. Therefore, the negative component detected pass through a first order filter that collects its mean value. Then, this signal is referred to the positive synchronous reference frame through the transformation and subtracted of positive sequence as feed-forward action. A similar methodology is used in negative sequence detection. Complete detection structure is presented in Figure 7. Influence of the filter in current dynamic response and increase in complexity of the control algorithm are also drawbacks of this structure. Furthermore, when current compensator has error in steady state, this technique cannot remove all oscillations [17]. ( )= D. Reference and error signals decoupling based method (RESD) Other option is use reference and error signals instead of measured signals to implement the decoupling network, as shown in Figure 8 [17]. This strategy results in a complete decoupling, even current compensator has error in steady state. As drawbacks, this strategy also increases the complexity of control algorithm. + + 2 + Where is resonant gain and this case, 120 Hz). (9) is resonance frequency (in Figure 9 – Control block diagram for RES strategy. IV. Figure 7 – Positive and negative sequence detector based on measured signals decoupling (MSD). METHODOLOGY Photovoltaic array considered presents 5 strings of 13 panels. Model KD250 manufactured by Kyocera is used. Table I present the studied system parameters. All simulations was performed in Matlab/Simulink environment. The control techniques described are implemented with sampling frequency of 12 kHz (2 times the switching frequency of the converter). Digital filters and integrators are discretized through trapezoidal method. All simulations considered constant irradiance at 1000 / on the panels. In order to compare the techniques, experimental data of unbalanced voltage dip are used. These data are presented in Figure 10. The harmonic spectrum of this signal around 1.8 seconds are presented in Figure 11. Therefore, results analyze the techniques performance during unbalanced voltage dip and influence of harmonics in each control strategy. TABLE I. PARAMETERS OF THE SIMULATED SYSTEM: POWER INVERTER. First inductance of LCL filter 2,9 Second inductance of LCL filter 95 Capacitance of LCL filter 45 Damping resistor of LCL filter 2,9 Ω DC bus Capacitance 3 Figure 8 – Positive and negative sequence detector based on reference and error signals decoupling (RESD). Switching frequency 6 Rated Power 17 E. Resonant Controller based method (RES) Other solution to compensate the oscillations generated by the negative sequence is using resonant controllers. In fact, a proportional integral resonant controller (PI-RES) is implemented in current loops, as shown in Figure 9. In this situation, resonant part of the controller compensates the negative sequence without detects separately positive and negative current components. The transfer function of the PIR controller is given by: Grid Voltage 220 500 200 400 VdcCONV 0 300 480 VdcMSD 460 VdcRESD 440 VdcRES VdcNOTCH -200 2 2.1 Vdc [V] VCONV 200 2.05 100 420 400 407.5 380 406.5 0 360 405.5 340 -100 320 0 -200 0.27 0.271 1 1 2 3 4 5 5 0 0 4 5 3 4 5 450 400 5 10 15 Harmonic order 20 25 MPPT Signal DC Voltage 350 300 0 0 3 500 Vdc CONV [V] VPhase A [%] VPhase B [%] VPhase C [%] Figure 10 – Experimental PCC voltage profile used in simulations: Complete waveform and detail of voltage dip at 2.05 seconds. 5 2 (a) Time [s] 0 0 0.273 Time [s] 0 5 0.272 0 1 2 Time [s] (b) 5 10 15 Harmonic order 5 10 15 Harmonic order 20 20 25 25 Figure 11 – Spectrum of PCC voltage. V. RESULTS Dynamic behavior of dc bus voltage are presented in Figure 12 (a). All Techniques have similar transient response and a small difference in steady state ripple. During the unbalanced voltage dip, the dc bus ripple increases. During voltage dip, dc bus voltage increases because the inverter cannot injects rated power in this situation, due to overcurrent limitation. Injected power decreases and MPPT algorithm try to reduce voltage reference. In this situation will appear error in steady state on dc bus voltage control loop. When the voltage dip finishes, the dc bus voltage follow the reference, which will return the system to the maximum power point. This behavior is presented in Figure 12 (b). Figure 12 – DC bus voltage dynamic behavior during voltage dip: (a) Comparison of 5 techniques; (b) DC bus voltage and reference calculated by MPPT algorithm for CONV strategy. Dynamic behavior of current components for CONV control strategy is presented in Figure 13. Negative components, which are not compensated, appears as oscillations in variables. On the other hand, the RES control technique remove the oscillations due to the resonant controller, as shown in Figure 14. Figure 15, Figure 16 and Figure 17 shows the positive and negative sequence components for MSD, RESD and NOTCH strategies. In these three techniques, negative grid current components are controlled to zero. RESD technique has a higher oscillation. Furthermore, MSD and NOTCH techniques presented similar dynamics. Behavior of active and reactive power injected to the grid are presented in Figure 18. During the voltage dip, the CONV control strategy has the smaller value of active power. Furthermore, MSD and NOTCH control strategies have a slower response if compared with CONV strategy. In terms of reactive power, the NOTCH control strategy has the worst behavior in transient response with a higher overshoot. The RESD control strategy has the best behavior with small overshoot and fast response. Finally, Figure 19 shows the current harmonic spectrum for each control technique at time 1.8 seconds. RES control strategy has the smaller current distortion followed by NOTCH and MSD control strategies. RESD control strategy has the worst behavior due to a large component of 17th harmonic. 100 100 80 80 60 60 40 Id-,q- MSD [A] Id,q CONV [A] 40 20 0 20 0 -20 -20 -40 -40 -60 -60 -80 -100 0 1 2 3 Id CONV -80 Iq CONV -100 4 5 Time [s] Figure 13 – Direct and Quadrature current dynamic behavior for CONV technique. Id- MSD Iq- MSD 0 1 2 3 4 5 Time [s] (b) Figure 15 – Direct and Quadrature current dynamic behavior for MSD technique: (a) Positive sequence; (b) Negative sequence. 100 100 80 80 60 60 40 Id+,q+ RESD [A] Id,q RES [A] 40 20 0 -20 20 0 -20 -40 -40 -60 -60 Id RES -80 Iq RES -100 0 1 2 3 4 Id+ RESD -80 Iq+ RESD -100 0 1 2 5 3 4 5 Time [s] Time [s] (a) Figure 14 – Direct and Quadrature current dynamic behavior for RES technique. 100 Id- RESD 80 60 80 40 Id-,q- RESD [A] 100 60 Id+,q+ MSD [A] 40 20 20 0 -20 -40 0 -60 -20 -80 -40 -100 -60 Id+ MSD -80 -100 Iq- RESD 1 2 3 Time [s] (a) 4 1 2 3 4 5 Time [s] Iq+ MSD 0 0 (b) 5 Figure 16 – Direct and Quadrature current dynamic behavior for RESD technique: (a) Positive sequence; (b) Negative sequence. 100 2 80 1 60 0 20 Q [KVAr] Id+,q+ NOTCH [A] 40 0 -20 QCONV -60 Id+ NOTCH -80 -3 Iq+ NOTCH 0 1 2 3 4 -4 5 Time [s] 0 QMSD QRESD QNOTCH 2.24 1 2 3 4 5 (b) Figure 18 – Active and reactive power dynamic behavior: (a) Active power; (b) Reactive power. 100 Id- NOTCH 80 Iq- NOTCH 4 IPhase A [%] 60 40 20 2 0 0 CONV MSD RESD (a) NOTCH RES CONV MSD RESD (b) NOTCH RES CONV MSD RESD (c) NOTCH RES -20 4 IPhase B [%] -40 -60 -80 -100 QRES 2.28 Time [s] (a) Id-,q- NOTCH [A] 0 -0.08 -0.16 -0.24 -0.32 -0.4 2.2 -2 -40 -100 -1 2 0 0 1 2 3 4 5 Time [s] Figure 17 – Direct and Quadrature current dynamic behavior for NOTCH technique: (a) Positive sequence; (b) Negative sequence. IPhase C [%] 4 (b) 2 0 20 Figure 19 – Current harmonic spectrum for each control technique at time 1.8 seconds. 15 VI. P [KW] 10 PCONV 5 PMSD 13.2 PRESD 0 PNOTCH 12.8 PRES -5 12.4 2.25 -10 0 1 2.3 2.35 2 3 Time [s] (a) 4 5 CONCLUSIONS This paper presented a comparison of 4 techniques of control negative sequence in a PV inverter and the conventional strategy. Comparisons during unbalanced voltage dips and harmonics are performed. During unbalanced voltage dips, RES and RESD control techniques presents a better performance in terms of reactive power dynamic response. However, all techniques controlled the negative sequence component, which appeared as a large fluctuation in CONV control strategy. Harmonic spectrum of injected current indicates RESD technique is more influenced by voltage harmonics than other techniques. In this situation, the RES control strategy obtains the best behavior in terms of unbalanced voltage dips and harmonic distortion in PCC voltage. Furthermore, this technique is simpler than others, because it do not control separately the positive and negative sequence. REFERENCES [1] J. Miret, M. Castilla, A. Camacho, L. G. d. Vicuña and J. Matas, "Control Scheme for Photovoltaic Three-Phase Inverters to Minimize Peak Currents During Unbalanced Grid-Voltage Sags," IEEE TRANSACTIONS ON POWER ELECTRONICS, vol. 27, no. 10, pp. 4262-4271, OCTOBER 2012. [2] N. Daratha, B. Das and J. Sharma, "Coordination Between OLTC and SVC for Voltage Regulation in Unbalanced Distribution System Distributed Generation," IEEE TRANSACTIONS ON POWER SYSTEMS, vol. 29, no. 1, pp. 289-299, JANUARY 2014. [3] X. Su, M. A. S. Masoum and P. J. Wolfs, "Optimal PV Inverter Reactive Power Control and Real Power Curtailment to Improve Performance of Unbalanced Four-Wire LV Distribution Networks," IEEE TRANSACTIONS ON SUSTAINABLE ENERGY, vol. 5, no. 3, pp. 967-977, JULY 2014. [4] J. Eloy-García, S. Arnaltes and J. R.-A. Amenedo, "Direct power control of voltage source inverters with unbalanced grid voltages," IET Power Electronics, vol. 1, no. 3, p. pp. 395–407, April 2008. [5] D. Roiu, R. I. Bojoi, L. R. Limongi and A. Tenconi, "New Stationary Frame Control Scheme for Three-Phase PWM Rectifiers Under Unbalanced Voltage Dips Conditions," IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, vol. 46, no. 1, pp. 268-277, JANUARY/FEBRUARY 2010. [6] Z. Li, Y. Li, P. Wang, H. Zhu, C. Liu and W. Xu, "Control of ThreePhase Boost-Type PWM Rectifier in Stationary Frame Under Unbalanced Input Voltage," IEEE TRANSACTIONS ON POWER ELECTRONICS, vol. 25, no. 10, pp. 2521-2530, OCTOBER 2010. [7] A. Mohd, E. Ortjohann, N. Hamsic, W. Sinsukthavorn, M. Lingemann, A. Schmelter and D. Morton, "Control strategy and space vector modulation for three-leg four-wire voltage source inverters under unbalanced load conditions," IET Power Electronv, vol. 3, no. Iss. 3, p. 323–333, 2010. [8] A. Rodríguez, E. J. Bueno, Á. Mayor, F. J. Rodríguez and A. GarcíaCerrada, "Voltage Support Provided by STATCOM in Unbalanced Power Systems," Energies, no. 7, pp. 1003-1026, 2014. [9] M. Chomat and L. Schreier, "Control method for DC-link voltage ripple cancellation in voltage source inverter under unbalanced three-phase voltage supply conditions," IEE Proc.-Electr. Power Appl., vol. 152, no. 3, pp. 494-500, May 2005. [10] M. Liserre, F. Blaabjerg and S. Hansen, "Design and Control of an LCLFilter-Based Three-Phase Active Rectifier," IEEE Transactions on Industry Applications, vol. 41, pp. 1281 - 1291 , Setembro 2005. [11] M. G. Villalva, J. R. Gazoli and E. R. Filho, "Comprehensive Approach to Modeling and Simulation of Photovoltaic Arrays," IEEE Transactions on Power Electronics, vol. 24, no. 1, pp. 1198-1208, March 2009. [12] E. M. d. S. Brito, A. F. Cupertino, L. P. Carlette, D. d. O. Filho, H. A. Pereira and P. F. Ribeiro, "Comparison of Solar Panel Models for Grid Integrations Studies," INDUSCON, 2012. [13] P. Rodriguez, J. Pou, J. Bergas, J. Candela, R. Burgos and D. Boroyevich, "Decoupled Double Synchronous Reference Frame PLL for Power," IEEE Transactions on Power Electronics, vol. 22, pp. 584 592, Março 2007. [14] A. Yazdani and R. Iravani, "Voltage-Sourced Converters in Power Systems - Modeling, Control, and Applications.," Power Engineering Society General Meeting, no. IEEE, 2006. [15] P. M. d. Almeida, Modelagem e Controle de Conversores Fonte de Tensão Utilizados em Sistemas de Geração Fotovoltaicos Conectados à Rede Elétrica de Distribuição, UFJF, Ed., Juiz de Fora, 2011. (Dissertação de Mestrado). [16] E. F. Cota, A. F. Cupertino and H. A. Pereira, "Análise Da Operação de Sistemas Fotovoltaicos durante Afundamentos desequilibrados De Tensão," CBENS, 2014. [17] R. Teodorescu, M. Liserre and P. Rodríguez, Grid Converters for Photovoltaic and Wind, John Wiley and Sons, 2011. BIOGRAPHIES Edmar Ferreira Cota received the B.S. degree in electrical engineering from the Federal University of Viçosa (UFV), Viçosa, Brazil, in 2014. Currently he is Master student from Federal University of Minas Gerais (UFMG), Belo Horizonte, Brazil., where developed works about power electronics applied in renewable energy systems. Lucas Santana Xavier was born in Viçosa -MG, Brazil. He is student of Electrical Engineering at Federal University of Viçosa (UFV), Viçosa, Brazil. He is integrant of GESEP, where developed works about power electronics applied on renewable energy systems. The interests of his research include photovoltaic systems and multifunctional power inverters. Allan Fagner Cupertino received the B.S. degree in electrical engineering from the Federal University of Viçosa (UFV), Viçosa, Brazil, in 2013. He is integrant of GESEP, where developed works about power electronics applied in renewable energy systems. Currently he is Master student from Federal University of Minas Gerais (UFMG), Belo Horizonte, Brazil. He has been with the Department of Materials Engineering of CEFET-MG, Brazil. His research interests include solar photovoltaic, wind energy, control applied in power electronics. Heverton Augusto Pereira received the B.S. degree in electrical engineering from the Universidade Federal de Viçosa (UFV), Brazil, in 2007 and M.Sc. degree in electrical engineering from the Universidade Estadual de Campinas (UNICAMP), Brazil, in 2009. Currently is Ph.D. student from the Universidade Federal de Minas Gerais (UFMG), Brazil. He has been assistant professor at the Electric Engineering Department in the UFV, Brazil, since 2009.