A standard-compliant prototype for PV module curve detection V. Boscaino*, G. Cipriani*, V. Di Dio*, R. Miceli*, A. Pulizzotto*, R. Rizzo † *Department of Energy, Information engineering and Mathematical models,University of Palermo, viale delle scienze, building 9, 90128 Palermo, Italy. Contact:
[email protected] † Department of Electrical Engineering and Information Technologies DIETI, University of Naples Federico II, Naples, Italy. E-mail:
[email protected]
the radiation level should be kept constant within ±1% and the operating temperature should be kept constant within ±2°C. Detected points close to short-circuit condition can be extrapolated from directly detected points only if at least one point is directly measured within [0, 3%] range of the opencircuit voltage. Fast detection is therefore required and constraints on the minimum input equivalent resistance are dictated to comply with the standard. In this paper, the use of a buck-boost converter for I-V curve detection of a PV module is investigated. The proposed electronic load is designed to comply with actual regulations in force. Design criteria of the power conversion stage are described for further applications. Performances of the electronic load in terms of detection time, accuracy and compliance with regulations are heavily affected by dutycycle patterns. Different duty-cycle patterns are designed and their effects on system performances are deeply investigated. Simulation and preliminary experimental results are shown to test performances of the proposed solution.
Keywords: PV module, Photovoltaic generator, photovoltaic I-V characteristics, buck-boost converter, MPPT, electronic load.
Abstract In this paper, a prototype for PV module curve detection which features low cost and size and ensures high accuracy is proposed. The proposed architecture is based on a buck-boost converter in open-loop configuration. The proposed architecture is fully compliant with actual standard IEC60904. The duty-cycle control algorithm and its own effects on system performances are accurately investigated. The comparison of different duty-cycle patterns is carried out and performance evaluation is performed. Simulation results are shown to perform a valuable comparison of control algorithms. Preliminary experimental results on a laboratory prototype are presented.
1 Introduction 2 Power architecture design
Today, research is focused on renewable energy sources. Even if fuel cells are still investigated for stationary applications [1-2], solar, wind and hydro sources are still mature technologies for installations. In the next future, the installed power capacity of solar plants is expected to still experience an increase due to several economic incentives and technology advances. Main concern relies in performance degradations due to changes in operating conditions, in solar radiation or even due to mismatch conditions and partial shading. Power derating is the main issue to deal with. Detection of the effective I-V curve of PV modules to identify abnormal operating conditions is usually required. In the literature, several architectures for PV module curve detection, often based on the use of DC-DC converters, are proposed [3-13]. Power architectures differ for components count, cost, size, efficiency and input current ripple performances. The aim is at controlling the DC-DC converter to emulate a time-variant non-linear load for the PV module, thus enabling fast, accurate and automated detection of the PV module I-V curve. In literature, control algorithms are seldom investigated. Furthermore, the detection of the PV module characteristic curves should comply with IEC 60904 standard, ruling the I − V characteristics curve detection both under natural and artificial light. A detailed description of constraints on apparatus and measurements is given [14]. According to standard specifications, during measurements
In previous works, authors have presented a system set-up for I-V curve detection of PV modules, fully-compliant with actual standards [15-19]. In [16], a new prototype based on the use of a DC-DC buck-boost converter has been proposed. In this paper, the investigation of duty cycle control laws for accurate and fast detection is proposed. The detection system is tested by simulation results on a 250 W commercial module Conergy E215P. By manufacturers’ data, a five-lumped parameter model of the commercial module is obtained to compare simulation results of the designed prototype [20-24]. Parameters of the five lumped parameters model are: diode quality factor ܣൌ ͳǡ͵; series resistance ܴ௦ ൌ Ͳǡ͵͵ͷπ;shunt resistance ܴ௦ ൌ ͳͲͶͺǡπ; saturation current ܫ ൌ ͳǡͲͺͲ ȉ ͳͲି ;ܣphotocurrent ܫ ൌ ͺǡʹͳ͵Ͳܣ. The simulation setup is shown in Fig.1. The buck-boost converter in open loop configuration, including input capacitive filter is connected in parallel to the model of the commercial Conergy E215P PV module. The duty-cycle pattern in a step-by-step pattern is implemented by a piecewise linear voltage source and a sample and hold block. The sampling frequency of the sample and hold is properly designed to achieve steady-state working conditions, as it will be further discussed. The duty-cycle signal is generated from the PWM comparator, which is fed by the duty cycle pattern and a 125 kHz saw-tooth signal. Signals of interest are monitored by means of voltage and current probes.
1
The aim is a fast detection of thhe I-V characteristics of the PV module under test. Furtherm more, constraints of actual regulations should be met. Finnally, accuracy is a key target especially close to the maxim mum power point. The PV module characteristic detection should not be affected by the transient response of the electronic e load. Steady-state constraints and fast detection are usually conflicting requirements. In order to solvee the trade-off, the duty-cycle pattern should be properly desiggned. The duty cycle pattern directlly affects the detection time, corresponding to the pattern perriod. The narrower the pattern period is, the faster detection is achieved. The risk of fast detection usually relies in suddden duty cycle changes and therefore transients of the DC-DC D converter cannot be reasonably neglected thus leadding to poor accuracy in true detection of PV module curves. The accuracy is here evaluatted by comparing simulation results. A commercial module is modelled by a five-lumped parameters model. Simulation results of the proposed electronic load performed wiith the proposed duty-cycle patterns are compared to evvaluate performances of the designed system. Fig.2 shows the first analyzed duty-cycle pattern and the corresponding profile of the equivalent load resistance on the PV module seection. The pattern provides a saw-tooth step-by-step duty-cycle waveform, values ranging from 0.1 to 0.9. The slope of the saw-tooth waveform is t trade-off between steadyaccurately designed to solve the state condition and fast detectioon. A 1.2 s detection period is achieved. As shown by simulaation results, the entire set of load lines is almost completed after a 0.5 s.
Fig.1: PSIM model of the detection system.. The DC-DC converter is forced in a Conttinuous Conduction Mode (CCM) of operation to reduce conducction losses. The converter is properly controlled to emuulate a time-variant load line for the PV module curve detecttion. The load line correlation with the duty-cycle value is given by the relationship between input and output ressistance of a buckboost converter in CCM operation: ܴௗ ሺͳ െ ܦሻଶ (1) ܦଶ The duty-cycle control law directly affectss the sequence and time disposition of PV module load liness. Furthermore, the transient response of the buck-boost connverter should not affect performed measurements. Thereforee, by changing the duty-cycle pattern, sampled points crowding on the detected curve and the entire detection time consequuently change. The lowest current coordinate is fixed by the lowest l value of Rin and the maximum voltage coordinate is fixed by the maximum value of Rin. The duty-cycle range is selected within 0.1 0 and 0.9 values. The Rload valued is designed to achieeve minimum and maximum values of Rin to comply with stanndards. The buck-boost converter is designed to ennsure a Continuous Conduction Mode of operation over the whole duty-cycle working range. Converter parameters are listed in Table 1. ܴ ൌ
ܮൌ ͷͲߤ ܪ- ܴ ൌ ͳ͵݉π ܥ ൌ ͳͷߤ ܨ- ܴ ൌ ͳ͵݉π ܥ௨௧ ൌ ͳͳߤ ܨ- ܴ௨௧ ൌ ͺǡ͵݉ ݉π ܴௗ ൌ ͵ǡͻπ ܨ௦௪ ൌ ͳʹͷ݇ݖܪ ݎௗ௦ǡሺ௦௧ሻ ൌ ͳͳ݉π Table 1: Converter parameters.
Fig.2: Linear duty-cycle patternn. In Fig.3 settling time as a functtion of the duty-cycle value is plotted. As shown by simulaation results, the worst-case settling time is less than 3 ms, corresponding c to the maximum delay time in each point detectioon. The same duty-cycle value should be kept at least for 3 ms to ensure high detection accuracy. In the proposed setup, each duty-cycle value is kept constant for 5 ms.
3 Duty cycle patterns and effeects on system performances Performances of the designed electronicc load are heavily affected by the implemented duty cycle pattern. Effects of different patterns on system performancess are evaluated by simulation results.
2
Fig.3: Settling time versus duty-cycle values during the linear duty-cycle pattern. In Fig.4, simulation results under STC conditions are plotted. Simulated curve by the proposed electronic load in red and simulated curve by five-lumped parameters model in blue are plotted. The proposed electronic load features quite high accuracy. Fig. 5 shows the comparison of simulation results under ܩൌ ͷͲͲ ܹ Τ݉ଶ and ܶ ൌ ʹͷι ܥenvironmental conditions. In red the sampled I-V curve, in blue the fivelumped parameters simulated curve are plotted. As shown by simulation results, even if environmental conditions are quite far from STC conditions, high accuracy is achieved. The second duty-cycle pattern provides a smoother and lower profile of the input resistance. This pattern is obtained by combing two linear duty-cycle patterns of different slope, as shown in Fig.6. Fig.5: I-V and P-V curves ൌ ͷͲͲ Τଶ - ୡ ൌ ʹͷι with the linear duty-cycle pattern. 1 0.9
Duty cycle pattern
0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0
0
0.2
0.4
0.6 Tim e [s]
0.8
1
1.2
Fig.6: Variable slope linear duty-cycle pattern leading to a smoother Rin profile. Settling time curve as a function of the duty cycle value is plotted in Fig. 7. Settling time curve, smoother duty pattern
-3
5
x 10
4.5 4 Settling time [s]
3.5 3 2.5 2 1.5 1 0.5 0 0.1
Fig.4: I-V and P-V curves at ൌ ͳͲͲͲ Τଶ - ୡ ൌ ʹͷι with the linear duty-cycle pattern.
0.2
0.3
0.4
0.5 Duty cycle
0.6
0.7
0.8
0.9
Fig.7: Settling time versus duty-cycle values during the linear duty-cycle pattern.
3
Simulation results with the second duty cycle pattern are shown in Fig.8 As shown by simulation results, samples density lowers close to the maximum power point of the P-V curve.
1 0.9 0.8
Duty cycle
0.7
Sampled I-V curve, smoother duty pattern
0.6 0.5 0.4
9 0.3
8 0.2
7
0.1
Current [A]
6
0
0
0.2
0.4
0.6
1
1.2
1.4
Fig. 9: Non linear duty cycle pattern to achieve a linear input resistance profile.
4 3
Settling time curve, non linear duty cycle pattern
-3
2
5
1
4.5
x 10
4
0
5
10
15
20 Voltage [V]
25
30
35
40
3.5 Settling time [s]
0
0.8 Tim e [s]
5
Sampled P-V curv e, smoother duty pattern 250
3 2.5 2
200
1.5
150
0.5
Power [W]
1
0 0.1
0.2
0.3
0.4
100
0
0.6
0.7
0.8
0.9
Fig.10: Settling time versus duty-cycle values during the linear duty-cycle pattern.
50
0
0.5 Duty cycle
5
10
15
20 Voltage [V]
25
30
35
Sam pled I-V curve, non linear duty pattern
40
9
Fig.8: I-V and P-V curves at ൌ ͳͲͲͲ Τଶ - ୡ ൌ ʹͷι with the smoother duty-cycle pattern.
8 7
Current [A]
6
As shown by simulation results, the variable slope duty cycle pattern leads to a smoother input resistance profile thus achieving better transient performances. The density of sampled points around the maximum power point is yet heavily reduced by the selected duty cycle profile. After all, the linear duty cycle pattern is preferred for practical implementations. As shown by simulation results, the variable slope duty cycle pattern leads to a smoother input resistance profile thus achieving better transient performances. The density of sampled points around the maximum power point is yet heavily reduced by the selected duty cycle profile. After all, the linear duty cycle pattern is preferred for practical implementations. The third duty cycle pattern provides a linear profile of the input equivalent resistance at the PV module section, as shown in Fig.9. The corresponding pattern is achieved by manipulating the relationship of input equivalent resistance as a function of the duty cycle and load resistance. Settling time as a function of the duty cycle value is shown in Fig.10. The corresponding sampled curve is shown in Fig.11.
5 4 3 2 1 0
0
5
10
15
20 Voltage[V]
25
30
35
40
Sam pled P-V curv e, non linear duty pattern 250
Power [W ]
200
150
100
50
0
0
5
10
15
20 Voltage [V]
25
30
35
40
ଶ
Fig.11: I-V and P-V curves at ൌ ͳͲͲͲ Τ - ୡ ൌ ʹͷι with the non linear duty-cycle pattern. As shown by simulation results, the non linear duty cycle pattern leads to higher crowding of detected points over the whole curve detection. Yet, after the maximum power point density of sampled points is heavily lowered. The linear duty cycle pattern still ensures better performances.
4
5 Experimental results A laboratory prototype has been realized at the University of Palermo to test the efficiency of the proposed solution. Preliminary experimental results are here presented to test the proper working of the prototype under each operating conditions and the correctness of system design. The prototype is tested by changing the input voltage value and the duty cycle value according to simulated PV module characteristics. Preliminary experimental results show a proper working of the prototype under each effective condition. Fig.12 shows experimental waveforms under 30% duty-cycle at 35.4 V input voltage. The inductor current (Ch1, 5 A/div), PWM comparator output (Ch2, 5 V/div), output voltage (Ch3, 10 V/div) and the MOSFET gate drive signal (Ch4, 20 V/div) are shown. Time base is set at 5 μs/div. Under the specified operating conditions, experimental results perfectly match simulation results: the average inductor current is equal to 5 A and the average output voltage is equal to 15 V. Experimental results match simulation results under the same operating conditions.
Fig.13: Experimental results at Vin= 3.7 V and D = 80%.
Fig.14: Experimental results at Vin= 2 V and D = 90%. As shown by preliminary experimental results under each effective operating condition, a proper working of the laboratory prototype is ensured.
Fig.12: Experimental results at Vin= 35.4 V and D = 30%. Fig.13 shows experimental waveforms under 80% duty-cycle at 3.7 V input voltage. The inductor current (Ch1, 10 A/div), PWM comparator output (Ch2, 5 V/div), output voltage (Ch3, 10 V/div) and the MOSFET gate drive signal (Ch4, 20 V/div) are shown. Time base is set at 5 μs/div. Under the specified operating conditions, experimental results perfectly match simulation results: the average inductor current is equal to 7A and the average output voltage is equal to 6 V. A proper working of the laboratory prototype is ensured even if the input voltage is set to 3.7V. Prototype efficiency for high duty-cycle values heavily affects matching from simulation and experimental results. Fig.14 shows experimental waveforms under 90% duty-cycle at 2 V input voltage. The inductor current (Ch1, 10 A/div), PWM comparator output (Ch2, 5 V/div), output voltage (Ch3, 5 V/div) and the MOSFET gate drive signal (Ch4, 10 V/div) are shown. Time base is set at 5 μs/div. Under the specified operating conditions, the average inductor current is equal to 6 A and the average output voltage is equal to 2.5 V. A proper working of the laboratory prototype is ensured even if the input voltage is set to 2V. The discrepancy between simulation and experimental results increases as far as the duty cycle value approaches a unit value due to the decreasing efficiency with duty-cycle.
4 Conclusions In this paper, a low cost, low components count electronic load for fast and accurate PV module curve detection has been proposed. A buck-boost converter has been designed in open loop configuration. Effects of three different duty-cycle patterns on performances in terms of detection time, accuracy and compliance with actual regulations have been accurately investigated. A comparison of simulation results and performance evaluation has been proposed. As shown by simulation results, the best suited pattern among the other is the saw-tooth one. For future developments, an adaptive pattern is actually investigated.
Acknowledgements This publication was partially supported by the PON PON04a2_H "i-NEXT" Italian research program. This work was realized with the contribution of SDES (Sustainable Development and Energy Savings) Laboratory UNINETLAB - University of Palermo.
5
controlled by a virtual instrument,” International Journal of Photoenergy, vol. 2012, 2012. [12] Seapan, Manit; Limsakul, Chamnan; Chayavanich, Tasanee; Kirtikara, Krissanapong; Chayavanich, Nattavut; Chenvidhya, Dhirayut, "Effects of dynamic parameters on measurements of IV curve,"Photovoltaic Specialists Conference, 2008. PVSC '08. 33rd IEEE , vol., no., pp.1,3, 11-16 May 2008 [13] Haverkamp, E. J.; Drozdowicz, Z.; Smith, A.; Mulder, P.; Bauhuis, G. J.; Schermer, J. J.; Bissels, G. M M W; Smeenk, N. J.; Vlieg, E., "Steps to minimize the characterization errors in steady state IV curve measurements taken with C.O.T.S. equipment," Photovoltaic Specialists Conference (PVSC), 2011 37th IEEE , vol., no., pp.002262,002267, 19-24 June 2011 [14] IEC 60904 1: Photovoltaic devices - Part 1: Measurement of photovoltaic current-voltage characteristics. 2006. [15] G. Cipriani, G. Ciulla, V. Di Dio, D. La Cascia, and R. Miceli, “A device for PV modules I-V characteristic detection,” in Clean Electrical Power, 2013. ICCEP ’13. International Conference on, 2013. [16] Boscaino V., Cipriani G., Di Dio V., Miceli R., Prestigiacomo G., Pulizzotto A., “A DC-DC power converter for PV module characterization”, International Symposium on Power Electronics, Electrical Drives, Automation and Motion, 2014, pp.:1-6, 18-20 June 2014, Italy. [17] V. Di Dio, G. Cipriani, D. La Cascia, and R. Miceli, “Design, Sizing and Set Up of a Specific Low Cost Electronic Load for PV Modules Characterization,” in Ecological Vehicles and Renewable Energies (EVER), International Conference & Exhibition on, 2013. [18] T. U. Townsend, “A method for estimating the long-term performance of direct coupled photovoltaic systems,” University of Wisconsis, Madison, 1989. [19] 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, pp. 1198–1208, 2009. [20] Conergy E215P PV Module Datasheet, Http://www.conergy.de. 2011. [22] V. Boscaino, G. Capponi, G. Cipriani, V. Di Dio, and R. Miceli, “A Simple and Accurate Model of Photovoltaic Modules for Power System Design,” in 2014 Ninth International Conference on Ecological Vehicles and Renewable Energies (EVER), 2014. [23] G. Cipriani, V. Di Dio, D. La Cascia, R. Miceli, and R. Rizzo, “A novel approach for parameters determination in four lumped PV parametric model with operative range evaluations,” International Review of Electrical Engineering, vol. 8, pp. 1008–1017, 2013. [24] G. Ciulla, V. L. Brano, V. Di Dio, and G. Cipriani, “A comparison of different one-diode models for the representation of I–V characteristic of a PV cell ,” Renewable and Sustainable Energy Reviews , vol. 32, pp. 684–696, 2014.
References [1] V. Boscaino, R. Miceli, and G. Capponi, “A semiempirical multipurpose steady-state model of a fuel cell for household appliances,” in 4th International Conference on Clean Electrical Power: Renewable Energy Resources Impact, ICCEP 2013, 2013, pp. 490– 495. [2] V. Boscaino, R. Miceli, G. Capponi, and G. Ricco Galluzzo, “A review of fuel cell based hybrid power supply architectures and algorithms for household appliances,” International Journal of Hydrogen Energy, vol. 39, pp. 1195–1209, 2014. [3] E. D. Aranda, J. A. G. Galan, M. S. de Cardona, and J. M. A. Marquez, “Measuring the I-V curve of PV generators,” Industrial Electronics Magazine, IEEE, vol. 3, pp. 4–14, 2009. [4] E. Duran, M. B. Ferrera, J. M. Andu´jar, and M. S. Mesa, “I-V and P-V curves measuring system for PV modules based on DC-DC converters and portable graphical environment,” in Industrial Electronics (ISIE), 2010 IEEE International Symposium on, 2010, pp. 3323–3328. [5] L. Cristaldi, M. Faifer, M. Rossi, and F. Ponci, “A Simple Photovoltaic Panel Model: Characterization Procedure and Evaluation of the Role of Environmental Measurements,” Instrumentation and Measurement, IEEE Transactions on, vol. 61, pp. 2632–2641, Oct. 2012. [6] S. Upadhyay, S. Mishra, and A. Joshi, “A Wide Bandwidth Electronic Load,” Industrial Electronics, IEEE Transactions on, vol. 59, pp. 733–739, Feb. 2012. [7] E. Duran, J. Galan, M. Sidrach-de-Cardona, and J. M. Andujar, “A New Application of the Buck-BoostDerived Converters to Obtain the I-V Curve of Photovoltaic Modules,” in Power Electronics Specialists Conference, 2007. PESC 2007. IEEE, 2007, pp. 413– 417. [8] E. Duran, M. Piliougine, M. Sidrach-de-Cardona, J. Gala´n, and J. M. Andujar, “Different methods to obtain the I-V curve of PV modules: A review,” in Photovoltaic Specialists Conference, 2008. PVSC ’08. 33rd IEEE, 2008, pp. 1–6. [9] E. Duran, J. Gala´n, M. Sidrach-de-Cardona, M. B. Ferrera, and J. M. Andujar, “A new application of Duty Cycle Sweep based on microcontroller to obtain the I-V characteristic curve of photovoltaic modules,” in Industrial Technology, 2008. ICIT 2008. IEEE International Conference on, 2008, pp. 1–6. [10] E. Duran, M. Sidrach-de-Cardona, J. Gala´n, and J. M. Andujar, “Comparative analysis of buck-boost converters used to obtain I-V characteristic curves of photovoltaic modules,” in Power Electronics Specialists Conference, 2008. PESC 2008. IEEE, 2008, pp. 2036– 2042. [11] E. Duran, J. M. Andújar, J. M. Enrique, and J. M. PérezOria, “Determination of PV generator I-V/P-V characteristic curves using a DC-DC converter
6
