TESTING OF SMART PV MODULES

Transcription

1 TESTING OF SMART PV Daniel Gfeller, Urs Muntwyler, Christian Renken, Luciano Borgna Berne University of Applied Sciences (BFH), Engineering and Information Technology Photovoltaic Laboratory (PV-Lab), Jlcoweg 1, CH-3400 Burgdorf, Switzerland Phone +41 (0) , Fax +41 (0) , ABSTRACT: There are more and more PV modules with integrated electronic devices on the market. These smart PV modules have their electronic devices either directly inside the junction box or in a box in close proximity to the module. Common applications for smart modules are MPP tracking on module level (with microinverters or power optimizers), monitoring of the module's electrical data or safety features such as disconnecting the module in case of fire. Standard tests for PV modules are not intended for modules containing electronics other than bypass diodes in their junction box. If there are for example large capacitors in parallel to the cells, it is no longer possible to measure the module's I/V curve with a flashed solar simulator. In case of microinverters or power optimizers in the junction box, the module's DC terminals might not even be accessible, especially when the junction box is potted. This paper proposes methods for classifying and testing of different types of smart PV modules. 1 INTRODUCTION Most PV modules simply have bypass diodes in their junction box. However, there are also modules with more sophisticated electronics on the market. These smart PV modules can have all kinds of functions (e.g. MPP tracking, monitoring or safety). Accordingly, there are many possible topologies and configurations of these electronics. Proper testing of these modules is sometimes difficult, because standard test procedures for PV modules do not foresee complex electronics in the module. In this paper, some typical problems at testing of smart PV modules are being discussed. In addition, a classification scheme for smart PV modules is presented and some possible testing methods are proposed. The focus lies on the characterization of the electrical data of the module (especially the MPP power) when the latter is considered to be a black-box. Mechanical stress tests are not considered. Moreover, even though the authors are somewhat skeptical concerning complex electronics in a rough environment such as the back of a PV module, possible problems with these devices in the field (e.g. endurance or reliability) are not discussed in this paper. An exception of this are the lightning current tests under section 8. 2 COMMON APPLICATIONS FOR SMART PV Microinverters Microinverters are one of the most popular reasons for electronics in a PV module. If each PV module has its own inverter, mismatch losses of the PV plant can be minimized. Also, the design of the plant becomes very flexible, because the modules have no mutual dependencies. Power optimizers Power optimizers basically aim for the same goal as microinverters. That is the MPP tracking on module level. Unlike microinverters, power optimizers just perform a dc-dc-conversion. Thus, there is still an inverter needed to feed-in the power into the mains. Monitoring There are some products on the market which allow monitoring of the module's electrical data. With this, defective modules and other flaws in a PV array can easily be detected. The transmission of the monitored data is usually performed by powerline or radio communication. Safety circuit breakers To place safety circuit breakers directly into the PV modules allows a complete shutdown of the PV array, e.g. in case of fire. By this, dangerous voltages, which might put the firefighters at risk, can be prevented. It is also possible that circuit breakers can autonomously detect faults, such as high temperatures (fire) or electrical arcs. Active bypass diodes Most PV modules use schottky diodes as bypass diodes. However, the power dissipation of these diodes is critical if the module is in full bypass operation (i.e. when all diodes carry the full string current). An alternative are active bypass diodes that emulate the function of normal diodes but with much better characteristics (very low forward voltage and negligible leakage current). Combined functions The applications listed so far can also be combined. Monitoring functions for example are also included in most microinverters and power optimizers. 3 POSSIBLE PROBLEMS AT TESTING OF SMART PV Power consumption The electronics in smart modules need a power supply. This leads to the question of how the MPP power shall be measured and specified with or without the losses caused by the module's own consumption. The fairest assessment would be to include the own consumption into the specifications because that is the real life situation. However, there is the problem that this power consumption can be time-variant and therefore in a black-box test not deterministic. Unknown components in parallel to the solar cells Often, smart PV modules contain electronic components in parallel to the solar cells. Of course this

2 has an impact on the electrical characteristics of the module, especially but not exclusively in transient operation. One example are capacitors in parallel to the solar cells who slow down the transient response of the module. This problem is explicitly discussed under section 5. However, in black-box tests these electronic circuits are unknown. Any attempt of measuring the electrical characteristics at the connections of the junction box might lead to undefined test results. Missing accessibility of the DC terminals Especially at microinverters and power optimizers, the DC terminals of the actual module (without electronics) might not even be accessible. If the electronics cannot be detached, it is virtually impossible to measure the I/V characteristics and with that the MPP power of the module. In some cases it is possible to open the housing of the electronics, disconnect them from the cells and make a direct connection (bye-bye black-box testing). However, if the housing is filled with potting mass, this becomes very difficult. 5 CURRENT MEASUREMENT ERROR BECAUSE OF CAPACITORS IN PARALLEL TO THE SOLAR CELLS Usually, PV modules are being measured with flashed solar simulators (so called flashers). With this kind of measurement, the solar cells are only shortly exposed to the simulated sunlight. During this time, there is a voltage sweep between zero and the module's open circuit voltage at the module's DC terminals, allowing a measurement device to scan the I/V curve. However, if there are capacitors in parallel to the solar cells, they are being charged or discharged during the sweep (open circuit to short circuit or vice versa). The total charge of the capacitor is the product of the capacitance (C) and the module's open circuit voltage (V OC ). The charging process leads to an error current (I ERROR ) that compromises the measurement. I ERROR is reciprocally proportional to the sweep time (t SWEEP ). It is calculated by formula I. 4 CLASSIFICATION SCHEME FOR SMART PV Figure 1 shows a simple classification scheme for smart PV modules with respect to black-box testing. This means that on the module only manipulations are needed that are intended to be done by a normal user as well. For white-box testing the scheme also applies, as long as the module under test can be modified to the testers needs (destructive or nondestructive). Formula I: Calculation of the error current Please note that formula I is only valid, if the voltage sweep is linear (constant dv/dt). If the sweep is nonlinear, I ERROR calculated by formula I is the mean value of the actual, non-constant error current. Assuming a 72-cell module with V OC = 45 V and a sweep time of 10 ms (linear), the current measurement error will be 4.5 ma per microfarad capacitor size. Assuming an MPP power of 300W, the measurement error because of the capacitor will be about 0.54 per microfarad. As capacitors for electronic devices may have several hundred microfarads, the measurement error can easily become unacceptable. By the way: The cell itself has a parasitic parallel capacitance. The sweep time for the measurement should therefore not be too low. With some cell types (mostly thin film) this can already be a problem with a sweep time of 10 ms. If the sweep is linear, the expected error current is not too high (e.g. <5% of I MPP ) and the size of the capacitor is known, the error caused by the capacitor can be corrected mathematically. 6 CHARACTERIZATION OF SMART PV USING A CONTINUOUS SOLAR SIMULATOR Figure 1: Classification scheme for smart PV modules This classification scheme differentiates between four types of smart PV modules (A to D). Which tests are suitable for which type of module will be discussed under section 9. In most cases, the common flasher-tests are not performable with smart PV modules. As an alternative, the smart PV module can be characterized as a complete system by its energy output over a certain amount of time (see section 6 and 7). However, this kind of test is time consuming and therefore only suitable for sample testing. With a continuous solar simulator it is possible to test most types of smart PV modules. Due to the constant irradiation of the solar cells, the potentially time-variant behavior of the electronics attached to the module is less troubling. For instance, the problem with capacitors in parallel to the solar cells discussed under section 5 can be avoided with a continuous solar simulator, because the voltage sweep for the I/V curve measurement can be very slow (I ERROR then becomes negligible). Even if the electronic device attached to the module has fluctuating power consumption, a test with a continuous solar simulator is possible because if the measurement period is long enough, the fluctuations are being time-averaged. Thus, the average value of the measurement is representative for real operation. Especially for tests of smart modules with microinverters or power optimizers

3 continuous solar simulators are congenial because they allow reproducible and stable test conditions for the complete system. If the simulated irradiation of the solar simulator is sufficiently accurate and stable, the power delivered to the module can be determined by calculating the product of irradiation (G) and module area (A M ). The efficiency of the system ( ) is simply the ratio of the electric output power (P EL ) to the power delivered to the module. If P EL and G are constant (static test), the efficiency is calculated by formula II (instantaneous value). Formula II: Calculation of the static efficiency However, with a real inverter or power optimizer P EL will not be perfectly constant because of the MPP tracking. Moreover, there is also the possibility for dynamic test patterns were G is time-variant. In this general case, the efficiency is the ratio between the electrical output energy from the Module to the radiation energy to the module. The energies again are the integrals of the powers. The static values G and P EL are being replaced by their instantaneous values g(t) and p EL (t). This leads to formula III. Formula III: Calculation of the dynamic efficiency The test profiles could be defined in analogy to EN 50530, the European standard for testing of the overall efficiency of PV inverters [1]. This standard comprises both, static and dynamic test patterns. For the static measurements, the static efficiency should therefore be measured at an irradiance G of 50, 100, 200, 300, 500, 750 and 1000 W/m 2. This would allow calculating an average efficiency analog to the European efficiency in EN This would give information about the overall efficiency of the system under realistic conditions. For the dynamic measurements, the test profiles of EN 50530, where the irradiation is following linear ramps between 100 and 500 W/m 2 as well as between 300 and 1000 W/m 2 can also be adapted. However, the temperature of the solar cells during these long tests is virtually uncontrollable (that's one of the nice features of flash-tests, because there the heating of the solar cells during the short measurement period is negligible). To ensure that every tested module is being treated equally, the physical test setup has to be defined explicitly (e.g. ambient temperature of 25 C, vertical mounting of the module and an air gap of 10 cm on both sides of the module). Moreover, the temperature of at least one cell should be measured and recorded as well during the tests. 7 OUTDOOR TESTS FOR SMART PV A good characterization of smart modules under real life conditions can be achieved with outdoor tests. However, in an outdoor test stand, precise and reproducible measurements similar to laboratory measurements are not possible, because on the outside there are uncontrollable interferences like weather, dirt on the modules or even partial shading (although a good test array shouldn't have any near shading). Therefore, it is impossible to determine an absolute and representative value for the device's performance (even though it is attempted sometimes). But with a clever test setup, it is possible to make a fair comparison of different systems. The tricky thing about this is that every module in the test setup should be treated equally. Because of the volatile environmental conditions on the outside, this is not entirely possible. However, with a sufficiently large number of modules of each type (sample size) statistical correlations can be found between module type and energy yield. But to achieve good, neutral data, the physical arrangement of the modules must be wellthought-out. Figure 1 shows an optimized layout for a test array with three different module types (numbered 1-3), each of them having a sample of three modules. In this case, the modules are being positioned in a single row. Figure 2: Optimized layout of a test array with PV modules in single row alignment Due to the alternating placement of the modules, interference factors (partial shadings, inhomogeneous thermal characteristics, etc.) are being distributed as evenly as possible between the different types of modules. However, a sample of three specimens per module type is a bit low. Of course this layout can be extended to an arbitrary number of modules, but in most cases it will be more useful to align the modules in a matrix. In this case, the alternating placement of the modules should be performed in both columns and rows. Figure 3 shows an example of such an array. Figure 3: Optimized layout of a test array with PV modules in matrix alignment The PV lab of BFH runs s smart PV module test stand for 20 PV modules in single row alignment, mounted on a bike shed. Due to the low mounting height, the modules are easily accessible for manipulations and measurements. So far, we made very good experiences with this outdoor test stand, which is shown in figure 4.

4 Figure 4: Smart PV module test stand of BFH's PV lab To compare the different types of smart modules, the Energy output of each individual module has to be monitored over a certain amount of time (ideally one year or longer). The energy has then to be normalized by dividing it by the module's nominal power. This way, the final yield Y f [4] of each module can be determined. Each of these yields represents a single, independent measurement. With this data, it is possible to make a statistical analysis for each type of smart module. This can be done by calculating the expected value, the standard deviation and the confidence interval of the energy yield for each module type. The results of the different types can then be compared. A visualization of the test results, e.g. with boxplots, is also possible. 8 LIGHTNING CURRENT TESTS FOR SMART PV Because of their large area, solar cells are quite immune against currents surges. In a standard PV module, usually the bypass diodes are the weakest parts [2, 3]. But even they are power electronic devices and as such to some degree robust. Things change when there are more sophisticated electronic devices inside the junction box, especially when there are parts based on CMOS technology. These circuits can already be damaged because of electrostatic discharges from a human body (ESD). Consequently, they can easily be destroyed by atmospheric discharges such as lightning currents. As smart PV modules may contain complex electronic circuits and are self-evidently exposed to the weather, a test of the immunity against lightning currents is reasonable. Before we are discussing a possible test setup, let's make one thing clear: An all-embracing protection against lightning currents is nearly impossible (or at least unpayable). If the lightning hits the module or the nearby DC lines directly, a damage of the module and any electronics that comes with it is almost certain. But as the average density of lightning strokes is quite low, the risk of a direct hit is small. Nearby lightning strokes however, may occasionally happen. A (smart) PV module should not be damaged by them. The test setup used by BFH's PV lab is simple: A simulated lightning current I L (10/350 µs) is generated in a pulse current generator. The module under test is being placed in a certain distance (e.g. 20 cm) to the lightning current conductor. Because of I L, an alternating magnetic field (B) develops around the conductor. This magnetic field induces a voltage (V I ) inside the module's cell loop that leads to a potentially destructive current surge (I I ). This test setup is shown in figure 5. Figure 5: Setup for lightning current tests Similar to the tests discussed under sections 6 and 7, with this setup the complete system (module and electronics) is being tested (as systems engineers we like to do that). With the pulse current generator in BFH's high voltage lab, simulated lightning currents with a peak value of up to 110 ka can be generated. The induced current surges can easily go up to a few kiloamperes. However, because of the short pulse duration (few microseconds), the induced charge is limited. Therefore, previous tests of smart PV modules (i.e. with microinverters) showed pleasantly good results, because the capacitors in parallel to the solar cells absorbed most of the current surge and by doing so protected the electronic circuits. In some cases, smart PV modules withstood higher simulated lightning currents than standard PV modules (damage of the bypass diodes). Figure 6 shows a test setup with a 60-cell PV module inside of BFH's pulse current generator. The tube on the left side is the lightning current conductor. At the bottom of the module, there is a receptacle for a microinverter, which for this test has been replaced with the measurement lines for the induced voltage and current. Figure 6: PV module inside of BFH's pulse current generator

5 9 CHOICE OF THE RIGHT TESTS FOR THE RIGHT TYPE OF SMART PV MODULE The classification of the smart PV modules (type A to D) in this section refers to the classification scheme under section 4. The lightning current tests under section 8 are suitable (and recommended by the authors) for each of the four types. Module type A If the "smart" part of the module can be detached, it is possible to simply test the bare module and the electronics separately. In most cases, this allows a good estimation of the behavior of the complete system. The bare module can be flash-tested like any normal PV module. The electronics have to be tested depending on their function (e.g. tests of EN for a microinverter). However, the whole truth (that is the behavior of the complete system under real conditions) can only be tested with an outdoor test, e.g. as proposed under section 7. Tests with a continuous solar simulator as proposed under section 6 might also prove to be useful, because they also characterize the complete system, but are much less time-consuming than an outdoor measurement. Module type B If the DC terminals are not accessible, measurement of the module's I/V characteristic is impossible in a black-box test. This is usually the case when a microinverter or power optimizer is permanently fixed to the module (if it's a power optimizer, there is a DC terminal, but it is not directly connected to the solar cells). These modules are best being characterized with a continuous solar simulator (section 6) and an outdoor performance test (section 7). Module type C This type of module is an exception, because in a black-box test it actually is impossible to tell how the (unknown) electronics will or will not compromise with the measurement. It is however possible that the manufacturer of the smart module reveals the module's inner life or specifies that this inner life does not or only negligibly influence the electrical data of the solar cells. Active bypass diodes can fall under this category. Such a module can be tested like any conventional PV module (i.e. flash testing). Module type D If the electronics do compromise with the measurement or it is unknown if they do so, measuring the I/V characteristic will be impossible in most cases. The only exception is when the behavior of the electronics is known with sufficient accuracy and their influence can be compensated. For instance, if the only influence is caused by (small) capacitors in parallel to the solar cells, the current measurement error in flash tests can be corrected mathematically (see section 5). However, in a black-box test, the required information will usually not be available or not with the accuracy required (e.g. if there is an electrolytic capacitor in parallel to the solar cells, it may have a tolerance of +/-20%). Therefore, these modules can only be characterized as complete systems under real-life conditions, namely with a continuous solar simulator (section 6) and an outdoor performance test (section 7). 10 CONCLUSIONS Smart PV modules can perform many tasks. However, in most cases, smart PV modules cannot be tested like normal modules. If and how they can be tested depends on the configuration and the function of the module. In this paper, a simple classification scheme is presented. Especially the well-known flash tests cannot be performed with most smart PV modules. As an alternative, I/V curve measurement with a continuous solar simulator is possible in some cases. Such a simulator can also be used for entire test profiles for smart PV modules with a microinverter or power optimizer. A very good characterization can be achieved with an outdoor performance test. A possible test setup is proposed in this paper. Also, the authors think that the immunity against nearby lightning currents should be tested, because the compelex electronics inside of smart modules might be sensitive to surge currents. DISCLAIMER Information contained in this paper is believed to be accurate. However, errors can never be completely excluded. Therefore, any liability for correctness and completeness of the information or from any damage that might result from its use is formally disclaimed ACKNOWLEDGEMENTS Thank goes to all the institutions that gave us financial support, particularly the Swiss Federal Office of Energy (SFOE), the Swiss Competence Centers for Energy Research (SCCER) and the Building Insurance Berne (GVB). REFERENCES [1] R. Bruendlinger, N. Henze, H. Haeberlin, B. Burger, A. Bergmann, F. Baumgartner, "pren The new European Standard for Performance Characterisation of PV Inverters", 24 th European Photovoltaic Solar Energy Conference and Exhibition, Hamburg, Germany, [2] H. Haeberlin, M. Kämpfer, "Measurements of Damages at Bypass Diodes by Induced Voltages and Currents in PV Modules Caused by Nearby Lightning Currents with Standard Waveform", 23 rd European Photovoltaic Solar Energy Conference and Exhibition, Valencia, Spain, 2008 [3] H. Haeberlin, "Damages at Bypass Diodes by Induced Voltages and Currents in PV Modules Caused by Nearby Lightning Currents", 22 nd European Photovoltaic Solar Energy Conference and Exhibition, Milano, Italy, 2007 [4] H. Haeberlin, C. Beutler,"Normalized Representation of Energy and Power for Analysis of Performance and On-line Error Detection in PV-Systems", 13 th European Conference on Photovoltaic Solar Energy Conversion, Nice, France, 1995 All papers can be downloaded under