Contents lists available at ScienceDirect Solar Energy journal homepage: www.elsevier.com/locate/solener Degradation analysis of crystalline silicon photovoltaic modules exposed over 30 years in hot-humid climate in China Huili Han a,b , Xian Dong b , Bingzhi Li b , Huan Yan a , Pierre J. Verlinden c , Jiangfeng Liu d , Jiapei Huang b , Zongcun Liang a,b, ⁎ , Hui Shen a,b,e, ⁎ a Sun Yat-Sen University, China b ShunDe SYSU Institute for Solar Energy, China c Trina Solar, State Key Laboratory of PV Science and Technology, Changzhou, China d Xin Yang Normal University, Collaborative Innovation Center of Henan Province for Energy-Saving Building Materials, China e Jiangsu Collaborative Innovation Center of Photovoltaic Science and Engineering, China A R T I C L E I N F O Keywords: PV module Visual inspection Degradation performance Degradation analysis Optical loss A B S T R A C T The main object of this paper is to present the analyze of the degradation mechanism of the electrical properties and polymeric materials for a batch of crystalline-silicon photovoltaic modules, which were installed in the hot- humid region in the south of China for 30 years. Although signi fi cant degradation of polymers (EVA and back- sheet) is observed, the average power output is only 6.53% below the name plate of the modules after 30 years in the fi eld. The analysis of electrical performance indicated that the decline of short-circuit current (I sc ) is, in this case, the main cause of power degradation. The degradation of polymeric materials, fi rst observed through visual inspection followed by several analytical methods, such as XPS, optical measurements, or measurement of the degree of crosslinking and mechanical properties, is characterized by a high degree of yellowing of the Ethylene Vinyl Acetate(EVA) and cracks of the back-sheet. In addition, some corrosion of Ag grid is also observed. The high water vapor transmission rate (WVTR) of back-sheet, and the presence of cracks, accelerated the corrosion of metal, but it does not directly result in a decrease of the power output, and no obvious degradation of the fi lling factor is observed. The increase of the Yellow Index of the EVA directly results in optical loss, which is believed to be the main cause of the decline of the short-circuit current. The loss in short-circuit current caused by EVA discoloration is 12.6% in average, which agrees well with that of electrical performance analysis. Failure cause analysis reveals that there is no direct relationship between the power degradation and the degradation behavior of packaging materials except EVA discoloration. 1. Introduction With the photovoltaic (PV) technology becoming a signi fi cant por- tion of the new electricity generation capacity, the reliability of PV modules in di ff erent climates must be extensively studied and the causes or mechanisms of failure must be analyzed in relationship with the climate where the modules have been installed. For the developers and the operators of PV systems, the long-term energy yield is de- termined by the long-term reliability of the PV modules. Reliability and lifetime of PV modules are key factors to the system performance, the revenues generated by the system, the estimation of the fi nancial return on investment and, if necessary, the possible warranty claims. To im- prove the attractiveness of PV technology and improve the return on investment, it is very important to improve the stability and reliability of the performance of PV modules, both for PV companies and users (Laronde et al., 2012; Antonella, 2003; Anand and Wang, 2003). Over many years, the PV scienti fi c community has been able to access degradation data for multitude of modules installed in temperate climates (Antonella, 2003; Vikrant and Chandel, 2013; Davide et al., 2013; Pramod et al., 2016; So fi ane et al., 2016; Husam et al., 2017; Ma et al., 2017; Resis et al., 2002; Carr and Prior, 2004; Jordan and Kurtz, 2013; Rahnuma et al., 2016; Meyer and Van Dyk, 2004; D. Jordan et al., 2017). Unfortunately, less data is available for tropical climates (Pramod et al., 2016; Jordan and Kurtz, 2013; D. Jordan et al., 2017), which is recognized by many scientists as the harshest type of climates for PV modules. As observed by D. Jordan et al. (2017), modules in hot and humid climates show considerably higher degradation modes than those in desert and moderate climates. In typical PV systems, PV modules are often considered to be the most reliable part of the whole system, with an observation of higher failure rates for inverters and https://doi.org/10.1016/j.solener.2018.05.027 Received 1 December 2017; Received in revised form 3 May 2018; Accepted 7 May 2018 ⁎ Corresponding authors at: Sun Yat-Sen University, China. E-mail address: shenhui1956@163.com (H. Shen). Solar Energy 170 (2018) 510–519 0038-092X/ © 2018 Elsevier Ltd. All rights reserved. T connectors. However, the reliability of PV modules is highly dependent on the climate, the manufacturing technology and materials, and, to a lesser degree, to the installation method. Up to now, the long-term reliability of PV modules is still plaguing manufacturers and users, and a good reliability model is far from established (Laronde et al., 2012; Antonella, 2003; Anand and Wang, 2003; Vikrant and Chandel, 2013; Kempe, 2006; William et al., 1993). It also seems to appear that PV modules manufactured after the year 2000 have a signi fi cantly better reliability and lower degradation rate than earlier modules (Jordan and Kurtz, 2013; D. Jordan et al., 2017). In general, degradation percen- tages are reported to decrease appreciably in newer installations that are deployed after the year 2000. It is widely acknowledged that the real operating performance of PV modules is substantially a ff ected by various environmental conditions such as temperature, relative humidity, radiation intensity, dust storm, spectrum distribution, degradation and maintenance (Vikrant and Chandel, 2013; Davide et al., 2013; Pramod et al., 2016; So fi ane et al., 2016; Husam et al., 2017; Ma et al., 2017; Resis et al., 2002; Carr and Prior, 2004). Most manufacturers of PV modules guarantee the minimum performance of their modules for 20 – 25 years, with an as- sumed linear degradation. A 30 – years warranty has been introduced, and 40 – year warranties will be anticipated soon (Jordan and Kurtz, 2013). In this context, the understanding of long term reliability of PV modules is the utmost importance. Quali fi cation test, as a method of assessing the long-term reliability of PV modules, is done for rapid detection of failures of PV modules in a controlled environment. The IEC (International Electrotechnical Commission) 61215 standard is the early quality assessment standard of crystalline silicon PV modules (IEC 61215, 2005). It is allows for the early detection of design fl aws of the PV modules. However, it is not a set of reliability tests for estimating long-term degradation and lifetime of the modules. A methodology for estimating the lifetime of a PV module using an accelerated degradation model was proposed, which demonstrated how a failure time dis- tribution may be obtained by means of several accelerated degradation tests in a continuous-time stochastic process, called Wiener process (Charki et al., 2013). The Wiener process in conjunction with the ac- celerated failure time model makes it possible to carry out numerous simulations and thus to determine the failure time distribution based on the pre-determined threshold value. By this means, the failure time distribution and the lifetime (mean and uncertainty) can be evaluated. The comparison of di ff erent technologies for solar PV module outdoor performance using indoor accelerated aging tests for long term relia- bility were studied, the results indicated that mono-crystalline Si has shown best performances in terms of small variation in maximum power and a-Si technology has shown poorest performances in terms of large variation in maximum power (Rahnuma et al., 2016). The quali fi cation tests, such as IEC 61215, can only evaluate the possibility of failure in the early stage of the life of the PV modules, caused by fl aws in the design of modules or materials, and are not su ffi cient to estimate the module lifetime under di ff erent geographical conditions. In order to ensure the quality and reliability of the PV modules, many manufacturing companies, customers or testing agen- cies have stricter quality testing standards. For example, higher UV irradiation is often required for packaging materials, such as back-sheet or EVA (Thomas et al., 2009). Despite the e ff orts to create a valid test sequence, so far scientists have not been able to duplicate the de- gradation of PV modules in real conditions with the degradation of PV modules under indoor accelerated testing, due to the lack of complete understanding of the degradation mechanisms and activation energy. Some environmental parameters are also di ffi cult to accelerate. For example, the degradation mechanism of polymer materials in the fi eld with composite environmental factors, such as UV temperature and humidity, cannot be repeated indoor in an accelerated test. Therefore, the PV industry and the PV scienti fi c community requires more data and analysis of the reliability and degradation of the performance of PV modules after a long-term exposure in the fi eld under di ff erent climates. In this paper, the degradation analysis of crystalline-silicon PV modules after 30 years of outdoor exposure in a hot and humid climate is reported. The analysis of electrical performance analysis indicates that the decline of the short-circuit current (I sc ) is the main cause of the Fig. 1. The installation information of Solarex modules in the fi eld: (a) The geographical location of installation in two periods. (b) The original installation site in Jianfengling, Hainan (1996 – 2008). (c) The re-installed site in Guangzhou, of Guangdong (2010 – ). H. Han et al. Solar Energy 170 (2018) 510–519 511 power degradation. The degradation mechanism of packaging materials is discussed and the loss in optical properties of the EVA, due to yel- lowing, corresponds to a 12.6% loss in transmittance. 2. Experimental platform A PV system composed of 196 modules was initially installed in 1986 in Jianfengling (Latitude 108°55 ′ E, Latitude 18°35 ′ N), in the Hainan provinceof China. Fig. 1 shows the geographical positions of installation and the photovoltaic array in di ff erent periods, The batch of modules were produced in 1982 by the company Solarex using mono- crystalline silicon solar cells, ethyl vinyl acetate (EVA) as encapsulant and a single layer of Tedlar as back-sheet. The island of Hainan presents a tropical island monsoon climate, classi fi ed Am in the Köppen climate classi fi cation scheme, characterized bya hot and dry climate with temperatures varying between 20 °C and 35 °C throughout the year, the average temperature being almost always above 25 °C. The average rainfall in Hainan about 2000 mm, with a dry season in winter and spring, and a wet season in summer and autumn. Hainan bene fi ts from about 2200 h of sunshine in average per year. The size of a solarex module is 970 ∗ 445 mm, and the size of a cell from the Solarex moules is 101 ∗ 101 mm. Each Solarex module consists of four columns of cells connected in series, and each column consists of 9 cells. Fig. 2 shows a panorama of a Solarex module. In 2008, the 196 Solarex modules were removed and 177 modules were collected by the Institute for Solar Energy system of Sun Yat-sen University (SYSU), i.e. 22 years after the original installation (other 19 modules were damaged so seriously that they were rejected). A series of testing including visual inspection, I-V characteristics and EL imaging of every Solarex module were performed during the period of 2008 – 2010, while the modules were kept indoor. The junction boxes of most modules looked badly degraded and were replaced. Finally, 144 modules were choosed and re-installed in 2010 on the roof of College of engineering of SYSU in Guangzhou, Guangdong Province, for further exposure and testing. Over the last 8 years, the performance of every module has been carefully monitored while exposed to the environment in Guangzhou. The PV array under test is shown in Fig. 1c. The re- maining modules (33 modules in total) have been analyzed to under- stand the cause of failure and degradation mechanisms. The climate in Guangzhou is classi fi ed Humid Sub-tropical, or Cfa in the Köppen climate classi fi cation scheme. With an average temperature varying between 14 °C and 29 °C, an average rainfall of 1700 mm per year, and 1773 h of sunshine, Guangzhou has a similar climate as Hainan but just not as hot and not as humid. The next section will present a comprehensive analysis of the de- gradation of Solarex modules throughout their exposure to sunshine and the environment over 30 years. The study includes visual inspec- tion, electroluminescence (EL) characterization, electrical performance degradation analysis and failure analysis of the encapsulant materials. 3. Results and discussions 3.1. Visual inspection and corresponding EL image The visual inspection was performed according to the IEC 61215 standards. Fig. 3 showed the result of visual inspection of 177 modules and the frequency of defects. The same defect occurring in multiple positions in one single module was considered to be one defect for purpose of the present study. Yellowing of EVA and glass soiling were the most common defects, which occurred in every PV module. The corrosion of the busbar was also observed in 100% of the modules. Delamination, bubble in EVA, cracks in back-sheet and cracks in cells were observed in more than 60% of the modules. To complete the visual inspection, electroluminescence (EL) ima- ging f the modules was used to identi fi ed fi nger interruptions and “ dark areas ” of cracked cells (Thomas et al., 2009). Figs. 4 – 6 showed the corresponding positions of visual defects and EL defects in solar cells. It is interesting to note from Fig. 4a that cracks in cells are observed at the same location as a crack in the back-sheet. The cracking of back- sheet signi fi cantly increases moisture ingress, and subsequently causes some swelling or deformation of the EVA encapsulant. The moisture ingress and the swelling of the EVA may induce in mechanical stresses that may lead to the crack of cells (Sander et al., 2010). This kind of deformation can be understood by the EVA creeping under the condi- tions of humidity and heat (Miller et al., 2010). Fig. 4a shows the visual image (top) of a typical crack in the back-sheet of a module and the corresponding crack in the cell at the same location. The e ff ect of the crack on the performance of the cell and the module is con fi rmed by the EL image (Fig. 4a bottom). When the cracks of back-sheet are located at the gap between cells, the deformation of EVA and the mechanical stress do not generate cracks in cells (Fig. 4b). This is con fi rmed by inspection of the EL image (Fig. 4b bottom). Acetic acid is a well-known by-product of hydrolysis within EVA that occurs in the presence of moisture, heat and UV. Acetic acid acts as a catalyst in the corrosion of the cell metallization and metallic inter- connects (Kempe et al., 2006; Gagliardi et al., 2017). Acetic acid for- mation is related to the di ff usion process of water vapor, through the back-sheet or moisture through back-sheet cracks, into the module and is accelerated by heat and UV following a typical Arrhenius relationship dependent on temperature (Kraft et al., 2015; Hülsmann et al., 2014; Haillant et al., 2011). Fig. 5 shows dark areas in solar cells caused by corrosion of fi nger electrodes and solder ribbon (Peike et al., 2013). Fig. 5a shows that the corrosion often occurs at the edge of cells or where a crack in cell is observed because of water vapor and gas dif- fusing from the edge of cells (Kempe, 2006) or through the crack. An obvious question could be: “ Why does it happen only on one edge? ” The answer to this question is not absolutely certain. It is believed that the Fig. 2. The panorama of a Solarex module. Fig. 3. Visual inspection results of 177 modules. H. Han et al. Solar Energy 170 (2018) 510–519 512 corrosion of metal by the acetic acid is accelerated by the temperature, which sometimes can be larger at one edge of a cell due to either a large leakage current created by a shunt between front junction and Back Surface Field (BSF) or by a localized reverse breakdown current in case of partial shading. In some cases, as shown in Fig. 5b, more serious corrosion is ob- served in the center of cells. It is not well understood at this time why in some cases the corrosion happens at the edge, supposedly limited by di ff usion of moisture and the generation of acetic acid, and in some other cases it happens at the center of the cell, supposedly accelerated by the non-uniform temperature distribution (higher temperature in the center of cells) or limited by the out-di ff usion (or also sometimes called “ breathing ” ) of acetic acid (Weerasinghe et al., 2015). Anyway, in both examples of corrosion by acetic acid, solder bonds and grid fi ngers can be corroded resulting in an increased series resistance (R s ), which lowers the performance of PV modules by reducing the fi ll factor. The discoloration of EVA is a common visual defect of Solarex modules after 22 years of exposure to tropical climate. It a ff ects 100% of the modules. Studies have shown that a photochemical degradation of EVA may occur under UV exposure and at elevated temperature, by the generation of chromophores. This problem was very common for modules manufactured before 1990. It was also found that the dis- coloration is almost always located at the center of cells, as shown in Fig. 1(c), due to a so-called “ photo-bleaching ” process. Photobleaching occurrs where oxygen is provided in enough quantity, di ff using through the back-sheet between cells or through cells along the cracks, to bleach Fig. 4. Cracking of back-sheet (top) and EL image (bottom). (a) The crack occurred on the back of solar cell and the defect in EL image observed in corresponding position. (b) The crack occurred at the gap between cells and there was no defect in EL image. Fig. 5. Corrosion in fi nger and interconnect ribbon (top) and EL image (bottom). (a) The corrosion occurred at the edge of a cell. (b) The corrosion occurred in the center of a cell. H. Han et al. Solar Energy 170 (2018) 510–519 513 the EVA that had been discolored by UV (Pern and Glick, 2000; D.C. Jordan et al., 2017). Fig. 6 shows another two examples where the phenomenon of photobleaching occurred along the cracks of the cells. 3.2. Electrical performance The I – V characteristics of the PV modules were measured under standard test conditions (STC, incident power density: 1000 W/m 2 , spectrum: AM1.5G, module temperature: 25 °C), using a h.a.l.m cetis PV-XF2-M solar simulator. The I-V characteristics of the Solarex mod- ules were tested in 2008 when the PV array was dismantled from Jianfengling, then afterwards in 2010 after the replacement of the de- graded junction boxes of most modules, in 2014, 2015 and 2016. Because we were not able to obtain data regarding the initial power output, the power output on the nameplate was used as reference. The minimum, maximum, average, median and standard deviation of de- gradation of I-V parameters of Solarex modules ware shown in Table 1. After operating more than 30 years outdoor, the average degradation of the maximum power (P m ), open-circuit voltage (V oc ) and the short- circuit current (I sc ) were 6.53%, 1.54% and 14.3%, respectively. In- terestingly, the degradation of I sc and I m continued after the installation in Guangzhou with an almost linear degradation of about 0.5% per year. The average annual degradation rates of power was 0.18% in Jianfengling (1986 – 2008) and 0.43% in Sun Yet San University, Guangzhou (2010 – 2016). These modules had a greater degradation rate in Guangzhou where less harsher climatic conditions were im- pacting the modules than in Jianfengling. This result may be due to the cumulative degradation caused by the total thermal cycling and damp- heat, just as explained by the so-called reliability “ bathtub ” curve (Meyer and Van Dyk, 2004). Interestingly, both the power degradation after the two periods of sunshine exposurewere still within the normal warranty of the manufacturer. In comparison with the values of 2008, after changed the junction boxes of the modules, all the maximum power, open-circuit voltage and fi ll factor (FF) were increased. That indicates the failure of the junction boxes can a ff ect the electrical performance of the modules by increasing the series resistance. The original actual power output of the modules is assumed to be ± 10% of the nameplate value as it was typical at the time of manufacturing of these modules. This situation is possibly the reason why we observed a higher fi ll factor (FF) measured after 2008 com- pared to the nameplate value of FF. Without any more reliable data about the initial power output, we accept the uncertainty on the power loss that is calculated as the di ff erence between the measured output in our laboratory and the nameplate value. Unfortunately, we cannot have a more accurate measurement of the degradation. From a scienti fi c point of view, it is not very satisfying, but the di ff erence with respect to the nameplate is the more important information for the customer. In 2016, after 26 years on sun, the maximum power point voltage V m seems to have increased by 9.6%, while the open circuit voltage V oc had no signi fi cant change. The decrease of I sc seems to have resulted from the degradation of the optical properties of the encapsulant. With the uncertainty related to the actual original value of FF, it seems that the series resistance has not increased in the degradation process. It is also very well understood that the non-uniformity of the degradation of the short-circuit current I sc , each cell providing a quite di ff erent I sc than the other cells in the same module and creating an important current mismatch within each module, is also a very possible explanation of the increase in FF and in V m . This was already apparent in 2008 after the fi rst EVA degradation while the modules were in Jianfengling, and continued over the years, but to a lesser degree, when the modules were on sun in Guangzhou. The average IV parameters of the 144 modules normalized by the nameplate rating (Fig. 7) indicate that there is a similar degradation trend between P m and I sc . We concluded that the degradation of I sc , resulting from the degradation of the optical properties of EVA, is the main cause of power loss. This is analyzed in more details in the fol- lowing section. 3.3. Degradation analysis of polymer materials The frequency chart of visual defects presented in Fig. 3 clearly shows that yellowing or browning of the EVA is the most important cause of current and power loss. It is a well-known problem for PV modules manufactured before the 1990s, caused by a photochemical degradation of EVA that occurs under UV exposure and at elevated temperature, by the generation of chromophores. The EVA can be subsequently bleached by oxygen di ff using through the back-sheet, Fig. 6. Delamination of EVA and cracks in solar cells (top) and EL image (bottom). (a) Delamination occurring between glass and EVA, and dendritic-type cracks in solar cell. (b) Single and cross cracks occurring in solar cells. H. Han et al. Solar Energy 170 (2018) 510–519 514 between cells and through cell cracks. Cracks in back-sheet and cracks in cells can accelerate the bleaching process (Fig. 5). We discuss in the following sections the physical analysis of the polymer materials (EVA and back-sheet) used in the Solarex modules. 3.3.1. The degradation behavior analysis of back-sheet The back-sheet of the Solarex module is a polyvinyl fl uoride (PVF) fi lm, also called Tedlar, that was mixed with titanium dioxide TiO 2 and some other additives to improve its resistance to ultraviolet light (UV). It was extensively used as back-sheet for PV modules until about 2010 when cheaper multi-layer fi lms having similar properties and resistance to weather were developed. Table 2 presents the physical parameter values of back-sheet from Solarex modules, that were measured according to IEC62788. The water vapor transmittance rate (WVTR) was measured and a value of 8.21 g/(m 2 day) was obtained, which was higher than that of new Te- dlar PV2001 with no aging (2.4 g/(m 2 day)). We also measured the mechanical strength of the back-sheet, in both directions (MD for ma- chine direction and TD for transverse direction) and we observed that the tensile strength seriously degraded from the typical original values. Cracking of the back-sheet, as a result of the decrease in tensile strength, combined with the di ff usion of water vapor, the swelling of EVA and repeated numerous thermal cycling in tropical regions, is the most catastrophic failure mode observed in the Solarex modules, as well as among reported failure modes of back-sheets. The statistic result of defects (Fig. 3) shows that cracking in back-sheet occurred in nearly 80% of the Solarex modules after 22 years on sun. To clearly observe the degradation of back-sheet of Solarex mod- ules, the analysis of a new Tedlar PV2001 with no aging was performed as comparison. Fig. 8 shows the scanning electron microscope (SEM) images of the back-sheet. It can be seen that the surface structure of the Tedlar back-sheet exposed in the fi eld for more than 22 years was ex- tremely uneven and “ powdery ” . The back-sheet seems to be composed of a number of small particles (Fig. 8a). Compared to a unexposed Tedlar back-sheet with a dense structure (Fig. 8b), the morphology of the back-sheet of the Solarex modules seems to be destroyed by hy- drolysis and exposure to the environment. The degradation of the mi- croscopic structure of the Tedlar fi lm is thought to be the cause of the degradation of its mechanical properties. The characteristic absorption peak of Tedlar in the infrared spec- trum can re fl ect the main functional group of the polymer compounds. Fig. 8 shows the infrared spectrogram for di ff erent positions on the back-sheet of the Solarex module (a and c) and commercial Tedlar PV2001 with no aging (b). The absorption peak at 1700 cm − 1 in curves (a) and (c) corresponds to the stretching vibration peak of C ] O, and the absorption peak at 3400 cm − 1 was the stretching vibration peak of oxhydryl ( e OH), In addition, the stretching vibration peak of con- jugated alkene emerged at 1630 cm − 1 in curve (c), which may be the functional group of chromophores, caused by the chemical degradation of the back-sheet. It can be speculated that gaseous hydro fl uoric acid Table 1 Values of the minimum, maximum, average, median and standard deviation of degradation of P m , V oc , I sc , V m , I m and FF of Solarex modules in 2008, 2010, 2014 and 2016. Original values (from nameplate) P m /W V oc /V I sc /A V m /V I m /A FF 42.60 20.8 3.04 15.10 2.82 67.34 Year Parameter Minimum degradation (%) Average degradation (%) Median degradation (%) Maximum degradation (%) Standard deviation (%) 2008 P m /W 0.25 6.71 6.54 17.05 1.49 V oc /V 1.13 2.45 2.74 10.64 1.35 I sc /A 1.15 7.89 7.92 12.65 2.10 V m /V − 10.83 − 5.17 − 5.74 9.15 2.97 I m /A 6.88 11.35 10.27 11.10 2.93 FF − 9.90 − 3.77 − 4.12 7.95 3.71 2010 P m /W 0.12 4.04 3.65 11.79 3.42 V oc /V 1.25 2.26 2.29 3.52 0.46 I sc /A 5.80 9.44 9.47 13.17 1.62 V m /V − 13.89 − 9.47 − 9.35 − 5.27 1.72 I m /A 7.07 12.3 12.36 18.11 2.59 FF − 12.56 − 9.67 − 9.51 0.38 2.59 2014 P m /W 0.24 5.35 5.12 17.40 3.73 V oc /V 0.16 1.2 1.23 2.49 0.45 I sc /A 5.93 11.18 10.63 15.99 2.42 V m /V − 15.79 − 9.87 − 9.67 − 6.17 1.86 I m /A 7.86 13.83 13.78 27.52 3.03 FF − 10.86 − 8.25 − 8.28 5.02 1.97 2016 P m /W 0.22 6.53 6.52 15.46 3.77 V oc /V 0.45 1.54 1.51 2.89 0.43 I sc /A 6.92 12.5 12.14 17.35 2.53 V m /V − 14.9 − 9.6 − 9.45 − 6.08 1.79 I m /A 8.60 14.89 14.47 21.67 2.91 FF − 11.18 − 8.24 − 8.69 − 2.63 1.66 Fig. 7. The average relative value of the IV parameters normalized by the nameplate rating. H. Han et al. Solar Energy 170 (2018) 510–519 515 was separated, showing that conjugated double bonds were formed after a long-time operating in the fi eld of the environment with hu- midity and heat. The strong absorption peak at 580 cm − 1 corresponds to the stretching vibration peak of Ti e O, which may indicates the present of titanium dioxide. In order to study the change in the binding state of each element in the back-sheet, the X-ray photoelectron spectroscopy (XPS) method was used to analyse the back-sheet of the Solarex modules after exposure (a) and compared to a new Tedlar PV2001 fi lm (b). The C 1 s, F 1 s, O 1 s and Ti 2p features obtained from the two back-sheets are shown in Fig. 10. In the XPS spectra, one can study the characteristic electronic binding energy for each element. Depending on the chemical environ- ment of the atom, the electron binding energy will shift. In the Fig. 9, the C 1 s feature shown in Fig. 10(1)b for a new PVF fi lm consisted of two narrow peaks with binding energy of 284.5 eV and 287.2 eV cor- responding to C e H, C e C, while the spectrum in Fig. 10(1)a for the degraded PVF fi lm shows a new peak with a binding energy of 293.01 eV for F e C e F, which may be the by-product (C 2 H 4 F 2 ) of poly (vinyl fl uoride). The F 1 s feature is represented by a single peak with the binding energy of 686.8 eV, which is present and unchanged for both samples. The O1s feature in Fig. 10(3)b for a new Tedlar fi lm is also a single peak with a binding energy of 532.5 eV, while for the Solarex backsheet. There is another peak in Fig. 10(3)a with a binding energy of 529.5 eV, which belongs to hydroxyl oxygen and crystal lat- tice of O e Ti (Wang et al., 2010). The Ti2p feature obtained from the back-sheet of Solarex modules consisted of the characteristic peaks of Ti2p3/2 and Ti2p1/2 with the binding energy of 458.2 eV and 464.6 eV, respectively, which indicates the presence of Ti 4+ . It can be con fi rmed that the powder on the surface of back-sheet is titanium dioxide. The contents of the surface elements for the back-sheet were ana- lysed by XPS are shown in Table 3. The O/C atom ratio increased from 0.14 to 0.40, while the F/C atom ratio decreased from 0.25 to 0.09. This data reveals that the molecular chain has been broken due to the photo oxidation of polymers. The following reactions are most probable (William et al., 1993): Free radical generation: ⎯→ ⎯ = + CH CHFCH CHF CH CH CHCHF HF 2 2 UV 2 e e % (1) e CH 2 CHFCH 2 CHF e + R % → e CH 2 (CF % )CH 2 CHF (2) Chain reaction of oxidation: e CH 2 (CF % )CH 2 CHF + O 2 → e CH 2 (CFOO % )CH 2 CHF (3) Main chain break: e CH 2 C(FOO % )CH 2 CHF + H 2 O → e CH 2 COF + e CH 2 COOH + HF (4) From the result of XPS and FTIR, the change of the carbonaceous structure causes the change of physical and mechanical properties of the Tedlar sheet, such as tensile strength and maximum elongation at break, which fi nally leads to the failure of back-sheet. Table 2 Values of physical parameters of the back-sheet. Item Water vapor transmission rate (gm/[m 2 - day]) Tensile strength (MPa) Maximum elongation at fracture (%) Breakdown voltage (kV) MD TD MD TD The back-sheet of a Solarex module 8.21 46.52 73.18 184.4 122.9 6.66 New Tedlar PV2001 with no aging 2.4 150 180 26 Note: MD = Machine direction, TD = Transverse direction. Fig. 8. The SEM image of the back-sheet: (a) Back-sheet from Solarex modules after more than 30 years on sun. (b) Unexposed Tedlar back-sheet. Fig. 9. The infrared spectrogram for the area of back-sheet under the module label of Solarex (a), and for a new Tedlar PV2001 with no aging (b) and the exposured area of the back-sheet of a Solarex module (c). H. Han et al. Solar Energy 170 (2018) 510–519 516 (1) C1s scanning spectrum (2) F1s scanning spectrum (3) O1s scanning spectrum (4) Ti2p scanning spectrum Fig. 10. XPS survey spectra obtained from (a) The back-sheet of a Solarex module. (b) A new Tedlar PV2001 fi lm; and (1) – (4) for the high resolution scanning spectra of C1s, F1s, O1s and Ti2p, respectively. H. Han et al. Solar Energy 170 (2018) 510–519 517 3.3.2. The optical loss analysis of EVA The yellowing of EVA is the most signi fi cant defect for the Solarex modules. Table 4 showed the measured parameters of EVA compared to a typical unexposed EVA as reference. The higher yellowing index (YI = 11.5) and the decline in transmittance are signi fi cant signs of EVA aging. The transmittance of EVA measured at the edge of the cell module was higher than that in the center, for wavelengths from 380 nm to 1200 nm as shown in Fig. 11. However, the transmittance of EVA in the center was higher than that at the edge in the wavelengths in the 300 – 380 nm range. This di ff erence maybe because the ultravoilet ab- sorbers in EVA in the center have been completely consumed (Pern et al., 1991). The other and most probable explanation of the di ff erence between the center and the edge of the module is the bleaching of EVA due to the di ff usion of oxygen, which is more important at the edge of the module and at the edges of the cells than at the center of the cells. In PV device, a photon can generate an electron-hole pair. We can count the number of generated electrons according to the number of incident photos, and the photocurrent can be calculated by multiplying the number of electron with the quantity of electron charge. Then the photon fl ux at each wavelength can be transformed into current density (Park et al., 2013). The photo current loss corresponding to the degradation of the optical properties of EVA after UV exposure is was calculated using the measured spectral transmittance of EVA and using the AM1.5 solar spectrum, and integrating the photo current density in the wavelength range from 300 nm to 1100 nm, the results is plotted in Fig. 12. The total loss in short-circuit density of the edge and center of the Solarex module is 10.12% and 15.08%, respectively, compared to a typical EVA encapsulant before exposure to UV. The result of optical loss agrees with the result of the electrical performance analysis, which con fi rms that the power degradation of the Solarex modules mainly caused by yellowing of EVA and the degradation of the short circuit current. 4. Conclusions A PV array composed of 196 Solarex modules has been dismantled after a 22-year exposure to sunshine and the environment in a tropical region of China, and of which 144 modules were re-assembled in Guangzhou, a sub-tropical region of China, for continued operation and exposure to the environment. The modules were measured every 2 years after their re-installation. After the fi rst period of 22 years, the short-circuit current I sc and maximum power point current I m were degraded by 7.89% and 11.35% respectively. After re-installation, they both continued to degrade at a rate of about 0.5% per year to reach after 28 years of combined exposure a total degradation of 12.5% and Table 3 XPS for surface element analysis. Element New Tedlar PV2001 fi lm (%) Tedlar back-sheet of Solarex module (%) C 72.37 65.19 F 17.77 5.77 O 9.86 26.05 Ti 0 1.99 Table 4 The values of measured parameters of EVA compared to an exposed typical EVA. Item EVA from the Solarex module The typical unexposed EVA Crosslink degree 73 80 – 90 Transmittance (%) 63 91 – 93 Yellowing index 11.5 0.9 Tensile strength (Mpa) 3 18 Fracture growth rate (%) 697 800 Fig. 11. The transmittance of EVA in di ff erent regions of the modules. Fig. 12. The photo current loss based on di ff erent measured transmittances: (a) the current density versus wavelength based on the di ff erent transmittances at di ff erent regions of the Solarex module and (b) the short circuit current density based on the di ff erent measured transmittances at di ff erent regions of the Solarex module. H. Han et al. Solar Energy 170 (2018) 510–519 518 14.9% respectively compared to the nameplate value. The total power degradation was only 6.53% after 28 years compared to the nameplate value of the modules. The open-circuit voltage remained almost un- changed with a 1.54% degradation after 26 years. The maximum power-point voltage and fi ll factor increased by 9.6% and 8.2% re- spectively. The increased of V m and FF is possibly due to the non-uni- formity and mismatch in degraded photo-current between cells within a module, creating an arti fi cially “ improved ” FF and V m , and also due to the uncertainty of the exact original parameters of the modules. The main visually observed defects, after the fi rst 22 years of operation on sun, are (1) a signi fi cant yellowing of the EVA causing a degradation of the optical transmittance in agreement with the loss in I sc , (2) a severe degradation of the PVF back-sheet resulting in cracks causing cracks in cells and corrosion of the metallic fi ngers or interconnects, as well as causing bubbles and delamination of the EVA. An analysis of the PVF back-sheet has shown a signi fi cant degradation of the microstructure of the back-sheet ( “ powdering ” ) resulting in an increase in WVTR and a severe degradation of the tensile strength and maximum elongation as the cause of the cracks. Acknowledgment The authors gratefully acknowledge the fi nancial support from the National High-tech R&D Program of China (863 Program; Grant No. 2015AA050303). References Anand, P., Wang, J., 2003. Modi fi ed failure mode and e ff ects analysis using approximate reasoning. Reliab. Eng. Syst. Saf. 79, 69 – 85. Antonella R., 2003, Mean Time before Failure of Photovoltaic Modules. Federal O ffi ce for Education and Science Final Report BBW 99.0579. Carr, A.J., Prior, T.L., 2004. A comparison of the performance of di ff erent PV module types in temperate climates. Sol. Energy 76, 285 – 294. Charki, A., Laronde, R., Bigaud, D., 2013. Accelerated degradation testing of a photo- voltaic module. J. Photon. Energy, pp, 033099-1 – 033099-10. Davide, P., et al., 2013. Polycrystalline silicon PV modules performance and degradation over 20years. Prog. Photovolt. Res. Appl. 21, 1004 – 1015. Gagliardi, M., Lenarda, P., Paggi, M., 2017. A reaction-di ff usion formulation to simulate EVA polymer degradation in environmental and accelerated ageing conditions. Sol. Energy Mater. Sol. Cells 164, 93 – 106. Haillant, O., Dumbleton, D., Zielnik, A., 2011. An Arrhenius approach to estimating or- ganic photovoltaic module weathering acceleration factors. Sol. Energy Mater. Sol. Cells 95, 1889 – 1895. Hülsmann, P., Weiß, K.A., Köhl, M., 2014. Temperature-dependent water vapour and oxygen permeation through di ff erent polymeric materials used in photovoltaic- modules. Prog. Photovolt. 22, 415 – 421. Husam, M., Walwil, A., et al., 2017. Comparative studies of encapsulation and glass surface modi fi cation impacts on PV performance in a desert climate. Sol. Energy 142, 288 – 298. IEC 61215, 2005. Crystalline Silicon Terrestrial Photovoltaic (PV) Modules – Design Quali fi cation and Type Approval. International Electrotechnical Commission. Jordan, D.C., Kurtz, S.R., 2013. Photovoltaic degradation rates - an analytical review. Prog. Photovolt. Res. Appl. 21 (1), 12 – 29. Jordan, D.C., et al., 2017. PV degradation curves: non-linearities and failure modes. Prog. Photovolt. Res. Appl. 25, 583 – 591. Jordan, D., et al., 2017. Photovoltaic failure and degradation modes. Prog. Photovolt. Res. Appl. 25 (4), 318 – 326. Kempe, M.D., 2006. Modeling of rates of moisture ingress into