Application of Infrared Thermography in an Adequate Reusability Analysis of Photovoltaic Modules Affected by Hail

Featured Revitalization of PV power plants affected by natural disasters. Abstract: Infrared thermography, in the analysis of photovoltaic (PV) power plants, is a mature technical discipline. In the event of a hailstorm that leaves the PV system without the support of the power grid (and a signiﬁcant portion of the generation potential), thermography is the easiest way to determine the condition of the modules and revive the existing system with the available resources. This paper presents research conducted on a 30 kW part of a 420 kW PV power plant, and demonstrates the procedure for inspecting visually correct modules that have suffered from a major natural disaster. The severity of the disaster is shown by the fact that only 14% of the PV modules at the test site remained intact. Following the recommendations of the standard IEC TS 62446-3, a thermographic analysis was performed. The thermographic analysis was preceded by an analysis of the I-V curve, which was presented in detail using two characteristic modules as examples. I-V curve measurements are necessary to relate the measured values of the radiation and the measured contact temperature of the module to the thermal patterns. The analysis concluded that soiled modules must be cleaned, regardless of the degree of soiling. The test results clearly indicated defective module elements that would result in a safety violation if reused. The research shows that the validity criterion deﬁned on the basis of the analysis of the reference module can be supplemented, but can also be replaced by a statistical analysis of several modules. The comparison between the thermographic analysis and the visual inspection clearly conﬁrmed thermography as a complementary method for testing PV-s.


Introduction
Thanks to the EU's energy policy, PV plants in the EU are already paving the way to meet the requirements of nearly zero energy buildings (NZEB) under Article  PV systems are usually insured under a building's insurance policy. The occasion for the investigation conducted in this paper was a severe storm, followed by hail, that hit Požega and its surroundings at 4:20 pm on 25 June 2021. The disaster caused damage to houses, cars, and power poles, mainly in the town of Požega and the settlement of Treštanovci. Part of the town of Požega and its surrounding settlements were without electricity [1]. The severity of the disaster is shown by the fact that the walls of the local prison collapsed, trees were left without leaves, and buildings were left without roofs [2]. According to a report by the Croatian Meteorological and Hydrological Service [3], the

Methodology
Preparations for the tests needed to be carried out with logistical care due to the 99 km distance between the laboratory and the test site. The tests conducted could not interfere with the reconstruction of buildings and production processes of companies that suffered losses due to the disaster. Therefore, a pilot project to test PV modules was conducted prior to on-site deployment. The purpose of the pilot project was to pre-investigate the procedure for determining the correctness criteria, by conducting analyses of the PV modules on racks at an angle corresponding to the site characteristics. The aim of the pilot project was to identify the challenges in performing thermographic analyses. The main objective was to determine the applicability of the available thermal imaging cameras, their acceptance in relation to the regulations in force, and the definition of the equipment required to perform the measurements. A flow chart of the test procedures is shown in Figure 1. The test began by collecting the undamaged panels, cleaning them with a broom and water, and visually inspecting them for visible damage. After initial inspection of the panel, the I-V curve was analyzed. The process of determining the electrical properties was quickly accomplished with the appropriate test equipment. Infrared thermography analysis requires a considerable amount of time, namely 10-15 min, which is what it takes for a panel to reach thermal equilibrium. This was followed by an analysis of the parameters of the thermal pattern in search of the hotspots, which deviate significantly from the standards and the defined class of anomalies of the reference panel. In each of the individual steps, a decision was made regarding the technical correctness and safety of using the module. In this paper, each step is described in detail, focusing on thermography and the determination of the class of abnormality (CoA) criteria.

Description of the Test Site and PV Module Electrical Performance Evaluation
The test site represented a 30 kW part of a 420 kW PV plant, commissioned in November 2020, which was damaged by the hailstorm on 25 June 2021. The studied test site consisted of 94 PV monocrystalline silicon 340 Wp modules (30.08 kWp) and a 30 kW three-phase grid-tie inverter. The hailstorm visibly (totally) destroyed 81 out of 94 PV modules in the PV array, leaving 13 PV modules with no visible damage on PV module's glass cover. The damaged PV modules are being replaced with new 340 Wp PV modules. Figure 2 shows the damage to the PV array caused by the hailstorm, while Figure 3 shows a close-up of a part of the PV array test site. The damage shown in the figures can also be shown using numerical indicators that define the failure share of the total system part [9]. The test began by collecting the undamaged panels, cleaning them with a broom and water, and visually inspecting them for visible damage. After initial inspection of the panel, the I-V curve was analyzed. The process of determining the electrical properties was quickly accomplished with the appropriate test equipment. Infrared thermography analysis requires a considerable amount of time, namely 10-15 min, which is what it takes for a panel to reach thermal equilibrium. This was followed by an analysis of the parameters of the thermal pattern in search of the hotspots, which deviate significantly from the standards and the defined class of anomalies of the reference panel. In each of the individual steps, a decision was made regarding the technical correctness and safety of using the module. In this paper, each step is described in detail, focusing on thermography and the determination of the class of abnormality (CoA) criteria.

Description of the Test Site and PV Module Electrical Performance Evaluation
The test site represented a 30 kW part of a 420 kW PV plant, commissioned in November 2020, which was damaged by the hailstorm on 25 June 2021. The studied test site consisted of 94 PV monocrystalline silicon 340 Wp modules (30.08 kWp) and a 30 kW three-phase grid-tie inverter. The hailstorm visibly (totally) destroyed 81 out of 94 PV modules in the PV array, leaving 13 PV modules with no visible damage on PV module's glass cover. The damaged PV modules are being replaced with new 340 Wp PV modules. Figure 2 shows the damage to the PV array caused by the hailstorm, while Figure 3 shows a close-up of a part of the PV array test site. The damage shown in the figures can also be shown using numerical indicators that define the failure share of the total system part [9].  The layout of the 30 kW PV plant with its PV module configuration on the DC sid is given in Figure 4. The technical characteristics of the original 340 Wp PV modules an the new 345 Wp PV modules are provided in Table 1.     The layout of the 30 kW PV plant with its PV module configuration on the DC sid is given in Figure 4. The technical characteristics of the original 340 Wp PV modules an the new 345 Wp PV modules are provided in Table 1    The layout of the 30 kW PV plant with its PV module configuration on the DC side is given in Figure 4. The technical characteristics of the original 340 Wp PV modules and the new 345 Wp PV modules are provided in Table 1.  The layout of the 30 kW PV plant with its PV module configuration on the DC side is given in Figure 4. The technical characteristics of the original 340 Wp PV modules and the new 345 Wp PV modules are provided in Table 1

Pilot Project of Infrared Thermographic PV Analysis
According to IEC TS 62446-3 Photovoltaic (PV) systems-Requirements for testing, documentation and maintenance-Part 3: Photovoltaic modules and plants-outdoor infrared thermography use with an IR camera with a resolution of ≥320 × 240 pixels and a separate photo camera are recommended [10]. It is recommended to use cameras that have been calibrated within the previous two years. To conduct thermographic analyses, we used two infrared thermographic cameras, Flir E60bx and Flir E6. Both cameras were IP 54 and had an operating time of up to 4 h. The camera differences are listed in Table 2. The significant differences in the cameras (for analysis purposes), in addition to the resolutions, are the optical characteristics of the field of view (FOV) and the instantaneous field of view (IFOV). Figure 5 illustrates the minimum distances of the cameras from the module to fit the complete module, described by Table 2, taken at an angle of 90 • .
From this, we can conclude that the Flir E6 camera is not suitable for the analysis due to its lower resolution; however, in combination with the aforementioned IFOV optics, it can be used as a control and for additional analyses when the analysis object cannot be viewed from a greater distance. From the aspect of documenting the visual image, none of the cameras met the requirement of 9 Mpix visual image resolution. Therefore, photo documentation, for the purposes of visual analysis, was carried out using the cameras. Since the cameras have not been calibrated for more than two years, the analysis was performed using a VOLTCRAFT RS-350 with an uncertainty measurement of ± 0.5 • C at 100 • C. The calibration results are shown in Figure 6. The significant differences in the cameras (for analysis purposes), in addition to resolutions, are the optical characteristics of the field of view (FOV) and the instantane field of view (IFOV). Figure 5 illustrates the minimum distances of the cameras from module to fit the complete module, described by Table 2, taken at an angle of 90°. From this, we can conclude that the Flir E6 camera is not suitable for the analysis due to its lower resolution; however, in combination with the aforementioned IFOV optics, it can be used as a control and for additional analyses when the analysis object cannot be viewed from a greater distance. From the aspect of documenting the visual image, none of the cameras met the requirement of 9 Mpix visual image resolution. Therefore, photo documentation, for the purposes of visual analysis, was carried out using the cameras. Since the cameras have not been calibrated for more than two years, the analysis was performed using a VOLTCRAFT RS-350 with an uncertainty measurement of ± 0.5 °C at 100 °C. The calibration results are shown in Figure 6. When analyzing two cameras, calibration data are necessary to reconcile the measurement results because the FLIR E60bx camera shows an average temperature of 0.8 °C higher, while the FLIR E6 shows an average temperature of 2.9 °C higher. To prepare the measurement procedure, a pilot analysis of a small unused module with dimensions of 52 × 32.5 cm, placed on supports intended for field use, was performed. Unfortunately, the nameplate, which should be present in accordance with the EN 50380: 2003 Datasheet and nameplate information for the PV modules (or UL 4730), was not attached to the module. Figure 7 shows a photograph of the pilot setup, taken at 720 nm. The expected values of the temperature difference, according to the literature [11], were in the range of 10 °C. Figure 8 shows the temperature distribution on the module, recorded by the FLIR E60bx camera.  When analyzing two cameras, calibration data are necessary to reconcile the measurement results because the FLIR E60bx camera shows an average temperature of 0.8 • C higher, while the FLIR E6 shows an average temperature of 2.9 • C higher. To prepare the measurement procedure, a pilot analysis of a small unused module with dimensions of 52 × 32.5 cm, placed on supports intended for field use, was performed. Unfortunately, the nameplate, which should be present in accordance with the EN 50380: 2003 Datasheet and nameplate information for the PV modules (or UL 4730), was not attached to the module. Figure 7 shows a photograph of the pilot setup, taken at 720 nm. From this, we can conclude that the Flir E6 camera is not suitable for the analysis due to its lower resolution; however, in combination with the aforementioned IFOV optics, it can be used as a control and for additional analyses when the analysis object cannot be viewed from a greater distance. From the aspect of documenting the visual image, none of the cameras met the requirement of 9 Mpix visual image resolution. Therefore, photo documentation, for the purposes of visual analysis, was carried out using the cameras. Since the cameras have not been calibrated for more than two years, the analysis was performed using a VOLTCRAFT RS-350 with an uncertainty measurement of ± 0.5 °C at 100 °C. The calibration results are shown in Figure 6. When analyzing two cameras, calibration data are necessary to reconcile the measurement results because the FLIR E60bx camera shows an average temperature of 0.8 °C higher, while the FLIR E6 shows an average temperature of 2.9 °C higher. To prepare the measurement procedure, a pilot analysis of a small unused module with dimensions of 52 × 32.5 cm, placed on supports intended for field use, was performed. Unfortunately, the nameplate, which should be present in accordance with the EN 50380: 2003 Datasheet and nameplate information for the PV modules (or UL 4730), was not attached to the module. Figure 7 shows a photograph of the pilot setup, taken at 720 nm. The expected values of the temperature difference, according to the literature [11], were in the range of 10 °C. Figure 8 shows the temperature distribution on the module, recorded by the FLIR E60bx camera. The expected values of the temperature difference, according to the literature [11], were in the range of 10 • C. Figure 8 shows the temperature distribution on the module, recorded by the FLIR E60bx camera. Appl. Sci. 2022, 12, x FOR PEER REVIEW 7 of 24  Figure 9 shows the thermal patterns of the bottom and top sides of the module, taken with a wide-angle camera, showing higher temperature values on the glass than a module shot with a narrow-angle camera at a higher distance. This analysis indicates that the influence of operator radiation during the thermographic analysis of the PV modules was not significant. A comparison of the thermograms of the modules recorded with the E6 and E60bx cameras clearly indicates that the apparent temperature of the glass surface will vary significantly depending on the shooting angle, due to the environmental effects. Under real conditions expected in the field, it would not be possible to determine the exact share of reflected radiation that may occur at microlocations. This will be especially true for changes in the position of the sun, which heats the environment, and the elements of which become thermal emitters. The extent to which the camera position affects the reading of the temperature value is best illustrated in Table 3, which provides a comparison of the results of recording the apparent temperature of the glass surface and the substrate temperature using two cameras.  Figure 9 shows the thermal patterns of the bottom and top sides of the module, taken with a wide-angle camera, showing higher temperature values on the glass than a module shot with a narrow-angle camera at a higher distance. This analysis indicates that the influence of operator radiation during the thermographic analysis of the PV modules was not significant.  Figure 9 shows the thermal patterns of the bottom and top sides of the module, taken with a wide-angle camera, showing higher temperature values on the glass than a module shot with a narrow-angle camera at a higher distance. This analysis indicates that the influence of operator radiation during the thermographic analysis of the PV modules was not significant. A comparison of the thermograms of the modules recorded with the E6 and E60bx cameras clearly indicates that the apparent temperature of the glass surface will vary significantly depending on the shooting angle, due to the environmental effects. Under real conditions expected in the field, it would not be possible to determine the exact share of reflected radiation that may occur at microlocations. This will be especially true for changes in the position of the sun, which heats the environment, and the elements of which become thermal emitters. The extent to which the camera position affects the reading of the temperature value is best illustrated in Table 3, which provides a comparison of the results of recording the apparent temperature of the glass surface and the substrate temperature using two cameras. A comparison of the thermograms of the modules recorded with the E6 and E60bx cameras clearly indicates that the apparent temperature of the glass surface will vary significantly depending on the shooting angle, due to the environmental effects. Under real conditions expected in the field, it would not be possible to determine the exact share of reflected radiation that may occur at microlocations. This will be especially true for changes in the position of the sun, which heats the environment, and the elements of which become thermal emitters. The extent to which the camera position affects the reading of the temperature value is best illustrated in Table 3, which provides a comparison of the results of recording the apparent temperature of the glass surface and the substrate temperature using two cameras.  Table 3 shows a significant deviation in the measured apparent temperature values on glass compared to those measured on the plastic base of the module. In addition to the emissivity itself, the camera position had a significantly greater impact, which, in the case of the E6, was on the right side and, in the case of the E60bx, was on the left side of the module. The results indicated that the impact of the environment was significantly expressed in the process of thermographic analyses. While the difference in the reading on larger plastic surfaces of 0.95 • C was below the camera limit accuracy, the difference in the reading on the glass surfaces was more than 10 • C on average, which was a significant deviation from the definition of the ∆T (temperature difference) criteria [12].
The indicators of the detailed analysis of the PV module at a short-circuit state current of 1.07 A, carried out at points corresponding to the position of all individual semiconductor elements (cells) of the module, are shown in Table 4. Table 4 shows the temperature that can be measured on the board as a whole, but also on the semiconductor elements. In order to compare the measured temperature values, it was necessary to know the limitations provided by the standards. According to IEC 61215 [13], modules must undergo several thermal cycles, from −40 • C to 85 • C, and hail in test IV 9 impacts, 3 ⁄4 "-45 mph and test V 10 impacts 1"-52 mph (typical assumed in PV standards 25-75 mm hail, 7.53-39.5 m/s [24]). According to IEC TS 63126 [14], the upper limit of temperature cycling is 95 ± 2 • C for rating modules as temperature level 1 and 105 ± 2 • C for rating modules as temperature level 2. Another standard that can be applied in setting limits can be the IEC 61730-2 Photovoltaic (PV) module safety qualification-Part 2: Requirements for testing [25]. IEC 61730-2 sets temperature limits for components, wiring compartments, and fibers at 90 • C, laminated phenolic composition at 125 • C, molded phenolic composition of 150 • C and field wiring terminals, and metal parts at 30 • C above ambient. As can be seen in the thermograms, recordings were within the regulations. The floor coverings in the background of the modules were, on average, between 38 • C and 42 • C, so none of the elements met the criterion of a temperature higher than 30 • C above ambient, i.e., 68 • C-72 • C, allowed by the standard. The answer to the question of the maximum temperature value the module can withstand was also sought in the production process [26]. During the production process, modules are heated up to 170 • C. Cure reaction temperatures of 150 • C and 160 • C are not adequate; high cure temperatures of up to 170 • C and/or long cure times can generate acetophenone, which causes yellowing in the EVA (Ethylene vinyl acetate). According to more recent literature [21], lamination takes place at 150 • C. According to [15], the operating temperature range to be expected in practice is between 30 • C and 80 • C. The reason for the analysis of potential temperature extremes is the fact that the power plant is no longer under normal operations, and testing is only possible with modules experiencing short circuit, i.e., at the highest possible current, which also leads to the highest thermal stress. Otherwise, under normal operations, the increase in temperature of the module parts can be expressed by a simplified expression that takes into account wind speed, irradiance, and ambient temperature [18], given as (1): where T a is ambient temperature in K, T c is cell/module operating temperature in K, G T is solar irradiance in W/m 2 , and V f is wind speed, ranging between 1-15 m/s.

Thermographic Validity Criterion
The new modules used for the power plant renovation were subjected to analysis of the thermal pattern in the short-circuit state, which can be seen in Figure 10. At the time of recording, the solar irradiance was 677 W/m 2 and the contact measured module temperature was 43.8 • C. elements met the criterion of a temperature higher than 30 °C above ambient, i.e., 68 °C-72 °C, allowed by the standard. The answer to the question of the maximum temperature value the module can withstand was also sought in the production process [26]. During the production process, modules are heated up to 170 °C. Cure reaction temperatures o 150 °C and 160 °C are not adequate; high cure temperatures of up to 170 °C and/or long cure times can generate acetophenone, which causes yellowing in the EVA (Ethylene viny acetate). According to more recent literature [21], lamination takes place at 150 °C. Ac cording to [15], the operating temperature range to be expected in practice is between 30 °C and 80 °C. The reason for the analysis of potential temperature extremes is the fact tha the power plant is no longer under normal operations, and testing is only possible with modules experiencing short circuit, i.e., at the highest possible current, which also leads to the highest thermal stress. Otherwise, under normal operations, the increase in temper ature of the module parts can be expressed by a simplified expression that takes into ac count wind speed, irradiance, and ambient temperature [18], given as (1): where Ta is ambient temperature in K, Tc is cell/module operating temperature in K, GT is solar irradiance in W/m 2 , and Vf is wind speed, ranging between 1-15 m/s.

Thermographic Validity Criterion
The new modules used for the power plant renovation were subjected to analysis o the thermal pattern in the short-circuit state, which can be seen in Figure 10. At the time of recording, the solar irradiance was 677 W/m 2 and the contact measured module tem perature was 43.8 °C. When performing the thermographic analysis, it was necessary to continuously con sider the recording angle because the emissivity parameter takes on different values. Dif ferent sources give different information on the emissivity and the viewing angle perpen dicular to the module, where shooting at an angle of 5-60° is acceptable [19]. FLIR in par ticular points out that the emissivity of glass is a great challenge. "Even though glass has an emissivity of 0.85-0.90 in the 8-14 μm waveband, thermal measurements on glass sur faces are not easy to do". One of the first papers to provide information on temperature differences taken from a helicopter [20] showed a graphical representation of the emissiv ity, given in Figure 11. When performing the thermographic analysis, it was necessary to continuously consider the recording angle because the emissivity parameter takes on different values. Different sources give different information on the emissivity and the viewing angle perpendicular to the module, where shooting at an angle of 5-60 • is acceptable [19]. FLIR in particular points out that the emissivity of glass is a great challenge. "Even though glass has an emissivity of 0.85-0.90 in the 8-14 µm waveband, thermal measurements on glass surfaces are not easy to do". One of the first papers to provide information on temperature differences taken from a helicopter [20] showed a graphical representation of the emissivity, given in Figure 11. The analysis of all previous findings, presented in [21], from the emissivit can be seen in Figure 12. The work in [21] states that emissivity should be selected as 0.85 for gla for a polymer backsheet, if the view angle is within 90-60° (glass) and 90-45° FLIR's recommendation is visible in the form of the green space in Figure 12. I on emissivity changes is especially important when conducting an analysis of a composed of several different thermographic records is necessary because, this is the only way we can get complete information about the temperature d of the surface of a PV module. Table 5 provides information on the temperatu tion on the module shown in Figure 10. Table 5. Distribution of the apparent temperature value of the new (reference) PV mod The analysis of all previous findings, presented in [21], from the emissivity spectrum can be seen in Figure 12. The analysis of all previous findings, presented in [21], from the emissivit can be seen in Figure 12. The work in [21] states that emissivity should be selected as 0.85 for gla for a polymer backsheet, if the view angle is within 90-60° (glass) and 90-45° FLIR's recommendation is visible in the form of the green space in Figure 12. In on emissivity changes is especially important when conducting an analysis of a composed of several different thermographic records is necessary because, s this is the only way we can get complete information about the temperature d of the surface of a PV module. Table 5 provides information on the te distribution on the module shown in Figure 10. The work in [21] states that emissivity should be selected as 0.85 for glass and 0.95 for a polymer backsheet, if the view angle is within 90-60 • (glass) and 90-45 • (polymer). FLIR's recommendation is visible in the form of the green space in Figure 12. Information on emissivity changes is especially important when conducting an analysis of a larger area composed of several different thermographic records is necessary because, sometimes, this is the only way we can get complete information about the temperature distribution of the surface of a PV module. Table 5 provides information on the temperature distribution on the module shown in Figure 10. Analyzing the data, we can conclude that the maximum temperature difference of the individual parts of the module is 27.9 • C. Table 6 gives the statistical data of the apparent temperature of the new module. Considering the pilot project results, where the apparent temperature only increased by 17%, the new module in the short-circuit state shows a temperature increase of 67% in some parts. Considering that the short-circuit current of the new module is up to 8 times higher, it is clear that the pattern of determining the ∆T criterion depends on the short-circuit current and will be unique for each module. The limiting criterion will be the maximum temperature that individual parts of the module can withstand under normal operation. From the above, it is clear why the IEC TS 62446-3:2017 standard does not specify temperature values, and only the three basic classes of abnormality are specified, as listed in Table 7. The situation analyzed in this paper is described in [27], where the absolute temperatures of the hotspots were 62.1 • C and 101.4 • C, compared to the 58.4 • C of a healthy cell. The temperature difference between a low-temperature hotspot and a healthy cell was 3.7 • C, and the maximum temperature difference was 43 • C. The work in [16] gave similar data in the list of identified module defects, where the maximum registered temperature difference was 42.53 • C. In [17], an IR image revealed the existence of two hot spots reaching 86 • C, a relative increase of more than 35 • C with respect to the temperature of nearby cells. Furthermore, the cell temperatures measured at the back of the module were about 7-10 • C higher than those measured at the front.

I-V Curve Measurements of Studied Modules
Preparing the implementation of the on-site measurements is shown in Figure 13. The position of the modules was adjusted to the actual slope of the halls where the modules were located. The shadow created by the stands was used to adjust the position of the module. A photograph in Figure 13 shows the measurement procedure, in which the I-V curve of the PV modules was measured, followed by thermographic analysis of the modules in the short-circuit state.
Appl. Sci. 2022, 12, x FOR PEER REVIEW 12 of 24 °C, and the maximum temperature difference was 43 °C. The work in [16] gave similar data in the list of identified module defects, where the maximum registered temperature difference was 42.53 °C. In [17], an IR image revealed the existence of two hot spots reaching 86 °C, a relative increase of more than 35 °C with respect to the temperature of nearby cells. Furthermore, the cell temperatures measured at the back of the module were about 7-10 °C higher than those measured at the front.

I-V Curve Measurements of Studied Modules
Preparing the implementation of the on-site measurements is shown in Figure 13. The position of the modules was adjusted to the actual slope of the halls where the modules were located. The shadow created by the stands was used to adjust the position of the module. A photograph in Figure 13 shows the measurement procedure, in which the I-V curve of the PV modules was measured, followed by thermographic analysis of the modules in the short-circuit state. PV modules with no visible damage to the glass cover were subjected to I-V curve measurements to evaluate the influence of the hailstorm on their performance. The PV modules' I-V curve measurements were performed using an IEC (EN) 62446 standard compliant PV tester, the Metrel MI 3108 EurotestPV. This measurement procedure estimated a PV module's I-V curve under standard test conditions (STCs) based on the PV module technical characteristics and the current environmental conditions (solar irradiance and PV module temperature), which was easily comparable to the manufacturer's I-V curve under the STCs. This enabled an assessment of hailstorm damage on PV module performance. The technical characteristics of the PV tester I-V curve measurement module, along with the PV solar irradiance sensor (temperature compensated monocrystalline PV cell) and cell temperature sensor, are given in Table 8. The PV module I-V curve measurement procedure is given in Figure 14. The procedure for measuring the I-V curve with the PV tester is given in Figure 14.  PV modules with no visible damage to the glass cover were subjected to I-V curve measurements to evaluate the influence of the hailstorm on their performance. The PV modules' I-V curve measurements were performed using an IEC (EN) 62446 standard compliant PV tester, the Metrel MI 3108 EurotestPV. This measurement procedure estimated a PV module's I-V curve under standard test conditions (STCs) based on the PV module technical characteristics and the current environmental conditions (solar irradiance and PV module temperature), which was easily comparable to the manufacturer's I-V curve under the STCs. This enabled an assessment of hailstorm damage on PV module performance. The technical characteristics of the PV tester I-V curve measurement module, along with the PV solar irradiance sensor (temperature compensated monocrystalline PV cell) and cell temperature sensor, are given in Table 8. The PV module I-V curve measurement procedure is given in Figure 14. The procedure for measuring the I-V curve with the PV tester is given in Figure 14.  Figure 14. PV module I-V curve measurement procedure with the PV tester.
PV module I-V curve measurements are given for 2 out of 13 PV modules visible damage to the glass surface. The environmental conditions at the mome curve measurement for both PV modules are given in Table 9. The numerical and graphical results of the I-V curve measurements of (washed) PV module are given in Table 10 and Figure 15; Figure 15 shows both P-V curves. The results are given for three scenarios-measured values, calcula values (measured data estimated for STC values based on manufacturer data), a inal STC values representing data from the manufacturer's datasheet.  PV module I-V curve measurements are given for 2 out of 13 PV modules with no visible damage to the glass surface. The environmental conditions at the moment of I-V curve measurement for both PV modules are given in Table 9. The numerical and graphical results of the I-V curve measurements of the first (washed) PV module are given in Table 10 and Figure 15; Figure 15 shows both I-V and P-V curves. The results are given for three scenarios-measured values, calculated STC values (measured data estimated for STC values based on manufacturer data), and nominal STC values representing data from the manufacturer's datasheet.  The numerical and graphical results of the I-V curve measurements of the second PV module are given in Table 11 and Figure 16.   The numerical and graphical results of the I-V curve measurements of the second PV module are given in Table 11 and Figure 16.  The numerical and graphical results of the I-V curve measurements of the second PV module are given in Table 11 and Figure 16.   The results indicate that both PV modules degrade significantly in terms of electrical performance in comparison to the manufacturer's data even though the modules have been operating for only one year and their degradation should not be as high as presented levels. The most significant deviation is observed in the output power, which decreases by more than 20 % for both PV modules. Figure 17 shows the thermogram of the module removed from the dust-covered roof. Upon washing, the apparent temperature values were significantly reduced. A temperature increase in the upper left corner of the triangle was also observed as a result of the reflected radiation from the production hall on the left side of the test site, which can be seen in Figure 13 of the measurement setup. The reflected temperature was read from the chromeplated stand with the emissivity set to 1. During the thermographic measurements, it was found that the used cameras were not suitable for measurements at high solar irradiance levels due to the poor visibility of the displays. It is recommended to use cameras with optical viewfinder. The results indicate that both PV modules degrade significantly in terms of electrical performance in comparison to the manufacturer's data even though the modules have been operating for only one year and their degradation should not be as high as presented levels. The most significant deviation is observed in the output power, which decreases by more than 20 % for both PV modules. Figure 17 shows the thermogram of the module removed from the dust-covered roof. Upon washing, the apparent temperature values were significantly reduced. A temperature increase in the upper left corner of the triangle was also observed as a result of the reflected radiation from the production hall on the left side of the test site, which can be seen in Figure 13 of the measurement setup. The reflected temperature was read from the chrome-plated stand with the emissivity set to 1. During the thermographic measurements, it was found that the used cameras were not suitable for measurements at high solar irradiance levels due to the poor visibility of the displays. It is recommended to use cameras with optical viewfinder. Due to the observed significant effect of shading, because of soiling, on the thermal pattern, all modules were subjected to the washing process. The procedure was carried out in three steps. The first stage was the removal of large particles and the rest of the glass with a whisk, followed by washing, which dissolves the stubborn deposits of particles from the module. This was followed by rinsing the surface with clean water and wiping with a clean cloth. Figure 18 shows the appearance of the module before washing and the necessary equipment on the ground. On average, one litre of water per module was needed to clean the modules.  Figure 19 shows the need for washing, i.e., the effects of impurities on the thermal pattern of the module, which is common due to shading [28]. Figure 19 illustrates the Due to the observed significant effect of shading, because of soiling, on the thermal pattern, all modules were subjected to the washing process. The procedure was carried out in three steps. The first stage was the removal of large particles and the rest of the glass with a whisk, followed by washing, which dissolves the stubborn deposits of particles from the module. This was followed by rinsing the surface with clean water and wiping with a clean cloth. Figure 18 shows the appearance of the module before washing and the necessary equipment on the ground. On average, one litre of water per module was needed to clean the modules. The results indicate that both PV modules degrade significantly in terms of electrica performance in comparison to the manufacturer's data even though the modules hav been operating for only one year and their degradation should not be as high as presente levels. The most significant deviation is observed in the output power, which decrease by more than 20 % for both PV modules. Figure 17 shows the thermogram of the module removed from the dust-covered roo Upon washing, the apparent temperature values were significantly reduced. A tempera ture increase in the upper left corner of the triangle was also observed as a result of th reflected radiation from the production hall on the left side of the test site, which can b seen in Figure 13 of the measurement setup. The reflected temperature was read from th chrome-plated stand with the emissivity set to 1. During the thermographic measure ments, it was found that the used cameras were not suitable for measurements at hig solar irradiance levels due to the poor visibility of the displays. It is recommended to us cameras with optical viewfinder. Due to the observed significant effect of shading, because of soiling, on the therma pattern, all modules were subjected to the washing process. The procedure was carrie out in three steps. The first stage was the removal of large particles and the rest of th glass with a whisk, followed by washing, which dissolves the stubborn deposits of parti cles from the module. This was followed by rinsing the surface with clean water and wip ing with a clean cloth. Figure 18 shows the appearance of the module before washing an the necessary equipment on the ground. On average, one litre of water per module wa needed to clean the modules.  Figure 19 shows the need for washing, i.e., the effects of impurities on the therma pattern of the module, which is common due to shading [28]. Figure 19 illustrates th  Figure 19 shows the need for washing, i.e., the effects of impurities on the thermal pattern of the module, which is common due to shading [28]. Figure 19 illustrates the temperature deviation of the readings before and after cleaning. The analysis was performed on the first 5 rows whose measurement data were not affected by the reflected radiation of the production hall, which is difficult to see on the dirty module. The significance of the influence of shading is clearly seen in Figure 17, where the temperature values of individual cells after cleaning cannot determine the pattern of stochastic change.

Influence of Soiling on the Infrared Thermographic Pattern
Appl. Sci. 2022, 12, x FOR PEER REVIEW 16 of temperature deviation of the readings before and after cleaning. The analysis was pe formed on the first 5 rows whose measurement data were not affected by the reflecte radiation of the production hall, which is difficult to see on the dirty module. The signi cance of the influence of shading is clearly seen in Figure 17, where the temperature valu of individual cells after cleaning cannot determine the pattern of stochastic change.

Infrared Thermographic Analysis of Undamaged Modules
After the initial analysis, the identification of potential reflections on the site, an taking into account the time needed to inspect one module on average 27 min, it becam clear that only 50 % of the available modules can be inspected during the day under th same conditions. The following thermograms show the thermal patterns of the modul in the short-circuit state. To facilitate visual comparison, the temperature range of the pa let is set to a temperature value from 20 °C to 100 °C. Figure 20 shows the thermograph analysis of module number 2 (left) and number 3 (right).  Figure 21 shows the thermographic analysis of the front side (up) and the backsid (down) of module number 4. A hot spot was noticed on module number 4, which can lea to module degradation. The observed anomaly significantly exceeds the temperatur predicted by the standards. Based on previous findings, we decided to investigate th background of the module using a camera with a wider shooting angle FLIR E6 and d termine the exact temperature values. The reflected radiation of the substrate was take from the images recorded with E60bx.

Infrared Thermographic Analysis of Undamaged Modules
After the initial analysis, the identification of potential reflections on the site, and taking into account the time needed to inspect one module on average 27 min, it became clear that only 50 % of the available modules can be inspected during the day under the same conditions. The following thermograms show the thermal patterns of the modules in the short-circuit state. To facilitate visual comparison, the temperature range of the pallet is set to a temperature value from 20 • C to 100 • C. Figure 20 shows the thermographic analysis of module number 2 (left) and number 3 (right).
Appl. Sci. 2022, 12, x FOR PEER REVIEW 16 of 24 temperature deviation of the readings before and after cleaning. The analysis was performed on the first 5 rows whose measurement data were not affected by the reflected radiation of the production hall, which is difficult to see on the dirty module. The significance of the influence of shading is clearly seen in Figure 17, where the temperature values of individual cells after cleaning cannot determine the pattern of stochastic change.

Infrared Thermographic Analysis of Undamaged Modules
After the initial analysis, the identification of potential reflections on the site, and taking into account the time needed to inspect one module on average 27 min, it became clear that only 50 % of the available modules can be inspected during the day under the same conditions. The following thermograms show the thermal patterns of the modules in the short-circuit state. To facilitate visual comparison, the temperature range of the pallet is set to a temperature value from 20 °C to 100 °C. Figure 20 shows the thermographic analysis of module number 2 (left) and number 3 (right).  Figure 21 shows the thermographic analysis of the front side (up) and the backside (down) of module number 4. A hot spot was noticed on module number 4, which can lead to module degradation. The observed anomaly significantly exceeds the temperatures predicted by the standards. Based on previous findings, we decided to investigate the background of the module using a camera with a wider shooting angle FLIR E6 and determine the exact temperature values. The reflected radiation of the substrate was taken from the images recorded with E60bx.  Figure 21 shows the thermographic analysis of the front side (up) and the backside (down) of module number 4. A hot spot was noticed on module number 4, which can lead to module degradation. The observed anomaly significantly exceeds the temperatures predicted by the standards. Based on previous findings, we decided to investigate the background of the module using a camera with a wider shooting angle FLIR E6 and determine the exact temperature values. The reflected radiation of the substrate was taken from the images recorded with E60bx. The comparison of the sizes illustrated in the figures shows that rameters were well chosen. The learned hot spot, with its temperature °C, can lead to the destruction of the module, so it was declared defectiv Figure 22 shows thermographic analyses of modules 5 and 6, while   The comparison of the sizes illustrated in the figures shows that the recording parameters were well chosen. The learned hot spot, with its temperature magnitude of 114 • C, can lead to the destruction of the module, so it was declared defective for further use. Figure 22 shows thermographic analyses of modules 5 and 6, while Figure 23 shows the thermographic analyses of modules 7 and 8. The comparison of the sizes illustrated in the figures shows that the recording parameters were well chosen. The learned hot spot, with its temperature magnitude of 114 °C, can lead to the destruction of the module, so it was declared defective for further use. Figure 22 shows thermographic analyses of modules 5 and 6, while Figure 23 shows the thermographic analyses of modules 7 and 8.   The analyzed modules showed different thermal patterns, resulting from the shortcircuit current at different solar irradiance values. The recorded values corresponded to the information found in [18,20,29]. Table 12 shows the values of solar irradiance and contact measurement values of the module temperatures that preceded the thermographic test. A graphical representation of the solar irradiance over the thermographic measurements on the modules is provided in Figure 24. The graphical representation of solar irradiance over the thermographic measurements must be corrected in order to get precise data according to (1), using the data presented in Figure 25. The average wind speed for the studied location in September is 1.7 m/s, based on the Croatian Meteorological and Hydrological Service data. The analyzed modules showed different thermal patterns, resulting from the shortcircuit current at different solar irradiance values. The recorded values corresponded to the information found in [18,20,29]. Table 12 shows the values of solar irradiance and contact measurement values of the module temperatures that preceded the thermographic test. A graphical representation of the solar irradiance over the thermographic measurements on the modules is provided in Figure 24. The analyzed modules showed different thermal patterns, resulting f circuit current at different solar irradiance values. The recorded values co the information found in [18,20,29]. Table 12 shows the values of solar irrad tact measurement values of the module temperatures that preceded the t test. A graphical representation of the solar irradiance over the thermogra ments on the modules is provided in Figure 24. The graphical representation of solar irradiance over the thermogra ments must be corrected in order to get precise data according to (1), using sented in Figure 25. The average wind speed for the studied location in Se m/s, based on the Croatian Meteorological and Hydrological Service data.  The graphical representation of solar irradiance over the thermographic measurements must be corrected in order to get precise data according to (1), using the data presented in Figure 25. The average wind speed for the studied location in September is 1.7 m/s, based on the Croatian Meteorological and Hydrological Service data.  If we exclude the extreme hotspots on module number 4 and modu data shown in Table 13 indicate that the difference between the observe minimum temperature values is 20.5 °C, which is an average increase of imum temperature values, averaging 65.8 °C, are not a problem for mo under continuous operation. A comparison with the data from the an modules in Table 6 shows that the data in Table 13 have similar values.   If we exclude the extreme hotspots on module number 4 and module number 8, th data shown in Table 13 indicate that the difference between the observed maximum an minimum temperature values is 20.5 °C, which is an average increase of 45.16%. The max imum temperature values, averaging 65.8 °C, are not a problem for module component under continuous operation. A comparison with the data from the analysis of the new modules in Table 6 shows that the data in Table 13 have similar values. Considering the observed extremes of the hotspots (maximum temperature values o the modules are shown in Figure 27), we can conclude that module 4 is defective, i.e., no suitable for further use because the temperature difference is 65.3 °C. In addition, modul 8 should be monitored, considering the lower value of solar irradiance with a temperatur difference of 33.3 °C. If we exclude the extreme hotspots on module number 4 and module number 8, the data shown in Table 13 indicate that the difference between the observed maximum and minimum temperature values is 20.5 • C, which is an average increase of 45.16%. The maximum temperature values, averaging 65.8 • C, are not a problem for module components under continuous operation. A comparison with the data from the analysis of the new modules in Table 6 shows that the data in Table 13 have similar values. Considering the observed extremes of the hotspots (maximum temperature values of the modules are shown in Figure 27), we can conclude that module 4 is defective, i.e., not suitable for further use because the temperature difference is 65.3 • C. In addition, module 8 should be monitored, considering the lower value of solar irradiance with a temperature difference of 33.3 • C.  The analysis performed on 62% of the total studied modules indicates that another faulty module may occur. Another conclusion is that, if reusing entire modules, it is necessary to monitor at least two modules for further development of potential hotspots. This is due to damage that can develop as the module ages during operation. The results of the visual inspection described in [13] are shown in Figure 28, which illustrate the different forms of damage found on the modules. A detailed overview of all the different module damage types can be found in the NREL report [23]. Interesting examples of damage and appearance of a PV module after 27 years of operation can be found in [22]. A further analysis of the studied modules was done by classifying the visible damage into five levels. Level 1 represents a cell without any visible damage, level 2 represents barely visible damage, level 3 represents visible damage present in the right photograph of Figure 28, level 4 represents damage presented in the centre photograph in Figure 28, and level 5 is visible in the left photograph of Figure 28. The visual inspection and damage classification of module number 3 for each cell is given in Figure 29, along with the thermogram of the same module. The thermogram of module number 3, given in Figure 29 (right), is captured right after the thermographic analysis described in Section 3.2.4 and is given in Figure 20 (right) before disposal of the PV plant at the microlocation. The purpose of the thermogram given in Figure 29 (right) was not to conduct a thermographic analysis but rather to determine the thermal pattern for comparison with the damage classification via visual inspection. Comparison of the thermogram for module number 3 with the visual inspection damage classification results in Figure 29 shows that there is a significant difference. The thermal pattern does not correspond to the results of the visual inspection, i.e., conclusions that could be made on the basis of the visual inspection, whereby evaluation of the individual cell state via visual inspection emphasizes the complexity of the studied problem. The analysis performed on 62% of the total studied modules indicates that another faulty module may occur. Another conclusion is that, if reusing entire modules, it is necessary to monitor at least two modules for further development of potential hotspots. This is due to damage that can develop as the module ages during operation. The results of the visual inspection described in [13] are shown in Figure 28, which illustrate the different forms of damage found on the modules. A detailed overview of all the different module damage types can be found in the NREL report [23]. Interesting examples of damage and appearance of a PV module after 27 years of operation can be found in [22]. The analysis performed on 62% of the total studied modules indicates that another faulty module may occur. Another conclusion is that, if reusing entire modules, it is necessary to monitor at least two modules for further development of potential hotspots. This is due to damage that can develop as the module ages during operation. The results of the visual inspection described in [13] are shown in Figure 28, which illustrate the different forms of damage found on the modules. A detailed overview of all the different module damage types can be found in the NREL report [23]. Interesting examples of damage and appearance of a PV module after 27 years of operation can be found in [22]. A further analysis of the studied modules was done by classifying the visible damage into five levels. Level 1 represents a cell without any visible damage, level 2 represents barely visible damage, level 3 represents visible damage present in the right photograph of Figure 28, level 4 represents damage presented in the centre photograph in Figure 28, and level 5 is visible in the left photograph of Figure 28. The visual inspection and damage classification of module number 3 for each cell is given in Figure 29, along with the thermogram of the same module. The thermogram of module number 3, given in Figure 29 (right), is captured right after the thermographic analysis described in Section 3.2.4 and is given in Figure 20 (right) before disposal of the PV plant at the microlocation. The purpose of the thermogram given in Figure 29 (right) was not to conduct a thermographic analysis but rather to determine the thermal pattern for comparison with the damage classification via visual inspection. Comparison of the thermogram for module number 3 with the visual inspection damage classification results in Figure 29 shows that there is a significant difference. The thermal pattern does not correspond to the results of the visual inspection, i.e., conclusions that could be made on the basis of the visual inspection, whereby evaluation of the individual cell state via visual inspection emphasizes the complexity of the studied problem. A further analysis of the studied modules was done by classifying the visible damage into five levels. Level 1 represents a cell without any visible damage, level 2 represents barely visible damage, level 3 represents visible damage present in the right photograph of Figure 28, level 4 represents damage presented in the centre photograph in Figure 28, and level 5 is visible in the left photograph of Figure 28. The visual inspection and damage classification of module number 3 for each cell is given in Figure 29, along with the thermogram of the same module. The thermogram of module number 3, given in Figure 29 (right), is captured right after the thermographic analysis described in Section 3.2.4 and is given in Figure 20 (right) before disposal of the PV plant at the microlocation. The purpose of the thermogram given in Figure 29 (right) was not to conduct a thermographic analysis but rather to determine the thermal pattern for comparison with the damage classification via visual inspection. Comparison of the thermogram for module number 3 with the visual inspection damage classification results in Figure 29 shows that there is a significant difference. The thermal pattern does not correspond to the results of the visual inspection, i.e., conclusions that could be made on the basis of the visual inspection, whereby evaluation of the individual cell state via visual inspection emphasizes the complexity of the studied problem. Appl. Sci. 2022, 12, x FOR PEER REVIEW 21 of 24 In a review of PV module failure [21], in a section on documenting visual failures in the field, the observed patterns were classified as snail tracks. Snail tracks are discolorations of the silver paste used for the gridlines of cells. This discoloration appears along cell cracks. More detailed conclusions regarding the state of the remaining modules can be made only by applying electroluminescence in accordance with IEC TS 60904-13:2008 and detailed I-V curve analyses [30], which are also the subject of another study to be published in the near future.

Conclusions
PV power plants are one way to shift energy policy toward a green economy. Natural disasters disrupt the economic indicators of a PV power plant's life cycle. In the case of unforeseen events, such as hail, insurance plays a key role and covers the costs of replacing PV modules. The decision to replace all the modules or only the broken ones has a foundation, because even the modules that are visually intact can eventually develop a problem in operation due to mechanical shocks, and also manipulations when cleaning the cover of the surface covered in glass from other damaged modules. Using the example of the analyzed PV power plant, only 14% of the total number of modules remained visually intact. From the stated number, 13% was expected to malfunction. These problems cannot be directly related to the consequences of hail, but they can be used as a guideline if a module is reused for less demanding needs, i.e., at a location that allows intensive control at the time of commissioning.
The recommendations of IEC TS 62446-3 regarding the resolution of a thermal camera proved to be justified when it comes to choosing measuring points based on a smaller IFOV, which resulted in more accurate data. The effect of reflected radiation is particularly evident when a PV module is analyzed from the side of the glass cover. The temperature of the new module in the short-circuit state assumes temperature differences of up to 27 °C. The criterion for deciding the state of a module based on the ΔT criterion is not easy to determine, as it depends on the size of the module, short-circuit current, and the In a review of PV module failure [21], in a section on documenting visual failures in the field, the observed patterns were classified as snail tracks. Snail tracks are discolorations of the silver paste used for the gridlines of cells. This discoloration appears along cell cracks. More detailed conclusions regarding the state of the remaining modules can be made only by applying electroluminescence in accordance with IEC TS 60904-13:2008 and detailed I-V curve analyses [30], which are also the subject of another study to be published in the near future.

Conclusions
PV power plants are one way to shift energy policy toward a green economy. Natural disasters disrupt the economic indicators of a PV power plant's life cycle. In the case of unforeseen events, such as hail, insurance plays a key role and covers the costs of replacing PV modules. The decision to replace all the modules or only the broken ones has a foundation, because even the modules that are visually intact can eventually develop a problem in operation due to mechanical shocks, and also manipulations when cleaning the cover of the surface covered in glass from other damaged modules. Using the example of the analyzed PV power plant, only 14% of the total number of modules remained visually intact. From the stated number, 13% was expected to malfunction. These problems cannot be directly related to the consequences of hail, but they can be used as a guideline if a module is reused for less demanding needs, i.e., at a location that allows intensive control at the time of commissioning.
The recommendations of IEC TS 62446-3 regarding the resolution of a thermal camera proved to be justified when it comes to choosing measuring points based on a smaller IFOV, which resulted in more accurate data. The effect of reflected radiation is particularly evident when a PV module is analyzed from the side of the glass cover. The temperature of the new module in the short-circuit state assumes temperature differences of up to 27 • C. The criterion for deciding the state of a module based on the ∆T criterion is not easy to determine, as it depends on the size of the module, short-circuit current, and the amount of solar irradiance. The modules need to be cleaned prior to thermographic analyses. Cleaning a module can change the thermal pattern, depending on the degree of contamination, as removing sediment reduces shading. On average, one liter of water is needed to clean one module, and the procedure can take up to six minutes. Analysis of a single module, consisting of the determination of the I-V curve and the thermographic analysis of the module in short-circuit state, takes an average of 27 min. If we do not have a new or reference module for defining the ∆T criteria, the data of the mean values of the analysis of several modules can be used, and it is necessary to exclude protruding observations. Based on statistically defined criteria, it is possible to make a judgment about the current state of a module, as well as the allowable range of temperature differences. The observed thermal patterns cannot be assumed by identifying damage during visual inspection, which is best seen by comparing the conducted analysis of the third panel.
The comparison between the thermographic analysis and the visual inspection clearly confirmed thermography as a complementary method for testing PVs. In the case where there is a need for additional accuracy and precision of the thermographic pattern information, it is necessary to compensate for the influence of solar irradiance and ambient temperature fluctuations during the day. In most cases, this compensation is not necessary since the defects of the damaged module result in a significant deviation of the temperature in relation to the recorded parameters of an undamaged module.