Opto electronic system for real time health evaluation of photovoltaic panels

Abstract

In this study, a novel optoelectronic system for fault detection in photovoltaic (PV) cells has been developed. Three sensors, each with a photodiode, were manufactured and mathematical models developed to interpret the fault results from the sensors. The photodiodes sweep across the PV panel to identify areas of high light intensity. The goal is to produce diagnostic images of PV panels that are comparable to standard electroluminescence (EL) imaging. Each sensor was tested under two conditions: darkness and sunlight exposure. For all the sensors, the results obtained in darkness closely match the EL images. However, PV panel exposure to sunlight produces mixed results due to differences in light intensity across the PV cells. To address this issue, two enhancement techniques were developed. First, a collector was used to improve sunlight directionality, with an improved result shown in Sensor 3. Second, a voltage step was added to the PV panel, showing an improved result in all three sensors. Among the tested combinations, the combination of Sensor 3 with an alternative collector and a step-type voltage source produced the best performance. These results clearly indicate that the sensor-based approach can effectively diagnose the PV panel health condition.

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-025-20849-2.

Introduction

Currently there has been a rise in the installation of non-conventional or renewable energy sources^1–7. This is due to an increase in both global population as well as energy usage. Energy consumption and environmental problems can be considerably decreased by substituting these fossil fuels with these sources. Solar energy has garnered significant attention due to its vast potential, especially in countries with extensive clear sky regions, such as deserts^5,7. Solar photovoltaic (PV) technology is advancing rapidly, with many countries integrating solar energy into their future energy strategies. However, to maintain optimal performance and meet life cycle operating requirements, PV facilities require regular maintenance. This includes not only cleaning the solar panels but also assessing their current operational conditions to ensure they produce the expected amount of electricity over their lifespan^8,9.

To evaluate the current state of PV panels, different diagnostic tools are being used: such as thermography^8,10–14, studies of current-voltage (I-V) curves^8,15–18, visual inspection^19,20, signal transmission technique⁸, ultraviolet (UV) fluorescence^21–23, and electroluminescence (EL)^24–29. Thermography detects hot spots in solar panels, which can be identified using specialized cameras. Failures, such as corroded circuits or defective solder joints, often cause localized temperature increases, making them easier to spot with this technology. Currently, thermography has expanded to include the use of drones, allowing for more efficient and rapid identification of such defects¹¹. The limitation of thermography is that it does not provide a quantitative assessment of how the detected anomaly affects the performance of the PV module. In contrast, I-V curves offer a detailed representation of the operational characteristics of each solar panel, establishing a baseline for comparison with future measurements. A key aspect of this system is the system curve, obtained by placing the solar panel on a test bench equipped with an artificial light source, controlled temperature, and a data acquisition system, while adjusting the module’s voltage or current using external systems. In field conditions, applying this approach according to the IEC 60904-3 standard is not feasible. However, if a pyranometer is used to measure irradiance, the collected data must be corrected according to the standard¹⁷. Nonetheless, it should be noted that using I-V characteristics requires turning off the PV generator in order to take measurements of the I-V characteristic, limiting the applicability of this approach to online monitoring. The visual inspection approach, specified in the IEC61215-1 and IEC61730-2 standards for new modules, cannot be applied to modules that have been degraded by usage. Strategies have been developed to standardize the inspection of modules that have been damaged, with the inspection being divided into components and types of PV panels. In addition, the primary defects are established for each type of panel, creating a baseline for the search of problems and improving the process. This is complemented by documentation guidelines that should be managed to comply with the notification of damage. One of the major drawbacks of the visual inspection method is that it does not work well with weathered modules. The UV fluorescence (UVF) is another technique used for PV fault detection. While UVF imaging with a camera provides information on the luminescence intensity and thus the fluorophore density, UVF spectroscopy determines the type of fluorophores present by analyzing the emitted spectrum at a specific location on the module. As a result, it allows inferences about the cell temperature history, for example, since greater temperatures result in additional peaks in the observed spectrum. This technique also has its drawbacks including that the recorded UVF signal can be modified by a variety of factors, such as module position, time of operation, actual temperature as well as temperature history, experienced doses of heat, humidity, and UV radiation.

In the EL diagnostic technique for PV panels, a voltage is applied to the solar panels, inducing a current that generates light in the infrared or near-infrared region of the spectrum. This process results in EL emission within the mid-infrared spectrum, typically around 1150 nm. A cooled charge-coupled device (CCD) camera is employed to capture the emitted light from the energized PV cell. Photodetectors sensitive to this specific wavelength range then convert the captured light into data that can be interpreted by the operator^24,25,30. The International Electrotechnical Commission (IEC) currently has only draft versions of international standards for the quantitative interpretation of PV cells using the EL technique. The IEC technical specification outlines procedures for capturing EL images of PV cells, processing these images to derive quantitative descriptors, and performing qualitative evaluations of the results²⁵. This standard applies to PV modules tested in a forward bias condition, i.e., by forcing current flow with a power supply whose leads are connected to that of the PV module of the same polarity. EL photography can also be done using specialized tripods or drones. However, the standard approach is to utilize cameras that are sensitive to the spectrum, as well as light filters to reduce the bandwidth of the light spectrum being captured²⁵. In the EL photography, defects are displayed as dark regions. Since EL imaging is not affected by blurring caused by lateral heat propagation, it can be used at high resolutions. Nevertheless, the analysis of EL images is typically a manual, costly, and time-consuming process that requires expert knowledge. CCD or complementary metal oxide semiconductors (CMOS) are frequently used as camera detectors. They can be cooled, usually through thermoelectric cooling, to lower device current originating from thermally generated charges and improve the signal to noise ratio. The number of pixels, noise, quantum efficiency at the desired wavelength, and dynamic range are important factors when selecting detectors. 

However, the EL technique faces challenges when applied in industrial settings. For accurate measurements, it is essential that the camera captures only the emission from the panel, minimizing interference from external light sources. Solar radiation and other ambient light can disrupt the measurements, making it crucial to isolate the panel’s emission from these extraneous sources³¹. A number of strategies are adopted to attain this goal, such as operating the PV module at night or constructing a shadow around the solar panel in question²⁵. All of the aforementioned techniques can negatively impact the solar panel’s continuous operation, as they necessitate halting the panel’s function during the assessment. To mitigate this issue, one approach is to conduct inspections and diagnostics during non-productive hours when the solar panel is not generating power²⁵. These circumstances provide an opportunity to explore alternative methods that complement existing approaches or serve as viable options when conventional techniques fall short.

This study introduces a novel EL analysis technique for PV modules using a photo-sensor. The photodiode performs a comprehensive sweep of the solar panel to collect data on regions of high light intensity, enabling the identification of active and inactive areas within the PV module. This method also facilitates the creation of a graphical representation of the emitted luminous intensity, closely resembling the images produced by traditional EL techniques. One advantage of this new technique is its ability to take measurements under conditions where traditional cameras may not yield reliable results. To assess its effectiveness, several photodiodes were characterized and evaluated for their suitability in this application. Additionally, various light collectors are employed to control the directionality of incident light on the sensor. The impact of applying a voltage step to the solar panel is examined using data from the photodiode. The primary aim of this work is to develop an optoelectronic monitoring system to assess the condition of PV cells according to the “IEC TS 60904-13:2018” standard. A secondary objective is to measure EL both under direct sunlight and in darkness. This is crucial for determining the acceptable ranges of luminosity variation as specified by the standard, based on the measurements conducted in this study.

This paper is organized as follows: Sect. 1 presents the introduction. Section 2 describes the methodology, and Sect. 3 presents the results and discussion.

Methodology

In this study, four distinct tests were conducted: one in darkness, one under sunlight exposure, one with an alternative collector, and one with an intermittent light source. Table 1 outlines the conditions for each of these tests.

Table 1.

Different tests carried out and their measurement conditions.

Test	Sunlight presence	DC Source	Collector
Darkness	No	Continuous	Original
Sunlight	Yes	Continuous	Original
Alternative collector	Yes	Continuous	Alternative
Intermittent source	Yes	Step	Original

Design of the test circuits

Six interchangeable LED boards were fabricated, each carrying nine identical through-hole LEDs (Young Sun LED Technology, 5 mm epoxy package). The boards were strictly monochromatic (red (≈ 625 nm), blue (≈ 470 nm), yellow (≈ 590 nm), white (broad-spectrum phosphor, CCT ≈ 6000 K), ultraviolet (≈ 395 nm) and infrared (≈ 850 nm)) so that no mixed-color illumination was introduced. During a given experiment only one board (i.e., one color) was operated, and the optical intensity was modulated solely by adjusting the BAKU BK-1502D + supply voltage.

Figure 1a illustrates the test bench setup used to characterize photodiodes and assess their suitability for detecting PV defects. The test bench features LEDs on one side and a measurement circuit, which employs photodiodes as the sensing element, on the other. The primary objective is to construct a sealed system to prevent light ingress and thereby minimize measurement interference. The system includes a light emitter, depicted on the left side of Fig. 1a and the gray section of Fig. 1b, consisting of nine LEDs of the same color. This LED lighting system is powered by a BAKU brand model BK-1502D + variable DC voltage source, which can supply a DC voltage of up to 15 V. The intensity of the light emitted by the system is proportional to the voltage of the source.

Fig. 1.

(a) Schematic of the sensor test bench (LED light source on the left, photodiode measurement circuit on the right); (b) Photograph of the test bench setup; (c) Detail of the sensor’s amplification circuit.

The sensor consists of an amplifier and a measurement circuit, featuring a socket and a mounting plate designed to hold the photodiodes. The circuit plate is shown in Fig. 1c. The sensor is powered by two DC voltage sources, namely 12 V and 9 V. The 12 V supply powers the amplification circuit, while the 9 V powers the measuring circuit, aiming to decrease measurement noise. A UNI-T brand oscilloscope model UT81B was utilized to measure voltage.

Characterization of the photodiodes used as sensors

In this study, three sensors were developed. Sensor 1 is a photodiode (SD003-151-001) with a sensitivity spectrum extending into the near-infrared (NIR) region. This characteristic makes it particularly suitable for use with InGaAs PV panels. Although the test panels used in this study are not made from InGaAs, a key objective is to assess the sensor’s performance with various PV panels and evaluate the impact of sunlight interference. Sensor 2 is a photodiode (VTP9812FH-ND) that, while not optimal for this application, offers a broader wavelength sensitivity and low power consumption. Sensor 3 is a photodiode (BP-104FS) with high sensitivity to the infrared spectrum, making it well-suited for the sensor development needs. However, due to its limited sensitivity to the blue spectrum, careful attention must be paid to avoid component saturation.

The characterization of the photodiodes was conducted under the following conditions and procedures. All measurements were performed at night to eliminate interference from sunlight. The test began at 0 V, with voltage increments of 0.5 V up to a maximum of 15 V. The voltage applied to the sensor was recorded, and measurements were repeated every 5 min for each photodiode. Additionally, each LED color was tested three times, and the average values were calculated. The averaged results for each LED color were then graphed separately for each photodiode.

To minimize stray light, the LED array and the photodiode under test were placed inside a light-tight black enclosure. The sensor PCB was fixed at a distance of 270 mm from the centre of the 3 × 3 LED array to ensure uniform illumination. The photodiode/amplifier stage was powered from two isolated DC rails (12 V for the trans-impedance amplifier and 9 V for the photodiode), while the LED array was driven by the BK-1502D + supply used for the voltage sweep. Output waveforms were monitored with a UNI-T UT81B digital oscilloscope and logged to a PC for off-line averaging.

Design of the sensor and collector

The sensor incorporated an amplifier to integrate the signal with the measuring instrument, specifically a UNI-T UT81B oscilloscope. The enclosure was designed for ease of handling during measurements. Initial tests revealed that the sensors were not fully opaque to all light spectra. To prevent interference from external light sources, the sensors were covered with aluminum foil and Kapton tape, as illustrated in the Appendix (Figure A.1). Additionally, the collector is mounted on the front opening, as shown in Figure A.1.

The opening that permits light entry into the enclosure was fitted with a collector made from flexible filament. To enhance opacity and control the directionality of incoming sunlight, two designs were developed to mitigate the effects of direct sunlight during measurements. The first design is the original collector, depicted in the Appendix (Figures A.2a and A.2b), while the second is an alternative collector shown in the Appendix (Figures A.2c and A.2d). The alternative collector features a slotted design to provide improved directionality.

Experimental procedure

To effectively use the sensors in this investigation, a measurement strategy was developed to gather sufficient data for a comprehensive diagnostic of the PV panel. The experimental procedure involves dividing the solar panel into 36 distinct sections. Each cell within these sections is scanned using the sensors. Following the scanning process, the solar panel is disconnected from the power supply for 30 min to ensure accurate measurements. This process is repeated until the desired number of measurements is obtained.

Measurements taken in darkness and under direct exposure to sunlight

In darkness, the sensor scans the PV panel and records the light intensity for each cell. This scanning process is repeated three times to determine the minimum intensity for each cell, as calculated using Eq. 1. The resulting data forms the initial matrix.

1

Where:

Where i is the row index and j is the column index of the PV cells. Then, using Eq. 2, a second matrix is produced in which the luminescence value of each cell is subtracted from the minimum luminance value derived from the first matrix.

2

In order to reduce the error, the mean light intensity for each cell was obtained, as described in the Eq. 3.

3

A gradient with three distinct points was created to visually match the electroluminescence (EL) photograph. The center point represents the average light intensity across all sections of the panel. The other two points correspond to the minimum and maximum light intensity values. Colors were selected as follows: purple for the lowest intensity, mauve for the average intensity, and magenta for the highest intensity. This color scheme was chosen to produce results visually comparable to the EL photograph. To ensure accurate measurements in darkness, the light sources were covered with aluminum foil to prevent interference. Measurements under sunlight were taken between 12:00 and 16:00 during the summer months to minimize interference and obtain the most reliable results. In the Appendix, Figure A.3 shows a graphical representation of an EL test conducted in darkness using Sensor 1 on a specific panel. The EL values displayed in the matrix indicate the light intensity measured by the sensor and reflect the difference between the panel’s active (on) and inactive (off) states. These values are influenced by the light intensity at the time of measurement.

Measurement taken using an alternative collector

To minimize errors caused by the shifting position of the sun, data were first collected using the original collector and subsequently with an alternative collector. This time-dependent error is a constant factor affecting the measurements. Because the collector’s effectiveness depends on the geometry of incoming sunlight, the measurement system must be robust enough to mitigate this issue. To ensure results comparable to the EL photograph, measurements were repeated both in darkness and under sunlight.

Measurement taken using an intermittent source

Two distinct datasets for light intensity were collected from the cells of the PV panel: one with the power supply turned on and another with it turned off. The matrices for these two datasets were then subtracted using the following Eq. (4).

4

Where:

matrix obtained from the light intensity values of the solar panel energized.

matrix obtained from the light intensity values of the solar panel not energized.

To obtain a result for the measurement, the previous three-point gradient system was adopted.

Results and discussion

This section presents the results of all measurements taken with the three sensors under the four specified conditions. To ensure robustness, five PV panels were tested; however, only the results from the first PV panel are included here. The results from the remaining panels are provided in the appendix. Figure 2 shows the functional block diagram of the optoelectronic system, illustrating the main subsystems: (i) BK-1502D + programmable DC source, (ii) photovoltaic panel, (iii) DC power source (iv) sensor housing with collector, photodiode, and TIA amplifier stage, (v) data acquisition unit/UNI-T UT81B oscilloscope, and (vi) computer for data processing.

Fig. 2.

Functional block diagram of the optoelectronic system.

Measurement protocol

There are two protocols for measurements, the first method is applied when measuring without the step in the power source.

The first diagnostic procedure follows the five-step sequence below:

• Segmentation: The PV module is divided into 36 cells arranged in a 4 × 9 matrix.

• Powering: The DC voltage is applied to the solar cells.

• Scanning: The sensor housing is placed sequentially over each cell (i, j); the photodiode output is sampled for 5 s and averaged.

• Data logging: The averaged value is recorded at the corresponding position M(i, j) in a 4 × 9 data matrix.

• Condition loop: Steps 1–3 are repeated under the four test conditions listed in Table 1, yielding one matrix per condition.

The second diagnostic procedure is applied when using the step power source, with the protocol as follows:

• Segmentation: The PV module is divided into 36 cells.

• Scanning: The sensor housing is placed sequentially over each cell (i, j); the photodiode output is sampled for 5 s.

• Step up source: The DC voltage is applied to the solar cells.

• Scanning: The sensor housing is placed sequentially over each cell (i, j); the photodiode output is sampled for 5 s.

• Data logging: The difference between the first and second value is recorded at the corresponding position M(i, j) in a 4 × 9 data matrix.

• Condition loop: Steps 1–3 are repeated under the four test conditions listed in Table 1, yielding one matrix per condition.

These matrices are subsequently compared with the EL photograph to assess the spatial correlation between sensor data and standard EL imaging.

Results of measurement taken in darkness

Figure 3 presents the light intensity measurement results for PV panel 1 under darkness. Sensor 1 (Fig. 3a) demonstrates the lowest responsiveness to the light spectrum among the sensors, showing minimal variation between maximum and minimum light intensity values. When compared to the EL photographs in Fig. 3d, the results for Sensor 1 are notably similar, indicating a successful test. The small offset observed in the sensor’s output does not affect the results, indicating that the amplification circuit’s stability is adequate.

Fig. 3.

(a–c) graphical representation of the measurements obtained from the EL test on PV panel 1, in darkness, (d) results obtained from the EL photograph for comparison, each area is a size of 35 × 55[mm].

The results indicate that Sensor 2 (Fig. 3b) outperforms Sensor 1 (Fig. 3a), as evidenced by its greater variability in light intensity measurements. This improved performance is attributed to the higher quality of the photodiode used, which is more commonly employed in practice. Sensor 3 (Fig. 3c) demonstrates the broadest range of light intensity values and exhibits the closest correlation with the EL photographs. Due to its high contrast, Sensor 3 allows for the implementation of complementary techniques, which could potentially reduce light intensity measurements. Additional results from testing the three sensors on four different PV panels are provided in Appendix A. More technical specifications and selection criteria for the sensors is given below:

• Sensor 1 (SD003-151-001): It is a surface-mount InGaAs photodiode with an approximate spectral range from 800 nm to 1700 nm. The wide infrared range that particularly includes the NIR spectrum makes it especially suitable for capturing electroluminescence emissions from InGaAs photovoltaic cells. It has been indicated previously that this device was specifically selected for its high sensitivity in the NIR, considering potential future diagnostic applications for InGaAs panels.

• Sensor 2 (VTP9812FH-nd): It is an “ambient light” silicon photodiode with an IR-blocking filter. Its peak sensitivity is in the visible region (~ 580 nm), covering an approximate spectral range from 400 nm to 700 nm. This sensor features a very high shunt resistance, low capacitance, and minimal dark current, resulting in low noise and low power consumption. It has been explained previously that this photodiode was chosen to represent a broader-spectrum sensor in the visible range and to serve as a comparative reference, although it is not optimal for detecting the infrared electroluminescence of solar cells. Its inclusion allowed us to verify the system’s performance with a wide-spectrum visible sensor and to confirm the importance of infrared sensitivity for this type of diagnostic.

• Sensor 3 (BP-104FS): This sensor is a silicon PIN photodiode equipped with a daylight filter, highly sensitive in the near-infrared range (approximately 780 nm to 1100 nm). We have added that this sensor has a peak response around 950 nm with a typical high sensitivity (~ 0.7 A/W at 950 nm) and an acceptance angle of approximately 60°. Its integrated daylight filter effectively rejects most visible light (particularly blue wavelengths, to which the sensor is especially sensitive), enhancing infrared detection. This sensor was chosen due to its high infrared sensitivity, making it ideal for capturing electroluminescence from silicon panels while minimizing interference from ambient visible light. Due to its limited blue sensitivity, precautions were taken to avoid saturating the photodiode with luminous intensities outside its range (although, in practice, its filter prevents most visible light from affecting it).

Overall, all sensors provided satisfactory results across various sections of the PV panel, closely resembling the images obtained using the standard EL approach. However, high-intensity areas do not necessarily indicate faults in the PV panel. When panel conduction is disrupted, the remaining material may experience increased current density, leading to bright spots that could be mistakenly identified as hot spots during defect analysis. Similarly, dark regions resulting from complete cell disconnections can cause adjacent areas to exhibit increased intensity. Therefore, examining both low-intensity areas and their surroundings is crucial, as these observations can help in making preliminary diagnoses and identifying potential issues within the panel.

Results of measurement taken under sunlight

Figure 4a–c shows the light intensity measurements on PV panel 1 under sunlight exposure. Compared to the results in Fig. 3 there is a noticeable loss of similarity to the EL photographs (Fig. 4d). Additional results from testing the three sensors on four different PV panels are presented in Appendix B. Significant color discrepancies are observed, particularly in rows 1 and 9 of panels 2, 3, and 5, with complete mismatches in panel 4.

Fig. 4.

(a–c) Graphical representation of the measurements obtained from the EL test on PV panel 1 conducted under the exposure to sunlight, (d) results obtained from the EL photograph for comparison, each area is a size of 35 × 55[mm].

The results from Sensor 1 (Fig. 4a) and Sensor 2 (Fig. 4b) exhibit a stark contrast with the EL photograph, particularly in panels 3 and 5 (Appendix B). This discrepancy is especially evident in rows 1 and 9, as well as column A. Sensor 3 (Fig. 4c) shows similar results to Sensors 1 and 2, with notable discrepancies in rows 1 and 9 and column A of PV panel 1. These findings indicate that sunlight exposure significantly distorts the light intensity values across different areas of the PV panel, making it challenging to compare with EL photographs. Additionally, the errors are concentrated mainly along the edge sections of the solar panel, likely due to the angle of sunlight incidence.

Each sensor exhibits distinct properties in capturing light intensity. Sensor 1 does not fully align with the EL photographs due to its narrower sensitivity range to the various spectra of natural sunlight. Sensor 2, while more sensitive to the different spectrum of sunlight, produces lower-quality results compared to Sensor 1. Sensor 3, with its adequate sensitivity to the infrared spectrum, is not narrow enough to filter out sunlight effectively. As a result, Sensor 3 shows the greatest variation in light intensity, which negatively impacts the results. The study anticipated the influence of sunlight on the sensors and proposes two mitigation measures, the effects of which are discussed in the following results.

Sensor 1, being an InGaAs photodiode, is mainly sensitive to the longer-wavelength infrared range (approximately > 800 nm) and shows virtually no response to the visible spectrum. Under sunlight conditions, this means it captures part of the panel’s infrared electroluminescence emission (the useful signal) but also a significant portion of the incident solar infrared radiation (background IR). This dual contribution (useful signal + solar IR background) raises the baseline intensity, reducing similarity with the reference EL image when the panel is under sunlight. However, its lack of visible sensitivity prevents it from detecting shorter-wavelength solar reflections or scattered visible light, meaning that the discrepancy stems primarily from the reduced intensity response. This reduction in contrast between the maximum and minimum values makes the measurements more susceptible to the limitations of the instrumentation.

Sensor 2, by contrast, is more sensitive to the solar visible spectrum (approximately 400–700 nm, with peak sensitivity around 580 nm) due to its IR-blocking filter. Therefore, under solar illumination, this sensor primarily registers solar visible light reflected from the panel cells, capturing almost none of the infrared electroluminescence emission (which falls outside its spectral range). Consequently, the measurements obtained in sunlight conditions deviate significantly from the EL image, as the sensor essentially measures solar illumination variations across the panel rather than the cells’ infrared emission. In practice, Sensor 2 exhibited the greatest loss of similarity to the EL photograph under sunlight due to this spectral misalignment with the electroluminescence emission.

Sensor 3 exhibits an intermediate behavior: its daylight filter allows it to ignore most of the visible (particularly blue) light, focusing instead on the near-infrared spectrum (780–1100 nm). Thus, under sunlight, Sensor 3 is not significantly affected by the visible component of solar irradiance but still captures a considerable portion of solar infrared radiation (extending up to approximately 1100 nm). Moreover, we note that the spectral range of Sensor 3 does not fully encompass the electroluminescence wavelength of silicon cells (1050 ~ 1150 nm), falling slightly short of this peak. Consequently, under solar conditions, Sensor 3 captures variable IR intensities from sunlight through its slots and may lose part of the effective EL signal, thus increasing the variability of its measurements and explaining why it initially showed the greatest dispersion and uncertainty in daytime results.

In summary, it has been clarified that each sensor responds differently to sunlight due to its spectral band: infrared-filtered sensors (Sensors 1 and 3) avoid visible-light noise but still suffer from solar IR interference, whereas the visible-spectrum-oriented sensor (Sensor 2) essentially measures reflected solar irradiance rather than electroluminescence emission.

Results of measurement with the alternative collector

Two separate measurements were conducted: one at 11:00 and another at 17:00. To facilitate a clearer interpretation of the results, measurements using an alternative collector were compared with those obtained using the original collector. Figure 5 presents the light intensity on PV panel 1 using the alternative collector at 11:00 AM, while Fig. 6 shows the intensity with the original collector at the same time. Figures 7 and 8 display similar measurements taken at 17:00 for both collectors.

Fig. 5.

(a–c) Graphical representation obtained after processing the measurements from the EL test, taken under condition of direct sunlight, with the alternative collector at 11:00 in the panel 1, (d) results obtained from the EL photograph for comparison, each area is a size of 35 × 55[mm].

Fig. 6.

(a–c): Graphical representation obtained after processing the measurements from the EL test, taken under condition of direct sunlight, with the original collector at 11:00 in the panel 1, (d) results obtained from the EL photograph for comparison, each area is a size of 35 × 55 [mm].

Fig. 7.

(a–c): Graphical representation obtained after processing the measurements from the EL test, taken under condition of direct sunlight, with the alternative collector at 17:00 in the PV panel 1, (d) results obtained from the EL photograph for comparison, each area is a size of 35 × 55 [mm].

Fig. 8.

(a–c) Graphic representation obtained after processing the measurements from the EL test, taken under condition of direct sunlight, with the original collector at 17:00 in the panel 1, (d) results obtained from the EL photograph for comparison, each area is a size of 35 × 55[mm].

To assess the effectiveness of the new collector in reducing interference from direct sunlight, we utilize two main tools: the intensity measurements depicted in the figures and the color variations of each cell. Comparing these measurements with the positions of the cells in relation to the EL photographs provides a comprehensive interpretation. By examining both the positioning and changes in intensity of adjacent cells, we can effectively evaluate the impact of the new collector.

The results indicate that for Sensor 1, the new collector did not improve performance compared to the original collector. In most cases, there was a noticeable reduction in light intensity with the new collector, particularly around cells C-5, C-6, and C-7, where high-intensity areas were visible in the measurements but not in the original photograph. The data obtained with the original collector did not exhibit this issue. Sensor 2 showed similar results to Sensor 1, with some improvement observed particularly in row 9. However, the alternative collector resulted in a loss of definition in regions of medium intensity, thus failing to reduce errors effectively and proving less suitable for the proposed EL test. Sensor 3 demonstrated overall improvement at 11:00 with the new collector. Although some discrepancies were noted at 17:00 likely due to reduced sunlight intensity, the alternative collector still provided better results compared to the original collector.

A practical way to judge the collector’s effectiveness is to inspect how the measured intensity of each cell compares with that of its immediate neighbors. Stray sunlight typically produces abrupt jumps at the panel edges (rows 1 and 9, column A) and local hot spots in the central region (e.g., cells C-6 and C-7). With the slotted collector, these artefacts were largely suppressed. For instance, on PV panel 1, the high-intensity patches seen at 11:00 h with the original collector (Fig. 6c) disappeared when the slotted collector was used (Fig. 5c), yielding a color distribution that matches the EL photograph much more closely. The improvement remained evident in the 17:00 h test, where the overall color scale became smoother and no out-of-range spikes persisted (compare Fig. 8c with Fig. 7c). Although Sensors 1 and 2 showed only marginal gains (attributable to their lower amplitude response), the qualitative analysis of neighboring-cell intensity confirms that the slotted collector significantly enhances the spatial fidelity of Sensor 3 under direct sunlight.

Overall, the sensor results vary primarily due to the collector’s impact on the EL intensity reaching the photodiodes. Sensor 1 experienced a significant reduction in EL intensity because its photodiode has a lower amplitude response to the light emitted by the solar cells. Sensor 2 showed mixed results, with some improvements but still inconsistencies. Sensor 3, however, demonstrated substantial improvement. These observations suggest that while the new collector helps, it alone is insufficient to fully mitigate sunlight interference issues. The effectiveness of the collector is largely dependent on the photodiode’s amplitude response to the light intensity from the solar cells.

Results of measurement taken with intermittent power source and the original collector

Figure 9 presents the results of the EL test conducted using an intermittent power source with the original collector on Panel 1. Results for PV panels 2, 3, 4, and 5 are detailed in Appendix C.

Fig. 9.

(a–c) Graphical representation of the measurements from the EL test, taken under condition of intermittent power source for the PV panel 1, with the original collector, (d) results obtained from the EL photograph for comparison, each area is a size of 35 × 55[mm].

The data in Fig. 9 indicate that the intermittent power source has enhanced the performance of Sensor 1 (Fig. 9a) compared to the results obtained with the original collector. The improved results with Sensor 1 are now more consistent with those observed under darkness (Sect. 3.1), preserving the distinction between high and low intensity points. However, variations in medium intensity points remain, indicating that while the overall quality has improved, there are still fluctuations in the medium intensity readings. This is regarded as having less resolution than other sensors, especially Sensor 3. The results from sensor 2 (Fig. 9b) are also comparable to those obtained in darkness, although in comparison with sensor 1 there is more variation of intensity, which permits a higher resolution in the color scale, making the results better when compared with the EL photography. Sensor 3 (Fig. 9c) shows an improvement in the results that is quite clear in all observed cases, with the exception of some points brighter than expected. This might be due to human error during the measurement process.

The intermittent-mode test consisted of two consecutive acquisitions under identical ambient light: (i) PV module forward-biased for 5 s and (ii) module open-circuited for 5 s. Subtracting the ‘OFF’ matrix from the ‘ON’ matrix isolates the electroluminescence signal and rejects the constant solar background.

Overall, employing an intermittent power source enhances the performance of the sensors, though the degree of improvement varies based on the sensitivity and spectral response of the photodiodes. Sensor 1 shows the least benefit from this technique compared to the other sensors.

Conclusion

Based on the results, it can be concluded that the developed sensor is a viable alternative to conventional methods for EL testing of PV panels, especially under the conditions outlined in the study objectives. This is particularly evident from the tests conducted in darkness.

The EL tests conducted in darkness demonstrated that the quality of results from the various sensors was comparable to that of conventional techniques. Despite significant variations in light intensity among the sensors, mathematical techniques were employed to process their data effectively. However, under direct sunlight, the sensors exhibited mixed performance. Sunlight introduced areas of high light intensity in some parts of the PV cells, while causing others to appear opaque. To improve performance under sunlight, two strategies were employed. The first involved using a new collector designed to enhance sunlight directionality, which yielded better results only for Sensor 3. The second approach involved applying a voltage step to the PV panel, which improved results across all sensors.

It can be concluded that the sensor-based approach to PV fault detection proposed in this study represents a valuable method for assessing the operational condition of PV panels. While it may not diagnose specific issues, it effectively provides insights into the overall health and functionality of the panels. Among the various strategies and sensors evaluated, the optimal configuration involves using Sensor 3 in conjunction with an alternative collector and a step-type power source. Notably, tests conducted with the step-type source revealed bright areas not observed during darkness tests or in EL photographs. These anomalies may indicate faults in the PV cells resulting from sudden changes in the power supply.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

Abdullahi Abubakar Mas’ud, Ignacio Cuadra, Jorge Ardila-Rey: Conceptualization, Methodology, Software, Visualization, Investigation, Writing- Original draft preparation. Abdullahi Abubakar Mas’ud, Rodrigo Barraza, Antonio Sanchez, Oscar Danilo Montoya. Data curation, Validation, Supervision, Resources, Writing - Review & Editing. Abdullahi Abubakar Mas’ud, Jorge Ardila-Rey, Ignacio Cuadra, Hassan Z. AlGarni: Project administration, Supervision, Resources, Writing - Review & Editing.

Data availability

The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.

Declarations

Competing interests

The authors declare no competing interests.