PV_EL_HDR_BW_Test_DB

Published: 9 February 2026| Version 1 | DOI: 10.17632/pxyzzkbs4b.1
Contributors:
Marcello Polenghi, Vittorio Sala, Emanuele Giovanni Carlo Ogliari, Roberto Caldelli, Claudio Loconsole, Anna Valente, Ambra Vandone

Description

This dataset contains high-resolution (500 DPI) images and experimental measurements of photovoltaic cells power output degradation under 84 progressively induced different damage conditions. Each damage condition is described by a 3-images-set (Near-IR Electroluminescence, Near-IR Black/White, Visible HDR for a total of 252 images); for each damage level are also reported the corresponding performance measurements under controlled conditions and a detailed verbal description, encompassing both categorical and qualitative aspects. This dataset is intended for research on degradation modeling and lifetime prediction, fault detection and diagnosis, damage–performance relationship analysis with machine learning and computer vision applications as well as an LLM-RAG approach. Further images and measures are intended to be added in future versions. The dataset includes: - 84 Experimental 3-image-sets (total 252 images; each images-set is composed by 3 images: HDR, EL, BW) (PNG format) - Image files data (CSV) - Electrical performance measurements (voltage, current, power output) and verbal description of damage as identified through visual inspection (CSV) - Metadata describing cell type, naming conventions, experimental parameters (TXT) ## Usage Notes: The dataset is provided **as-is**. Users are responsible for validating and interpreting the data. ## Licensing and Attribution: This dataset is licensed under the **Creative Commons Attribution 4.0 International (CC BY 4.0)** license. ## Possible PV_EL_HDR_BW_Test_DB Dataset Applications This dataset, comprising damaged solar cell images alongside corresponding power output recordings can support a wide range of research activities across engineering, materials science, and data science. Key possible research uses include: Degradation modeling and lifetime prediction; Fault detection and diagnosis; Damage–performance relationship analysis; Machine learning and computer vision applications, also with LLM-RAG approach; Reliability and robustness assessment; Maintenance and mitigation strategy evaluation;Simulation and model validation; Energy yield and economic impact studies; Synthetic data generation. Overall, this dataset is valuable for both fundamental research on degradation mechanisms and applied research aimed at improving monitoring, reliability, and cost-effectiveness of photovoltaic systems.

Files

Steps to reproduce

## Solar Cells Specifications |Parameter |Description |Units / Notes | Type | Panel technology | Polycrystalline | | Nominal Power | Rated power | 4.5 Wp (Watts peak) | | Nominal Voltage| | 5V | | Nominal Current| | 0.9A | | Dimensions | Physical size | 130mm × 150mm | | Manufacturer | Panel manufacturer | DAUERHAFT | | Model | PV cell model code | AP130X150MM | ## Environmental Conditions | Ambient Temperature | 12 – 20 °C | | Relative Humidity | 50 – 70 % | | Test Lamp Irradiance Power (extimated) | 50 W/m² | ## Measurement Protocols | Parameter | Description| Units / Notes| | Measurement Interval| Frequency of measurements | 5 minutes between each of the 3-image-set | | Duration| Experiment duration| 80 hours; 60 days timespan| | Environmental Conditions Monitoring| Ambient temperature/humidity | domestic thermometer/web data| | Power Calculation| P = V × I| Watts| | Electric Damage Detection| EL imaging| visual assessment | | Physical Damage Detection| HDR imaging| visual assessment | | Soiling Damage Detection| BW imaging|visual assessment | | Defect Inspection| Microcracks, delamination etc.| Categorical, visual inspection | ## Experimental Setup | Device | Model / Notes | | HDR Camera | LUCID Triton HDR5.4 | | Visible/NearIR Camera | IDS u3-36p0xcp | | BP800 Filter | MIDOPT BP 800-25.4 (FilBP800/C-mount #24495) | | Test Alogen Lamp | OSRAM DECOSTAR 51 ECO SUPERSTAR 47860 ECO SST 36°-cone Ra=100 1000cd 12V 25W 2900K; working distance 40cm; angles: X=10°, Y=15° | | Power Supply | KORAD KD3005D DIGITAL-CONTROL DC POWER SUPPLY 0-30V 0-5A | | MultiMeter | Metex Corporation M-3660D digital multimeter | HDR Camera Details (for HDR) CAMERA: LUCID Triton HDR5.4 Bin: 2x2 Exposure: 80e-6 sec Gain: 5 analogic Acquisition Software: ArenaView Lens: RICOH FL-BC2518-9M (F5.6/25mm/1.8”) Work distance: 62cm (manually measured) Angles: X=5° ; Y=1° (extimated geometrically form setup ortogonal imagery) CALIBRATION: CALIBRATION PLATE and HALCON software Visible/NearIR Camera (for EL and BW) CAMERA: IDS u3-36p0xcp Bin: 2x2 Exposure: 1 sec Gain: 5 analogic Acquisition Software: IDSpic Lens: OPT-CDP-1628-V10 (F2.8/16mm/1.1”/C sn C252080332) Filter: MIDOPT BP 800-25.4 (FilBP800/C-mount #24495) Work distance: 41.5cm (manually measured) Angles: X=-7° ; Y=2° (extimated geometrically form setup ortogonal imagery) CALIBRATION: CALIBRATION PLATE and HALCON software ## Calibration and Accuracy - Camera calibration by calibration plate and Halcon sw procedure - Power Supply Voltage accuracy: ±0.5 % + 20 mV - Power Supply Current accuracy: ±0.5 % + 10 mA - Multimeter Voltage accuracy: ±0.3 % of reading + 1 digit - Multimeter Current accuracy: ±0.5 % of reading + 1 digit - Temperature sensors: unknown °C ## Power Input Values V=6,8 [V] (EL is amplified and more clearly observable with V>5 V) A= variable 0,7-0,9 [A] (current<0.9[A] in order to limit image white saturation)

Institutions

Categories

Electric Power, Photovoltaics, Machine Learning, Electroluminescence, Ground Based PV Systems, Photovoltaic Performance, Photovoltaic Modules, Infrared Imaging, Database, Damage Classification, Large Language Model, Retrieval-Augmented LLM

Licence