Evaluation of Effect of Pre-Processing Techniques in Solar Panel Fault Detection
Sujata P. Pathak, Sonali Patil
Abstract
Solar energy is a clean and renewable source of energy produced by solar panels. Solar panels deteriorate over time, resulting in a generation of faults. Faults reduce the overall power generation capacity of photovoltaic (PV) plants. A variety of atmospheric and functional conditions contribute to the formation of hotspots on solar panels, indicating an increase in temperature and resulting in lower efficiency. Early detection of faults during PV module inspection and monitoring is critical for improving the efficiency, reliability, and safety of PV systems. Thermal imaging is a non-contact, non-destructive, efficient, and effective technique. With thermal image analysis, probable problem areas can be identified and fixed before actual failures or problems occur, resulting in lower costs and less human labor. Pre-processing is one of the steps in an automated fault detection system for removing noise or artefacts from thermal images. This study investigates the impact of pre-processing techniques such as filters and histogram equalization on fault detection and demarcation accuracy. Five different types of faults, such as single cell, multicell, diode, dust/shadow, and PID hotspot are detected. For fault detection, two segmentation techniques, histogram-based color thresholding and RGB color channel-based thresholding, are applied to thermal images of solar panels. Intersection over Union (IoU) is used to determine the efficiency of fault detection and demarcation techniques. Overall, images processed with a bilateral filter and histogram equalization performed better for fault detection and demarcation than other filters with IoU values ranging from 0.14 to 0.5.