Photovoltaic failure diagnosis using imaging techniques and electrical characterization

Abstract

Inspections of 48 photovoltaic (PV) modules within a 302.4 kWp solar array were undertaken to expose the presence of defects after 12 years of operation under the harsh environmental conditions of Djibouti. To this end, a multiple-technique testing protocol was conducted including visual inspection (VI), infrared thermography (IR), current-voltage curve characterization (I-V), ultraviolet fluorescence (UVFL) and electroluminescence imaging (EL). The main visible degradation features observed were discoloration, bubbling and snail trails with occurrences of 100%, 93.7% and 2.1% respectively. According to the IR imaging results, hotspots were observed on cells affected by snail trails. IR was combined with convolutional neural network (CNN) techniques to automatically detect the different classes of failures that PV modules may experience. EL imaging reveals that the cracks of the cells underlie the observed snail trails during visual inspection and UVFL imaging. In addition, a decrease in STC power was observed after 12 yr of operation with a median reaching 5.5% corresponding to an average degradation rate of 0.46%/years. Conclusively, fault diagnosis with combined approaches of imaging and electrical techniques is crucial to prevent defects and minimize the investment losses; this will ensure uninterrupted power generation, extended service life and high safety of photovoltaic modules.

Keywords

• photovoltaic fault diagnosis
• infrared imaging
• ultraviolet fluorescence
• electroluminescence
• current-voltage (i-v) curves
• convolutional neural network