Fault classification using deep learning based model and impact of dust accumulation on solar photovoltaic modules

Abstract

ABSTRACT Solar panel performance is affected by ambient temperature, sunlight, module surface temperature, dust, and shadows. Dust inhibits sunlight from reaching photovoltaic modules, reducing power generation. Dust is aerosol pollution from natural or human-made sources. Dust on photovoltaic (PV) panels reduces power generation and raises the surface temperature, shortening panel life. This work uses a Fluke TiS60 Thermal Imager for detection of hotspot created by the accumulation of dust and further classification of the faulty and healthy images is done. The image classifiers used in this paper are SqueezeNet and AlexNet transfer learning methods, the accuracy of both methods is compared, and AlexNet performs best with an accuracy of 99.3%. Further, the effect of dust on PV modules is studied, and hotspots are created intentionally with the help of dust for study purposes. Initially, 50% of a solar module is covered with dust and then 100% of the solar module is covered with dust particles to find the power loss, when a thin layer of dust was spread over 50% of a module power loss is of 4.7% and when a thick layer of dust is spread over 50% the power loss is 5.9%, when thin layer of dust was applied over 100% of a module power loss is of 6.49% and when thick layer of dust is spread over 100% the power loss is 10.17%. This analysis for a healthy and faulty panel is done that helps PV module inspection by facilitating a more precise and cost-effective identification of PV defects.