Fault Diagnosis Method for Photovoltaic Panels Based on Improved ShuffleNet V2 and Infrared Images

The long-term exposure of photovoltaic modules outdoors is prone to various faults that affect the safe and efficient operation of photovoltaic power plants. This paper studies a photovoltaic panel fault diagnosis method based on the improved ShuffleNetV2 network and infrared images of photovoltaic modules. First, the grayscale processing and data enhancement of the existing photovoltaic fault public data sets are used to obtain high-quality data sets. Then introduce the SENet attention mechanism in the lightweight network ShuffleNetV2 and improves the network unit structure, making it more suitable for the fault diagnosis task of photovoltaic modules. The experiment is compared with the MobileNetV2 and the original ShuffleNet V2 network. The improved method has a significantly improved fault diagnosis effect on the photovoltaic panel fault data set, and the overall accuracy rate reaches 84.06%, which has a certain practicability.