Thermal inspection of photovoltaic modules with deep convolutional neural networks on edge devices in AUV

•Proposed assessment method based on deep convolutional neural network for solar photovoltaic (PV) systems to reduce energy conversion lost and extend the life cycle.•Using an infrared radiation (IR) camera on a AUV to capture the IR image of the PV helps to fast, efficient, and remote inspection.•A Gradient-weighted class activation mapping module was implemented in the model that helps to determine the location of the anomaly and improve the reliability of the detection result.•The proposed method provides a fast, easy with high accuracy of inspection and is applicable for real-life applications.