Abstract: Aiming at the problems of a large number of components in large-scale photovoltaic fields,difficulty in defect detection,and low efficiency of operation and maintenance decision-making,an unmanned aerial vehicle（UAV）visual inspection and twin-driven decision-making method for intelligent operation and maintenance of photovoltaic fields is proposed.Firstly,to address the problem of unclear small-scale defect features such as micro cracks and pollution obstruction in photovoltaic modules,which are easily affected by complex background interference,a cross scale feature fusion network is constructed,and channel and spatial attention enhancement mechanisms are introduced to improve the recognition accuracy and robustness of small target defects.Secondly,based on the drone detection results,a digital twin of the photovoltaic field is constructed,and a dynamic update model for the component health index is designed to achieve real-time mapping between the physical component status and the virtual twin model.Finally,a model for prioritizing operations and risk propagation assessment is constructed to achieve closed-loop decision-making for defect identification,health assessment,and optimized allocation of operations and maintenance resources.The experimental results show that this method is superior to traditional methods in component health assessment,risk area identification accuracy,operation decision-making efficiency,and can effectively improve the intelligent operation and maintenance level and operational safety of large-scale photovoltaic fields.