Research on Photovoltaic Anomaly Detection Method Based on Unmanned Aerial Vehicle
A drone based intelligent photovoltaic inspection system is proposed to address the issues of high data dependence,large model parameters,and low execution efficiency during large-scale photovoltaic inspections.Firstly,the software and hardware structure of the system was designed,which can only upload the recognition results of the drone platform to the mo-nitoring center,thereby saving communication network bandwidth and improving inspection efficiency.Secondly,a lightweight photovoltaic anomaly detection model based on aerial images was proposed to meet the requirements of limited computing re-sources on unmanned aerial vehicle platforms.During the experimental phase,the performance of the ground server and drone platform models were compared.The experimental results of the ground server show that the proposed loss function leads to faster training convergence speed and better model performance.The experimental results of the drone platform show that the proposed model performs better than mainstream models,with an average detection accuracy of 86.83％.The simulation re-sults have verified that the proposed model provides a certain reference for the development of photovoltaic inspection and pow-er safety operation management.
NETL
NSTL
万方数据
