Study On The Deduction Of Bird Damage Distribution Of Overhead Transmission Line Based On Image Case Segmentation

With the rapid development of China’s economy,the industry has put forward a higher demand for the stable operation of the power system.Once the power failure occurs,it will bring great loss to people’s life.However,due to the favorable geographical environment and ecological climate in guangdong coastal hilly area,transmission line failures caused by bird damage occur frequently,and the safety of transmission lines is seriously threatened.Transmission lines are widely distributed,and the factors associated with bird activity are not unique and fixed.The harm range of birds will also change every year due to changes in the ecological environment,blind artificial qudiao is difficult to achieve effective prevention of bird damage.Therefore,how to reduce bird damage fault has become a new problem for overhead transmission line operation and maintenance.For wide distribution of transmission lines,the sound every year due to the ecological environment change to the characteristics of the example in this thesis,based on image segmentation algorithm for overhead transmission line deduction,an in-depth study is made on the sound distribution of maintenance personnel in power system in such aspects as the installation of Qu Niao provide instructional decisions,so as to avoid blindness to work on prevention and control of the sound malfunction.The contribution of the proposed bird damage distribution on overhead transmission lines is as follows:（1）Due to the influence of impurities such as fog and haze in the atmosphere,the tile map image directly taken by satellite will appear blurred.If it is directly trained in the example segmentation algorithm Mask R-CNN,the extracted geographical environment features are difficult to achieve ideal results.Therefore,the dark channel algorithm is firstly used to remove haze from the tiled map image collected in this thesis,which effectively improves the clarity of the image.（2）As an important factor to deduce the distribution of bird damage on transmission lines,the geographical environment features will consume a lot of manpower and material resources,so it is necessary to train the neural network to automatically extract the geographical environment features of the whole Guangdong region.The Mask R-CNN adopted in this thesis can not only effectively detect the target in the tile image,but also provide a high-quality segmentation result for the geographical environment features to achieve accurate extraction of the target.（3）In view of the uncertain factors such as complex geographical environment characteristics and unclear operation and maintenance situation,and the areas with bird damage faults on overhead transmission lines are in dynamic change,this thesis will collect the bird damage operation and maintenance data of transmission lines in Guangdong Power grid from 2015 to2019 and remove the bird nest data of poles and towers.Then,the relationship between geographical environment characteristics,tower characteristics,season and bird damage fault was analyzed,and the bird damage model was constructed by Stacking model integration.Stacking model fusion effectively improves prediction accuracy and generalization capability compared to a single model.（4）The bird hazard distribution of transmission lines in coastal hilly areas of Guangdong Province is deduced based on the geographical environment characteristics automatically adopted by Mask R-CNN and integrated learning Stacking model,which provides a guiding basis for transmission lines operation and bird hazard fault prevention to realize intelligent bird hazard prevention.