| With the continuous development of power technology and ubiquitous Internet of Things technology,the inefficient and backward power grid pattern has become an obstacle to the development of a smart society.Ultra high voltage network(UHV Network),as an efficient and mature power transmission technology,has become the source of power for grid structure,construction quality,large grid security,and innovative development.Safety is the foundation and primary goal to ensure the construction of UHV power grids.High-voltage circuit breakers are important control and protection equipment used in high-voltage and UHV power grids.The manual reading method has several defects,such as low efficiency,high cost,and poor safety.It has always been a difficult problem that plagued the safety of laser UHV power grids,which greatly affected the risks and costs of inspection during the actual operation and maintenance process,and severely hindered the automation of power grid development and intellectualization.Aiming at the above-mentioned key technical issues and the needs of the power grid’s worksite,this thesis designs a gas-insulated and metal-enclosed switchgear dial recognition system,which is based on related technical methods in the field of UHV power grids and artificial intelligence.Relevant experimental cases verify the effectiveness of the system.The main research contents of this thesis are as follows:(1)Research on image denoising model based on GAN NetworkAiming at the defects of manual inspection,this subject is oriented to the drone image deblurring method,and an image deblurring model is established.By studying the basic structure of the GAN,the network hyperparameters are optimized to optimize the network’s deblurring effect.Based on the trained GAN network,fuzzy samples are generated for model training and verification.(2)Research on pointer meter recognition method based on CNNAiming at the defects like poor generalization ability and low recognition accuracy of traditional shallow machine learning models,this thesis studied a CNN method for UAV inspection image analysis,basing on the application of CNN in the field of image classification and detection.The basic structure of CNN is designed and the network hyperparameters are optimized.(3)Prototype system development and case analysisBased on the aforementioned methods of image acquisition,image pre-processing,and image recognition,this thesis designed and developed a computer-aided prototype system for the design and implementation of the image dial reading recognition system for UAV inspection images.Intelligent analysis is taken as an example to verify the effectiveness of the theoretical method proposed in this thesis. |
