Research on Insulator Defect Recognition Algorithm for Overhead Transmission Line in Power Grid
Overhead transmission line insulators in power grid has been in the external exposure environment for a long time,which is prone to erosion,cracking,crushing and other problems,leading to a reduction in the reliability of grid operation.For this reason,the overhead transmission line in power grid insulator defect recognition algorithm is researched.The collected insulator image multiplicative noise is removed by filtering,and the surface defect features are extracted by utilizing different scale Gaussian masks for convolution processing,combined with the image gradient direction.The constructed deep feed-forward neural network is used as the basic architecture,and the self-encoder with three-layer network structure is set up.The penalty factor is incorporated into the optimization objective function of the feed-forward network,and the sparsity constraints are constructed.The initial weights of the self-encoder are optimized by genetic algorithm,and the surface defect recognition results are obtained by network iteration.The experimental results show that the proposed algorithm always has a recognition accuracy of insulator defects higher than 90％,and the average recognition time is 1.72 s,which has significant feasibility.This research is of great revelation to the fields of monitoring and security,image analysis and so on.
Power gridOverhead transmission linesDeep learningInsulatorGaussian maskSelf-enconderDefect identification
万方数据
维普
