Analysis And Processing Of Wind Dancing Monitoring Signals Of Overhead Transmission Lines

Power system plays an extremely important role in our daily life.Overhead transmission lines,as a carrier for transmitting electric energy,indirectly bear the responsibility of supplying power to thousands of households.The unusually large-amplitude and low-frequency wind-dance phenomenon of overhead power transmission lines due to ice-covering and other factors may lead to the breakage of transmission lines,which in turn causes power interruptions to cause large-scale blackouts.Therefore,the monitoring and early warning of abnormal wind-dance signals of overhead transmission lines is particularly important.The existing transmission lines wind dancing monitoring systems include network camera-based monitoring system,electrical sensor-based monitoring system,and monitoring system based on a fiber bragg grating(FBG)sensor array.The camera-based monitoring system can realize qualitative observation and simple quantitative monitoring of wind-dance signals,but it cannot obtain more accurate quantitative monitoring results,neither it cannot perform further quantitative analysis(such as spectrum analysis)based on the monitoring results.The monitoring system of the electrical sensor needs a large amount of power supply,but in actual situations often cannot meet this power supply demand,and the electrical sensor has weak anti-interference ability and is not easy to maintain.The fiber Bragg grating(FBG)sensor array can realize quantitative monitoring,is anti-interference and easy to maintain,supports distributed monitoring and has other advantages,but for the distributed monitoring scene with many sensor nodes,the load capacity of the transmission sensor is weak,so it is not suitable for long-distance monitoring scenarios such as transmission lines wind dancing monitoring.The distributed optical fiber sensing system using OTDR technology has the advantages of anti-jamming,easy maintenance,quantitative monitoring,and strong load capacity for large amounts of sensing data in long-distance scenarios.It is suitable for long-distance overhead transmission lines wind dancing online monitoring.This article uses distributed optical fiber vibration sensing system developed by the laboratory to collect wind dancing data of overhead power transmission lines,and then builds data sets,extracts features and fuses the extracted features,uses machine learning algorithms based on supervised analysis to design classification models and uses models to recognize normal wind dancing data and abnormal wind dancing data.Classification and identification of normal data and abnormal data,evaluates the classification and recognition effect of each classification model based on recall rate and precision and screening out the classification model with the best classification and identification effect as the classification model of transmission lines wind-dance anomaly monitoring system.The specific work is as follows:1.Summarizes the domestic and foreign research status of overhead transmission lines monitoring technology,distributed optical fiber sensing technology and sensing signal processing algorithms.2.Based on polarization-sensitive fiber-optic time domain reflectometry(P-OTDR)technology,a distributed fiber-optic vibration sensing system for over-the-air cable style monitoring is designed.This part includes the sensing mechanism,the acquisition method of the sensing signal,and the various functional modules and functions of the system.3.According to signals colected by the distributed optical fiber vibration sensing system to build data sets.The signal within 24 hours of the spatial sample point is divided into frames by 50 s to obtain minimum signal processing units.Based on box-pattern statistics and spectral correlation analysis methods,all framing signals divided by signals of spatial sample points are classified,that is,the framing signals are divided into normal wind dancing signals and abnormal wind dancing signals.The classified framing signal data are divided into a training set and a test set according to a ratio of 6:1 for training and testing of subsequent classification models.4.Feature extraction and feature fusion of the signal,and data augmentation of the abnormal data in the training set after feature extraction and feature fusion.The extracted features include time domain statistical features,frequency domain features,linear predictive coding coefficient features and time-frequency chromagram features.The feature selection is performed,ie,the features are analyzed,and the feature that can distinguish the normal signal from the abnormal signal is selected.For the selected different types of features,feature fusion is performed to obtain fusion features.The data of the abnormal signal features in the training set are augmented to obtain an augmented training set.The proportion of normal data and abnormal data in the augmented training set is 1:1.5.Uses classification algorithms based on supervised analysis,that is,nearest neighbor classification algorithm,logistic regression classification algorithm,support vector machine classification algorithm,and convolutional neural network classification algorithm to design the wind dancing signal classification model.Uses original training set data and augmentation training set data respectively when training classification models.Uses the training set data to train classification models and the test set data to test the classification effect of each classification model,and then evaluates the classification performance of each classification model based on the recall rate and precison,analyzes the advantages and disadvantages of each classification model,and finally selects the classification model with the best classification and recognition effect.The selected classification model is used as the classification model for anomaly monitoring system for overhead transmission lines.