Prediction Of Fault In High Voltage Overhead Transmission Line Based On Clustering Method

The power transmission line is a very important part of the power system for reliability and safe operation.The fault in transmission directly threatens the safe operation and causes power failure and blackout.The accurate prediction and assessment of the transmission line operating status can provide technical support to the economy and improve the efficiency of an electric power system.The running state of an overhead transmission line is affected by operating conditions and weather conditions.The major cause of transmission line outage is severe weather characteristics like wind,icing,and lightning.The thesis proposes the methodology of weather and fault correlation influence on transmission line failure in terms of unknown factors originating from the severe weather conditions.The previous method proposed has some limitations due to low-quality data is collected from the online monitoring systems and sudden variation in meteorological conditions,and it requires proper information of historical faults events parameter.In this work,data is collected only from an online monitoring system to study the prediction model of transmission line failure.An unsupervised machine learning prediction algorithm is proposed based on Principal Component Analysis(PCA)and K-Means clustering algorithm.Firstly,the principal component analysis is used to eliminate the less influencing factor and reduce the dimension of data then,Next,the cluster is constructed using the K-Means algorithm,and weather fault occurrence is predicted by measuring the distance between the centroid to farthest point.To evaluate the performance of result,the other unsupervised machine learning algorithms,One-Class Support Vector Machine(OC-SVM)and Isolation Forest,are used to get comparative results and model accuracy.Finally,the comparison shows when considering the mean and standard deviation value of K-means clustering,and Isolation Forest algorithm is mostly similar with the error of approximately 1%while,the difference of K-mean and OC-SVM is more.This add detail that the prediction of failure detected from K-mean,and Isolation Forest is similar.