Machine Learning Based Fault Distance Estimation of Underground Transmission Using Impedance Relay

Fast development of electric transmission infrastructure necessitates efficient methods of transmission and distribution for reliable cost effectiveness. Between the two main types of transmission, Overhead Transmission Lines involve extensive maintenance cost, use of extensive protection schemes, earthing, and lightning protection. In contrast, Underground Transmission Lines with their insulated and securely laid design are highly resistant to external factors such as lightning, wind, and open conductor faults, showing higher reliability and efficiency. However, faults such as symmetrical and unsymmetrical faults can occur due to conductors being bound together. The traditional fault distance estimation techniques are often inefficient for high-power transmission when using impedance-based fault distance relays. The constant variation in fault characteristics leads to impedance fluctuations, and therefore, fault distance calculations become inaccurate. The current research introduces a Machine Learning-based Fault Distance Estimation system for Underground Transmission Lines using Impedance Relays to address the above-mentioned challenges. Through data fed into the training machine learning model based on various fault scenarios, data mining is extracted for predicting precise impedance in faulty states and accurately determining fault locations by utilizing trained models with high-performance K-Nearest Neighbour (KNN), Decision Trees (DT), Support Vector Regression (SVR), Random Forest (RF), Long Short-Term Memory (LSTM), and Artificial Neural Networks (ANNs). Through these algorithms and approaches, implementations that emphasize optimal preprocessing, effective feature selection and tuning of specific hyperparameters determine their respective modeling efficiencies. The results show that the models generate very accurate distance estimations, allowing for more accurate fault locations and quick rectification. Not only does it bypass the deficiency of the methods in use today, but this method also goes further to bring forth intelligent and adaptive fault management systems that provide underground transmission network reliability and efficiency.