Machine Learning-Assisted Sampling Frequency Optimization for Traveling-Wave Fault Analysis in Two-Ended Power Transmission Networks

The continuous growth in electrical power demand necessitates highly reliable and fast fault detection and localization techniques for modern transmission networks. This study presents an advanced Traveling Wave (TW) based Fault Detection and Localization Algorithm (FDLA) for a two-ended transmission network comprising two synchronous generators with parallel-connected loads. The suggested system setup is initially verified for realistic operating conditions using steadystate Load Flow Analysis (LFA). After that, the TW voltage and current signals from both ends of the transmission line are used to find faults and locate their exact location in a number of different fault scenarios. A Machine Learning (ML) method based on Polynomial Regression (PR) is used to find the Optimum Sampling Frequency (OSF) for the faulty signals, which is employed by a Python program. This method focuses on the sampling rate point that gives the lowest percentage of localization error (<=0.001%). This optimization with the help of ML makes the TW-based fault analysis much more reliable and accurate. A MATLAB code is used to implement the full FDLA framework, which includes signal processing, fault distance, and error computation. The results in MATLAB/Simulink show that the proposed method finds faults with high accuracy and low computational error, which is superior to traditional methods. The suggested method is excellent for protecting high-speed connections in modern twoended transmission networks.