Hybrid Framework for Fault Location Estimation in Transmission Lines using Linear Regression

Accurate and rapid fault location in transmission lines is critical to ensuring the reliability and stability of power systems. Faults in transmission lines may result from insulation failure, lightning strikes, or external damage, and they must be identified and resolved swiftly to reduce downtime. This paper proposes a linear regression-based fault location estimation model using machine learning techniques for transmission lines. The objective is to develop a predictive model that accurately estimates the distance to the fault from a single-end measurement, enabling timely and cost-effective maintenance. The proposed method employs MATLAB/Simulink to simulate various fault scenarios in a 400 kV, 100 km three-phase transmission system, while Python is utilized for building and training a machine learning model based on the collected current data. The methodology involves injecting faults at multiple points along the line and capturing current signals at the sending end. A dataset is then constructed to establish a relationship between time and current, which is analyzed using linear regression under the Least Squares Estimation (LSE) framework. The results indicate a strong correlation between the time and current variables, validating the applicability of the linear model. Both simulated and calculated outputs yield similar slope and intercept values, affirming the accuracy of the Python-based predictive model. Visual analyses using regression plots and current waveforms for different fault types, single line-to-ground (SLG), line-to-line (LL), double line-to-ground (LLG), and three-phase (LLL) demonstrate the model's reliability. However, the model shows limitations when applied to complex three-phase faults, suggesting future enhancements using multiple regression or neural network models. This study confirms the effectiveness of linear regression as a predictive tool for fault location in transmission lines and provides a foundation for integrating more advanced machine learning models for improved fault diagnosis.