Fault Diagnosis in Transmission Line Using Empirical Mode Decomposition

This paper proposes the concept of fault diagnosis (both detection and localization) in a power transmission line using the method of empirical mode decomposition (EMD) and M.L. algorithms like random forest classifier and recurrent neural network (RNN). Firstly, signal processing is performed on the current signals, captured during a fault event, to generate IMFs. Using the IMFs, EMD is used to decompose the current signals. FFT analysis is also performed to differentiate between the signal frequencies in faulted line and the healthy line. Finally, a web application is hosted using Streamlit to display the actual fault distance, predicted fault distance, and error in meters. This method of EMD, random forest classifier, and RNN give an appreciable accuracy.