An Ensemble Machine Learning Based Fault Classification Method for Faults During Power Swing

In this paper, a novel method for phase selection of faults during power swing is presented. The proposed method operates with current signals from one end of the protected transmission line. The most popular and powerful ensemble machine learning technique Random Forest is used for fault classifica...

Full description

Saved in:
Bibliographic Details
Main Authors: Patil, Dinesh, Naidu, OD, Yalla, Preetham, Hida, S
Format: Conference Proceeding
Language:English
Subjects:
Decision trees
Fault classification
Fault diagnosis
Impedance
Moving average
Power swing Fault
Power swing unblocking
Protective relaying
Radio frequency
Random Forest
Online Access:Request full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:In this paper, a novel method for phase selection of faults during power swing is presented. The proposed method operates with current signals from one end of the protected transmission line. The most popular and powerful ensemble machine learning technique Random Forest is used for fault classification. The magnitude of the three-phase incremental quantities of currents is used as input to random forest for classifying the fault. The incremental quantity of the current signal is computed using a moving average filter. The proposed method is validated with 9216 cases covering varying swing frequency, inception angle, fault type, fault resistance, source to line impedance ratio and fault distance. Achieved fault classification within 10msec with an accuracy of 99.8%.
ISSN:2378-8542
DOI:10.1109/ISGT-Asia.2019.8881359