TY - GEN
T1 - Automated fuse identification within an operational smart power grid
AU - Wilson, Aaron J.
AU - Reising, Donald R.
AU - Hay, Robert W.
AU - Johnson, Ray C.
N1 - Publisher Copyright:
© 2018 IEEE.
PY - 2018/12/6
Y1 - 2018/12/6
N2 - During normal operation for power utilities an average of 2,100 events are recorded each month. Currently, utility personnel employ an archaic by-hand approach in the analysis of these events. This limits the analysis to roughly 2% of the total recorded events. This work applies pattern recognition and machine learning to perform automated identification of faults cleared by fuses. A critical step in the automation is the accurate detection of the waveform samples associated with the fault itself. Poor detection of these samples negatively impacts the identification performance. This work investigated three detection methods: Variance trajectory, binary vectors, and the analytic signal. A naïve Bayes classifier was employed to identify seven fuse sizes using let-through energy. Naïve Bayes identification performance of 94.3%, 93.6%, and 94.7% was achieved for the variance trajectory, binary vectors, and analytic signal detection approaches, respectively.
AB - During normal operation for power utilities an average of 2,100 events are recorded each month. Currently, utility personnel employ an archaic by-hand approach in the analysis of these events. This limits the analysis to roughly 2% of the total recorded events. This work applies pattern recognition and machine learning to perform automated identification of faults cleared by fuses. A critical step in the automation is the accurate detection of the waveform samples associated with the fault itself. Poor detection of these samples negatively impacts the identification performance. This work investigated three detection methods: Variance trajectory, binary vectors, and the analytic signal. A naïve Bayes classifier was employed to identify seven fuse sizes using let-through energy. Naïve Bayes identification performance of 94.3%, 93.6%, and 94.7% was achieved for the variance trajectory, binary vectors, and analytic signal detection approaches, respectively.
UR - http://www.scopus.com/inward/record.url?scp=85060389637&partnerID=8YFLogxK
U2 - 10.1109/APPEEC.2018.8566414
DO - 10.1109/APPEEC.2018.8566414
M3 - Conference contribution
AN - SCOPUS:85060389637
T3 - Asia-Pacific Power and Energy Engineering Conference, APPEEC
SP - 719
EP - 723
BT - IEEE PES Asia-Pacific Power and Energy Engineering Conference, APPEEC 2018
PB - IEEE Computer Society
T2 - 10th IEEE PES Asia-Pacific Power and Energy Engineering Conference, APPEEC 2018
Y2 - 7 October 2018 through 10 October 2018
ER -