Partial discharge detection on aerial covered conductors using time-series decomposition and long short-term memory network

Abstract

Nowadays, aerial covered conductors (CC) are increasingly used in many places of the world due to their higher operational reliability, reduced construction space and effective protection for wildlife animals. In spite of these advantages, a major challenge of using CC is that the ordinary upstream protection devices are not able to detect the phase-to-ground faults and the frequent tree/tree branch hitting conductor events on such conductors. This is because these events only lead to partial discharge (PD) activities rather than overcurrent signatures typically seen on bare conductors. To solve this problem, in recent years, ENET Centre in Czech Republic (ENET) devised a simple meter to measure the voltage signal of the electric stray field along CC, aiming to detect the above hazardous PD activities. In 2018, ENET shared a large amount of waveform data recorded by their meter on Kaggle, the world's largest data science collaboration platform, encouraging worldwide experts to develop an effective pattern recognition method for the acquired signals.For this challenge, we developed a unique method based on time-series decomposition and Long Short-Term Memory Network (LSTM) in addition to unique feature engineering process to recognize PD activities on aerial covered conductors. The proposed method is tested on the ENET public dataset and compared to various traditional classification methods. It demonstrated superior performance and great practicality.

Interpretable Detection of Partial Discharge in Power Lines with Deep Learning

“…To the best of our knowledge, no other published study outside of the competition leaderboard reported results on the second test dataset. In [ 12 , 18 , 19 ], the reported results are computed on a subset of the labeled dataset. In [ 12 ], results are reported on the full training set and might therefore be overfitted.…”

“…In [ 12 ], results are reported on the full training set and might therefore be overfitted. In [ 18 ], results are reported on an artificially augmented set containing 807 non-PD signals and 935 signals with PD, which might also therefore suffer overfitting. We report anyway their results in Table 2, where we recompute the value of the metrics they would achieve on our set, assuming constant sensitivity and specificity of their model.…”

“…They aim at characterizing the pulses with engineered features. The features become a multidimensional space where decision boundaries are established, for example with traditional classifiers such as random forest [ 7 ], support vector machines (SVM) [ 11 , 12 ], or deep learning methods such as artificial neural networks (ANN) [ 13 ], convolutional neural networks (CNN) [ 14 , 15 ], autoencoders [ 16 ], and recurrent neural networks (long short-term memory networks (LSTM)) [ 17 , 18 , 19 ]. For a more exhaustive overview, the reader is referred to the literature reviews in [ 20 , 21 ].…”

Neural Network Based Country Wise Risk Prediction of COVID-19

“…(C) Multivariate Regression in AI: The key point in time series study [40] is forecasting. Time series analysis for business prediction helps to forecast the probable future values of a practical field in the industry [41][42][43][44].…”

Neural Network Based Country Wise Risk Prediction of COVID-19

“…The key point in time series study [35] is forecasting. Time Series analysis for business prediction helps to forecast the probable future values of a practical field in the industry [36][37][38][39].…”