论文信息 - Application of Principal Component Analysis (PCA) and SVMs for Discharges Radiated Fields Discrimination

Application of Principal Component Analysis (PCA) and SVMs for Discharges Radiated Fields Discrimination

This paper proposes the use of a fast method of discriminating magnetic signals radiated from electrical discharges occurring in overhead lines string insulations. Two types of electrical discharges exist: dangerous that can lead to arcing discharges and those without danger that auto extinguish. We propose as a new method of discriminating and classifying partial discharges principal component analysis (PCA) combined with support vector machines (SVMs) which have proved their robustness in several disciplines. However, the used database is composed of two classes, the majority representing 130 harmless radiated magnetic field signals, while the second minority class represents 31 dangerous signals. The learning and test sets correspond respectively to 2/3 and 1/3 of the database. The SVMs application to the test set shows that no dangerous signal is detected, this being due to the fact that the two classes are unbalanced. We were then asked to apply the Principal Component Analysis (PCA) even before classification, which made it able to select the most relevant variables. The results show that by using PCA and then SVMs, the detection rate of a dangerous signal is 90%.

[1] Cheng-Lin Liu,et al. Pattern Classification Methods , 2007 .

[2] Mourad Oussalah,et al. A new classification strategy for human activity recognition using cost sensitive support vector machines for imbalanced data , 2014, Kybernetes.

[3] Abderrahmane Beroual,et al. ANN and wavelet-based discrimination technique between discharge currents in transformer mineral oils , 2012 .

[4] Jean-Louis Sanchez,et al. Aide au diagnostic de défauts des transformateurs de puissance , 2011 .

[5] Nitesh V. Chawla,et al. Editorial: special issue on learning from imbalanced data sets , 2004, SKDD.

[6] David G. Stork,et al. Pattern Classification (2nd ed.) , 1999 .

[7] Vladimir N. Vapnik,et al. The Nature of Statistical Learning Theory , 2000, Statistics for Engineering and Information Science.

[8] A. Beroual,et al. Pre-breakdown current discrimination diagnose technique for transformer mineral oil , 2012, 2012 Annual Report Conference on Electrical Insulation and Dielectric Phenomena.