Transmission Line Fault Classification under High Noise in Signal: A Direct PCA-Threshold-Based Approach

Transmission line faults are most common in long distance power transmission system. Classification of faults is crucial in removal of the faulted line from supply end in order to discontinue the unwanted flow of power through the fault point. The proposed work illustrates a simple method of classification of faults in transmission line using Principal Component Analysis- (PCA) based approach. The proposed method uses to extract fault features in terms of Principal Component Index (PCI), followed by a threshold-based analysis of the PCI values. Development of two threshold values helps in segregating the three different levels of fault disturbance in terms of the PCI values, thereby, developing fault signatures for classification. A 150 km transmission line has been modeled in EMTP for simulating the different fault prototypes, followed by analysis in MATLAB environment. Fault locations and fault resistances are also varied in order to enhance robustness in the classifier. Most importantly, noise level of the fault signals is varied in wide practical range; especially, signal to noise ratio (SNR) is increased to 15 dB to examine the robustness of the proposed classifier under severe noisy condition, where the same is found to produce 99.78% classifier accuracy using half cycle post-fault line currents only; thus, establishing the insensitivity of PCA toward noise.