Skip to Main Content
This paper presents the development and hardware implementation of a classification scheme to distinguish the transients originated by faults from other types of transients. In the proposed scheme, a set of Hidden Markov Model-based classifiers is employed to recognize the fault transients. Input features for the classifiers are the energy contained in wavelet coefficients of the measured current waveforms. A laboratory prototype of the fault recognition system was implemented on a floating-point digital-signal-processor (DSP)-based hardware platform. The classification system was tested using the transient signals generated by a real-time waveform playback unit. The test waveforms were generated by simulating an actual extra-high-voltage transmission system on an electromagnetic transient simulation program. The operation of the classification system was further verified using waveforms obtained from an actual fault recorder. The performance of the classifier was investigated under different practical scenarios, such as current transformer saturation, measurement noise, and lightning faults.