Complementary Feature Extractions for Event Identification in Power Systems Using Multi-Channel Convolutional Neural Network

This paper presents an event identification process in complementary feature extractions via convolutional neural network (CNN)-based classification. The CNN is a suitable deep learning technique for addressing the two-dimensional power system data as it directly derives information from measurement signal database instead of modeling transient phenomena, where measured synchrophasor systems are allocated by time and space domains. dynamic signatures phasor unit (PMU) signals analyzed based on starting point subtransient signals, well fluctuation signature signal. For fast decision protective operations, use narrow band window recommended to reduce acquisition delay, wide provides high accuracy due large amounts data. In this study, two separate preprocessing methods multichannel structures constructed provide validation, successive conditions. result includes pertaining various types locations delays operation. Finally, work verifies method through case study analyzes effects events addition classification accuracy.