Classification of Multiple Electromagnetic Interference Events in High-Voltage Power Plant

This paper addresses condition assessment of electrical assets contained in High-Voltage (HV) electrical power plants. Our work introduces a novel analysis approach of multiple event signals related to faults, and which are measured using Electro-Magnetic Interference method. The proposed method transfers the expert's knowledge on events presence in the signals to an intelligent system which could potentially be used for automatic EMI diagnosis. Cyclic spectrum analysis is used as feature extraction to efficiently extract the repetitive rate and the dynamic discharge level of the events, and multi-class support vector machine is adopted for their classification. This first and novel method achieved successful results which may have potential implications on developing a framework for automatic diagnosis tool of EMI events.