Affiliations: Department of Electrical and computer Engineering, Semnan University, Semnan, Iran | Department of Electrical and computer Engineering, Babol Noshirvani University of Technology, Babol, Mazandaran, Iran
Note: [] Corresponding author. Asghar Akbari Foroud, Department of Electrical and computer Engineering, Semnan University, Postal Code 35131-19111, Semnan, Iran. Tel.: +989124618433; Fax: +0231 3354123; E-mails: aakbari@semnan.ac.ir; akbari_asghar@yahoo.com
Abstract: In the literature of power transformer protection, the essential issue consists of discriminating between transformer inrush current and internal fault current. This paper presents a novel scheme for the protection of power transformers utilizing Multi Resolution Analysis (MRA) and Smooth Supported Vector Machine (SSVM). Some important characteristics of differential currents are extracted based on MRA. The combination of MRA and SSVM is proposed as an effective and fast technique for discrimination of the faulted and unfaulted conditions. The Particle Swarm Optimization (PSO) is used to obtain optimal values of adjustable parameters of the applied classifier. The efficacy of the proposed scheme has been tested using numerous inrush and internal fault currents. The results confirm that the combined MRA-SSVM scheme yields high detection accuracy even in the noisy conditions.
Keywords: Transformer protection scheme, pattern recognition, multi resolution analysis (MRA), smooth supported vector machine (SSVM), inrush and internal fault currents
