A novel application of wavelet based SVM to transient phenomena identification of power transformers

• Published in: Energy conversion and management Vol. 52; no. 2; pp. 1354 - 1363
• Main Authors: Jazebi, S., Vahidi, B., Jannati, M.
• Format: Journal Article
• Language: English
• Published: Kidlington Elsevier Ltd 01-02-2011; Elsevier
• Subjects: Accuracy; Applied sciences; Computer programs; Computer simulation; Connection and protection apparatus; Differential protection; Electrical engineering. Electrical power engineering; Exact sciences and technology; Faults; Inrush current; Particle swarm optimization; Power electronics, power supplies; Power transformer; Support vector machine; Support vector machines; Transformers; Transformers and inductors; Wavelet; Wavelet transforms; Support vector machine; Power transformer; Particle swarm optimization; Inrush current; Differential protection; Performance evaluation; Cost minimization; High precision; Power electronics; Vector method; Computational complexity; Switching conditions; Wavelet transformation; Feature extraction; System simulation; Software; Economic aspect; Electric fault; Recovery (properties)
• ISSN: 0196-8904; 1879-2227
• DOI: 10.1016/j.enconman.2010.09.033

A novel differential protection approach is introduced in the present paper. The proposed scheme is a combination of Support Vector Machine (SVM) and wavelet transform theories. Two common transients such as magnetizing inrush current and internal fault are considered. A new wavelet feature is extracted which reduces the computational cost and enhances the discrimination accuracy of SVM. Particle swarm optimization technique (PSO) has been applied to tune SVM parameters. The suitable performance of this method is demonstrated by simulation of different faults and switching conditions on a power transformer in PSCAD/EMTDC software. The method has the advantages of high accuracy and low computational burden (less than a quarter of a cycle). The other advantage is that the method is not dependent on a specific threshold. Sympathetic and recovery inrush currents also have been simulated and investigated. Results show that the proposed method could remain stable even in noisy environments.