Power Quality Event Detection Using a Fast Extreme Learning Machine

• Main Authors: Ferhat Ucar, Omer F. Alcin, Besir Dandil, Fikret Ata
• Format: Article
• Language: English
• Published: MDPI AG 2018-01-01
• Series: Energies
• Subjects: event detection; power quality; histogram; machine learning; wavelet transform
• Online Access: http://www.mdpi.com/1996-1073/11/1/145

Monitoring Power Quality Events (PQE) is a crucial task for sustainable and resilient smart grid. This paper proposes a fast and accurate algorithm for monitoring PQEs from a pattern recognition perspective. The proposed method consists of two stages: feature extraction (FE) and decision-making. In the first phase, this paper focuses on utilizing a histogram based method that can detect the majority of PQE classes while combining it with a Discrete Wavelet Transform (DWT) based technique that uses a multi-resolution analysis to boost its performance. In the decision stage, Extreme Learning Machine (ELM) classifies the PQE dataset, resulting in high detection performance. A real-world like PQE database is used for a thorough test performance analysis. Results of the study show that the proposed intelligent pattern recognition system makes the classification task accurately. For validation and comparison purposes, a classic neural network based classifier is applied.