Evaluating windowing based continuous ST with deep learning for detection and classifying PQDs

• Published in: IOP conference series. Materials Science and Engineering Vol. 1088; no. 1; p. 12060
• Main Authors: Daud, K, Mansor, MBM, Soh, ZH Che, Samat, AA Abd, Shafie, MA, Ismail, AP, Abdullah, MH
• Format: Journal Article
• Language: English
• Published: Bristol IOP Publishing 01.02.2021
• Subjects: Classification; Deep learning; Disturbances; Feature extraction; Machine learning; Neural networks; Performance evaluation; Programming languages; Scattering; Signal processing
• ISSN: 1757-8981; 1757-899X
• DOI: 10.1088/1757-899X/1088/1/012060

Abstract This paper discusses the performance of evaluating a windowing based continuous S-Transform (ST) with deep learning classifier for the detection and classifying of interrupt and transient disturbances. The primary purpose is to analyze the detection and classification of voltage interrupt and transient using ST as a signal processing technique. The detection technique is divided into half-cycle and one-cycle windowing techniques (WT with both cycles used for the purpose of comparison. The disturbances signal was create using MATLAB programming language and set in the form of m-file. ST was used to extract the significant feature in a form of scattering data from the disturbances signal. Then, the scattering data was used to build the detection interface inside the disturbances signal. The scattering data is an input for neural network (NN) to classify the percentage accuracy of the disturbances signal. This analysis presents the suitable windowing technique that can provide smooth detection and suitable characteristics to produce high accuracy percentages in the classification of power quality disturbances (PQDs).