A new methodology for multiple incipient fault diagnosis in transmission lines using QTA and Naïve Bayes classifier. (December 2018)

• Record Type: Journal Article
• Authors: da Silva, Paula Renatha Nunes; Gabbar, Hossam A.; Vieira Junior, Petrônio; da Costa Junior, Carlos Tavares
• Is Part Of: International journal of electrical power & energy systems. Volume 103(2018)
• Page Start: 326
• Page End: 346
• Publication Date: 2018-12
• Subjects: Fault diagnosis -- Multiples faults -- Qualitative trend analysis -- Naïve Bayes -- Leakage current -- Transmission lines; Electrical engineering -- Periodicals; Electric power systems -- Periodicals; Électrotechnique -- Périodiques; Réseaux électriques (Énergie) -- Périodiques; Electric power systems; Electrical engineering; Periodicals; 621.3
• DOI: 10.1016/j.ijepes.2018.05.036
• Languages: English
• ISSNs: 0142-0615

Abstract

Highlights: Diagnosis methodology for multiple incipient faults in transmission lines. Experimental Leakage current (LC)-based predication of different fault scenarios modeled by equivalents circuits. Features extraction to define qualitative trends and multiple fault modes based on polynomial approximations. NB classifier designed to identify the most prominent fault in a multiple faults. Approach NB-QTA was applied to the detection and discrimination of both single and multiple faults with 95% accuracy. Abstract: Transmission lines' monitoring systems are hybrid-dynamical structures produce a large amount of data that renders faults diagnosis difficult. Moreover, multiple faults of unknown nature can occur simultaneously impeding their discrimination. In addition, the feature extraction step constitutes an inherent limitation to the diagnosis of multiple faults. Indeed, a feature extraction algorithm avoiding any masking effect among between different multiple faults is hard to devise. This paper proposes a methodology dedicated to the diagnosis of multiples faults in transmission lines using both and experimental and calculated Leakage Current (LC) signals' harmonics as residuals. Measured data from the real transmission line was used to modelling normal operating mode. Besides, different scenarios, including insulator chains contamination with different types and concentrations of pollutants were modeled by equivalents circuits to generate a multiple faults scenario. LCs from deteriorated insulators were inserted in the transmission line normal operating mode to implement a faulty operating conditions model. Qualitative trend analysis (QTA) was used to generate primitives and by exploiting the leakage current difference between normal and faulty operating conditions to define a features extraction algorithm for the diagnosis of specific fault modes. A Naïve Bayes classifier was designed in order to identify the most prominent fault in a multiple faults scenario by means of LC data. The methodology manages to split multiple faults into single faults and reaches a classification accuracy for multiple faults of 95%.