Análise do processo de wavelet shrinkage na extração de ruído de sinais de descargas parciais e separação dos defeitos associados

• Ano de defesa: 2019
• Autor(a) principal: Cunha, Caio Fleming Ferreira de Carvalho
• Tipo de documento: Tese
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal do Rio de Janeiro; Brasil; Instituto Alberto Luiz Coimbra de Pós-Graduação e Pesquisa de Engenharia; Programa de Pós-Graduação em Engenharia Elétrica; UFRJ
• Palavras-chave em Português: Descargas parciais; Transformada wavelet; Redução de ruído; Detecção de polaridade; Separação por agrupamento; CNPQ::ENGENHARIAS::ENGENHARIA ELETRICA::CIRCUITOS ELETRICOS, MAGNETICOS E ELETRONICOS
• Link de acesso: http://hdl.handle.net/11422/22719

In this work we deal with the digital processing of signals related to the phe- nomenon of partial discharges (PD), which manifests itself in high voltage equip- ment. Initially, in the field of noise extraction using the wavelet transform, new optimized thresholding functions are proposed to treat the wavelet coefficients of the PD pulses, which presented superior results compared to those obtained with several functions commonly applied. In addition, the performances of different com- binations of mother-wavelet functions in PD filtering are verified. Subsequently, special attention is given to the pattern separation method by normalized autocor- relation function (NACF), obtaining a better performance in relation to the original proposal. We also describe the design of a PD data analysis and separation system. In addition to the usual signal processing techniques, the system includes a new pulse polarity identification technique based on the correlation coefficient between the evaluated pulse and a standard pulse, providing an expressive gain in the quality of the phase resolved partial discharge (PRPD) map generated in a measurement. In addition, a method is presented to estimate the frequency response of the whole measurement system, from its origin to the acquisition hardware, an alternative that presented promising results. The system also has a set of techniques to extract char- acteristics of the signals presented to it, for later separation with clustering methods, allowing the comparison between them.