Submitted by Automa??o e Estat?stica (sst@bczm.ufrn.br) on 2018-04-03T14:45:11Z No. of bitstreams: 1 FrankelenePinheiroDeSouza_DISSERT.pdf: 1803265 bytes, checksum: c75d05de7793272bb03cf2417c8e6d84 (MD5) / Approved for entry into archive by Arlan Eloi Leite Silva (eloihistoriador@yahoo.com.br) on 2018-04-10T19:21:14Z (GMT) No. of bitstreams: 1 FrankelenePinheiroDeSouza_DISSERT.pdf: 1803265 bytes, checksum: c75d05de7793272bb03cf2417c8e6d84 (MD5) / Made available in DSpace on 2018-04-10T19:21:14Z (GMT). No. of bitstreams: 1 FrankelenePinheiroDeSouza_DISSERT.pdf: 1803265 bytes, checksum: c75d05de7793272bb03cf2417c8e6d84 (MD5) Previous issue date: 2017-12-20 / O Conselho Nacional de Desenvolvimento Cient?fico e Tecnol?gico - CNPq / Coordena??o de Aperfei?oamento de Pessoal de N?vel Superior (CAPES) / Os sistemas de prote??o convencionais baseados em sobrecorrente, de um modo geral, n?o se sensibilizam com a ocorr?ncia de faltas de alta imped?ncia, pois essas correntes apresentam baixa amplitude. Esse tipo de dist?rbio causa preju?zos ?s concession?rias, como tamb?m, pode causar mortes de pessoas e animais. Diante disso, distintos m?todos para a identifica??o das faltas de alta imped?ncia em sistemas de distribui??o de energia el?trica v?m sendo propostos. A principal dificuldade encontrada por esses m?todos ? devido a interfer?ncia do ru?do no sinal, assim como o baixo n?vel de sobrecorrente de falta. Diante da busca por um m?todo seguro, eficiente e confi?vel para identificar as faltas de alta imped?ncia, e que n?o apresente os inconvenientes encontrados nos m?todos proposto na literatura, prop?e-se um m?todo inovador que minimiza a interfer?ncia do ru?do no sinal, e utiliza caracter?sticas estat?sticas presentes nos sinais com falta de alta imped?ncia para os identificar, e assim os diferenciar de outros fen?menos presentes nos sistemas el?tricos. Especificamente, o m?todo usa an?lise de processos cicloestacion?rios para extrair informa??es de autocorrela??o c?clica das amostras dos sinais de interesse, pelo c?lculo da fun??o densidade espectral c?clica. A partir dessas informa??es, podese obter descritores que permitem classificar em uma etapa posterior as faltas de alta imped?ncia. Resultados indicaram que o m?todo ? robusto ao ru?do e eficiente em termos de taxa de classifica??o correta dos eventos de falta de alta imped?ncia, como tamb?m o m?todo n?o emite falsos alertas ao sistema. / Conventional overcurrent protection systems are often not sensitized to the occurrence of high impedance faults, because these currents have a low amplitude. This type of disturbance causes damages to the ulilities, as well as can cause deaths of people and animals. Therefore, different identification methods of high impedance faults in electric power distribution systems have been proposed. The main difficulty found by these methods is due to noise interference on the signal, as well as the low fault overcurrent level. Therefore to search for efficient and reliable method to identify high impedance faults, which does not present the drawbacks found in the methods proposed in the literature, an innovative method is proposed that minimizes the noise interference on the signal, and uses characteristics statistics present on the signals with high impedance fault to identify them, and thus to differentiate them from other phenomena present in the electrical systems. Specifically, the method uses analysis of cyclostationary processes to extract cyclic autocorrelation information from the samples of the signals of interest by calculating the cyclic spectral density function. From this information, we can obtain descriptors that allow to classify at a later stage the faults of high impedance. Results indicated that the method is robust to noise and efficient in terms of correct classification rate of high impedance fault events, as well as the method does not emit false alerts to the system.