Souza, André N. [UNESP]Da Costa Jr., Pedro [UNESP]Da Silva, Paulo S. [UNESP]Ramos, Caio C. O.Papa, João Paulo [UNESP]2014-05-272014-05-272011-12-212011 16th International Conference on Intelligent System Applications to Power Systems, ISAP 2011.http://hdl.handle.net/11449/73076In this paper we propose an accurate method for fault location in underground distribution systems by means of an Optimum-Path Forest (OPF) classifier. We applied the Time Domains Reflectometry method for signal acquisition, which was further analyzed by OPF and several other well known pattern recognition techniques. The results indicated that OPF and Support Vector Machines outperformed Artificial Neural Networks classifier. However, OPF has been much more efficient than all classifiers for training, and the second one faster for classification. © 2011 IEEE.engFault LocationOptimum-Path ForestPattern RecognitionUnderground SystemsArtificial Neural NetworkPattern recognition techniquesReflectometrySignal acquisitionsTime domainUnderground distribution systemUnderground systemsElectric fault locationForestryIntelligent systemsNeural networksPattern recognitionPower transmissionSignal processingTime domain analysisAlgorithmsClassificationDefectsElectric Power DistributionForestsNeural NetworksTrabalho apresentado em evento10.1109/ISAP.2011.6082204Acesso aberto2-s2.0-836551976679039182932747194