การวิเคราะห์ซิกเนเจอร์ของความผิดพร่องในระบบส่งไฟฟ้าโดยใช้การรู้จำรูปแบบ / พลสัณห์ พงษ์ประยูร = Fault signature analysis in a transmission system using pattern recognition / Ponson Pongprayoon
To propose a method to analyze fault signatures which results from insulation flashover, burning smoke, service crane and faults in adjacent distribution using data recorded by Digital Fault Recorders (DFR) and fault event analysis report of the Electricity Generating Authority of Thailand (EGAT), during B.E. 2550-2554. Applying feature extraction to seven types of input data, the Artificial Neural Network and the Decision Tree have been constructed for classifying the four fault types. Then, selecting the sufficient features using Sequential Forward Selection (SFS) reveals that only 5 features can satisfy, which are fault clearing times, recloser operation, the number of affected phases, fault developing characteristic and Root Mean Squares (RMS) of the neutral current. Comparative test results shows that both the Artificial Neural Network and the Decision Tree can provide high accuracy of the classification in which the highest accuracy of 96.67% can be obtained when testing with the 90 actual fault events. Additionally, the signature of those four fault types are analyzed and discussed.