부산대학교 도서관

Accurate and reliable electric arc fault identification ensures the safety of personnel and equipment through in-field operations and maintenance work. However, due to the application scenarios and nonlinear arc dynamics, the distortion degree of the detectable signals is influenced by network/circuit structure, system parameter, operation manner, etc., the salience of the fault signatures may change significantly, which poses challenges to develop the “generic scheme” in arc faults identification. To address such issues, this paper proposed a robustness-improved arc fault identification approach that incorporates the modeling of dynamic impedance and chaotic behavior in faulting systems. For describing the “unique” nonlinear characteristics of arc faults among different operation scenarios, the HT-based dynamic impedance (HTDI) representation method is developed to extract the “signatures” of arc faults, and the mechanism of how system parameters impact the presentation form of arc fault features are also revealed. Meanwhile, for quantitively extracting the signatures, the chaotic behavior of arcing fault system has been analyzed in phase space in detail, with a series of chaos indicators. Additionally, HTDI and chaos indicators implemented LSTM classifier has been designed to achieve accurate and rapid arcing fault identification with time-series inputs. With the series of actual arc fault cases and simulation cases under different configurations, the effectiveness and robustness of the proposed method have been thoroughly validated.