| Title |
AI-based Fault Detection Algorithm Using Wavelet Transformation in DC System |
| Authors |
Jun-Soo Che ; Su-Han Pyo ; Tae-Hun Kim ; Byeong-Hyeon An ; Jae-Deok Park ; Tae-Sik Park |
| DOI |
http://doi.org/10.5207/JIEIE.2023.37.1.033 |
| Keywords |
AI algorithm; Artificial neural network; DC fault detection; High impedance fault; Wavelet transformation |
| Abstract |
DC grids are being considered for renewable energy connection and underwater power transmission. There are overcurrent-based methods and methods using frequency thresholds as fault diagnosis methods currently used in DC systems. However, since the fault current is affected by the ground impedance, it may be difficult to accurately diagnose the fault with the conventional method in the case of HIF. In addition, it is difficult to expect high accuracy when diagnosing a fault location because the ground impedance value varies depending on the weather and season. Therefore, this study proposes an AI-based fault detection algorithm. When a fault occurs in a DC system having a ripple component connected to a rectifier, it has a different frequency pattern depending on the fault location. In this study, a ground fault was simulated through MATLAB Simulink and the fault signal was analyzed in the time and frequency domains. Based on the analysis results, the fault current was wavelet transformed to train the DNN. Then, the model was validated through additional fault data and HIF failures. |