Fault detection scheme based on mathematical morphology in last mile radial low voltage DC distribution networks

This paper presents a fast fault detection scheme based on mathematical morphology (MM), which is a nonlinear signal transformation tool in a time-domain, in order to avoid the damage on diodes of power electronic devices in low voltage DC (LVDC) distribution networks. The MM is utilized to extract abnormal features from fault currents when short circuit faults with capacitive discharging characteristic or high impedance faults (HIFs) with arcing phenomenon occur in the network. The presented scheme utilizes a dilation erosion difference filter (DEDF) and closing opening difference filter (CODE) for detecting capacitive discharging current from the AC/DC converter and arc current with repetitive extinction and re-ignition, respectively. The two filters work in a coordinated way to meet the critical fault detection time (CFDT) for preventing damages on diodes in the AC/DC converter and avoid unfavorable tripping by applying different length of structuring element (SE). The framework for using the proposed scheme including a guideline for selecting the SE for each filter is also presented to make the best use of the proposed scheme. The presented scheme is implemented using the electromagnetic transients program (EMTP) and computer simulations are conducted on a test last mile radial LVDC distribution network.