When a fault occurs in a DC system, the fault current rises rapidly with no zero-crossing point which makes fault-line selection and fault-type identification difficult. In this paper, an online detection and protection method based on graph theory, namely the “double D method”, is proposed for fault-line selection and fault-type identification in DC systems. In the proposed method, the entire distribution network is visualized as a “map” with vertices representing the line convergence points and edges representing the connection lines. A network topology matrix “D” is formed by detecting the current directions as the current directions are altered following a fault, whereas the current directions at the ends of non-fault lines remain the same. In order to prevent misjudgment problems arising from power flow reversal, the rates of change of the fault currents are used to further determine whether a fault has occurred and the “double D method” is introduced to identify the fault type. Simulations results with different fault types verify the effectiveness and reliability of the proposed method.