With increasing scale of usage for renewable energy resource, DC micro-grids system becomes more and more popular for the application of different kinds of power electronics in DC. However DC arc fault introduces major safety concerns due to the randomness and instability, especially absence of zero current crossing, which makes it is hard to be detected and eliminated. In this thesis, a novel proposed series dc arc detection methods were validated both in theoretical simulation by MATLAB and in real-time experiments by creating real arcing in the constructed equivalent DC system. By changing the parameters and type of both the circuit and the setting of detection method, the arc detection performs with sensitivity, selectivity, speed and reliability harvested. Additionally, the research and practical tests have also been conducted to study the characteristics of series dc arc, which contributes to the determination of the appropriate setting value for the arc detection algorithm.