Arc fault, especially series arc fault, has been the primary cause of electrical fire, which is intermittent, hidden, random and uncertain, difficult to detect, and serious harm. The characteristics of arc fault are firstly analyzed, including the division of arc fault and its detection difficulties. Secondly, the methods for series arc fault detection at home and abroad are explained, which are divided into detection methods based on the physical phenomenon of arc fault, detection methods based on arc models and simulation technology, and detection methods based on time-frequency characteristics, voltage and current. Among them, the focus is on the recent hot new deep learning methods, introducing AC and DC arc fault detection technologies respectively. Finally, from the perspective of practical application, the problems that need to be solved urgently in the detection scheme of series arc fault are pointed out, and the research and optimization direction of series arc fault detection in the future are prospected.