The current waveform of a series arc fault is greatly affected by the types of load, making it difficult to construct a universal criterion for fault detection using current characteristics. To identify the arc voltage at the fault point, a series arc fault detection method based on voltage characteristic energy is proposed. First, the selection of characteristic frequency bands for fault information is demonstrated by analyzing the characteristics of arc voltage at the fault point and fault voltage at the monitoring point. Then, based on the classification of arc voltage waveform characteristics under different loads, a fault detection method is proposed based on the whole domain energy amplitude and sensitive phase domain energy phase information of the voltage characteristic frequency band. Finally, the construction of a comprehensive fault detection strategy is achieved through the total energy amplitude of the whole domain and the statistical ratio of phase mapping of the sensitive domain. The experimental results show that the fault detection accuracy of the method exceeds 98% with different line parameters and test loads, and with no misdetections, verifying the effectiveness of the method.