In order to solve the problem of inaccurate fault line selection in small current resonant grounding system, a method for line selection based on adaptive multi-scale morphological filtering and complementary ensemble empirical mode decomposition (CEEMD) is proposed. The adaptive morphological filtering method can effectively suppress the wideband white noise and isolated impulse noise associated with the power system fault process. After denoising, the line selection is conducted by comparing the first-order differential polarity and phase of the characteristic components which exist in the faulty line and the non-faulty line through the CEEMD decomposition and Hilbert-Huang transform (HHT) transform. Compared with empirical mode decomposition (EMD) and ensemble empirical mode decomposition (EEMD), CEEMD can reduce the influence of modal aliasing and the decomposition error better. In the MATLAB simulation software, different fault location, transition resistance and fault initial angle are analyzed, and the methods above were verified. A lot of simulation results show that the method can’t be influenced by initial fault angle, fault location and transition resistance.