The present invention provides a low current grounding fault line selection method, comprising: using wavelet packet transformation and Fourier transformation to perform feature parameter extraction on a zero-sequence current signal; using a fuzzy self-correcting algorithm to optimize a support vector machine model; and performing multi-source data fusion, to complete grounding fault line selection. According to the method, a problem that judging of multi-source data fusion fails to adapt to various complex faults is resolved. The wavelet packet transformation and the Fourier transformation are used to perform feature parameter extraction on the zero-sequence current signal, and then the fuzzy self-correcting algorithm is used to optimize the support vector machine model, so that adaptation to complex fault conditions is improved, and network precision and a convergence rate are improved, thereby effectively resolving a problem in a line selection method and improving line selection precision and speed.