The invention relates to an islanding detection key feature extraction method based on multi-resolution singular spectrum entropy and empirical mode decomposition. According to the characteristic thatsignal features are not easy to discover in an island and non-island interference condition, deep feature analysis is carried out on the signals. Common-point voltage signals are subjected to multi-resolution empirical mode decomposition. The decomposed signals are used for phase space reconstruction, singular spectrum decomposition is carried out on each layer of phase space matrix to obtain aneach-layer singular value, and each-layer phase space singular spectrum entropy is extracted according to the each-layer singular value. After combination of multiple layers of singular spectrum entropy, a feature vector of the multi-resolution singular spectrum entropy under the signal is constructed. In the method, the multi-resolution empirical mode decomposition realizes layered and meticulousanalysis on the signal spectrum, each-layer singular spectrum entropy carries out unified measurement for characterizing the unique feature of each layer of the signal, the combined feature vector quantitatively describes the essential features of the signal, stability is achieved for the same kind of signals and the same sample, and the method is applicable to islanding detection key feature extraction.