Aiming at the non-linear and non-stationary characteristics of power system signals, the adaptive decomposition method has efficient and simple advantages in processing signals.However,there is a shortcoming that the selection of additive noise parameters in such methods are required to be obtained artificially.To counter the above-mentioned shortage,an adaptively complementary LMD(AELMD) method was proposed.In the proposed method,the added-noise principle of noise assisted decomposition method is analyzed,and the two parameters(Standard Deviation and Signal-to-Noise) are taken as an evaluation index of adding noise to adaptively determine the optimal amplitude of additive white-noise and number of ensemble trials.And the additive noises are added in the form of positive and negative pairs to the targeted signal,which inhibits residual noise in the decomposition process.And then,the feature extraction steps in standard LMD method are modified,the decomposed product functions are extracted the instantaneous characteristic parameters by Hilbert transformation,which eliminates the limitation of the extraction range of pure frequency modulation function under step component interference.In the last, the modified method is utilized in the harmonic detection,through simulation experiments verify that this method is not only able to effectively extract harmonics characteristic parameters,but also accurately locate disturbances time, and has a certain anti-noise performance as well.