In order to deal with the adaptive shortcoming that the existing time-frequency analysis methods are not insufficient in processing non-linear and non-stationary signal, an Adaptively Complementary Ensemble Empirical Mode Decomposition (ACEEMD) method is proposed. In the proposed method, through the study of the added-noise principle of noise assisted decomposition method, where the two parameters of the relative RMSE and the signal-to-noise are introduced as added-noise evaluation indexes to adaptively determine the optimal amplitude of added-noise and the number of ensemble trials respectively. And the additive noise are added in the form of positive and negative pairs to the targeted signal, which overcomes the disadvantages of the mode mixing problem, the endpoint effect and the large residual noise existing in original decomposition method. Finally, the modified method combined with Hilbert transform is utilized in the power quality disturbance detection. Simulation experiments verify that the developed method is not only able to effectively extract disturbance frequency, amplitude and other characteristic parameters, but also accurately locate disturbances time, which provides a novel method for power quality detection and analysis. This work is supported by Natural Science Foundation of Jiangsu Universities (No. 15KJB510033).