In order to deal with the shortcoming that Empirical Mode Decomposition (EMD) is not insufficient in processing non-linear and non-stationary signals, an adaptive local mean EMD (ALMEMD) method is proposed in this paper. In the proposed method, the influence of how the distribution characteristics of signal extrema as well as the mean curve fitting affect decomposition results is analyzed, and only one spline interpolation is utilized to construct local mean curves, in which all of the data located between the zeros and extrema are employed as local characteristic time-scale by using mean value theorem for integrals. Furthermore, aiming at signal multicomponent feature, the advantages that low-frequency components and high-frequency components are respectively sifted out by local mean curves derived by higher-order extrema and low-order extrema, and orthogonality coefficient is taken as an evaluation index to adaptively select the optimal extreme order. Both simulation test and experiments data measured from the built power quality disturbance platform demonstrate the feasibility and effectiveness of the proposed method, which is suitable for applying in detection and analysis of harmonics as well. This work is supported by National Natural Science Foundation of China (No. 61374043).