To solve the problem that Ensemble Empirical Mode Decomposition (EEMD) is not insufficient in processing the non-linear and non-stationary signal. On this basis, this paper presents a new method of power quality disturbance detection based on Complete Ensemble Empirical Mode Decomposition (CEEMD). Firstly, the power quality disturbance signal with noise is decomposed into Intrinsic Mode Function (IMF) by CEEMD, whose instantaneous amplitude and phase characteristic are detected by Hilbert Transform. Furthermore, the obtained instantaneous amplitude is derived to get its modulus maximum points by second-order derivative, which improves the accuracy of positioning disturbance time. Given the high frequency compound disturbance, the CEEMD decomposition method is adopted twice to remove noise and false components. According to steady state disturbance, the detection method is to remove harmonic components firstly and then to extract the flicker envelope. At last, through Matlab simulation experiments and the data measured on AC voltage regulator load experiment and three-level experimental platform verify that this method is able to identify the unknown disturbance signal, and as well as determine the power quality disturbance time, type, frequency, amplitude and other characteristic parameters. This work is supported by National Natural Science Foundation of China (No. 61374043), Fundamental Research Funds for the Central Universities of Ministry of Education (No. 2013QNA50), and Natural Science Foundation of Jiangsu province (No. BK20130205).