Empirical mode decomposition (EMD) is an important part of Hilbert-Huang transform (HHT), in order to overcome its modal aliasing and endpoint effect in harmonic detection, a novel harmonic detection method combing adaptive noise complete integration of empirical mode decomposition (CEEMDAN) and Hilbert transform (HT) is proposed in this paper. This method firstly compares and analyzes EMD, EEMD and CEEMDAN algorithm in theory, and studies the characteristics of CEEMDAN algorithm. Then, CEEMDAN algorithm is adopted to decompose the original signals to obtain the intrinsic modal function (IMF). Finally, the HT algorithm is used to analyze the IMF component of each order, and the instantaneous amplitude-frequency information contained in the harmonic is detected. The simulation results show that compared with the signal processing capability of HHT algorithm, the proposed method overcomes the disadvantages of EMD algorithm in harmonic detection and improves the accuracy of signal decomposition.