Abstract: Electric load forecasting is concerned with power deployment and system operation. For the short-term load forecasting, this paper proposes a combined forecasting model based on the complementary ensemble empirical mode decomposition (CEEMD), which integrates the advantages of the classical and machine learning approaches. Firstly, the original data are denoised by the singular value decomposition (SVD), and the CEEMD is performed on the noise reduced sequence to obtain the intrinsic mode functions (IMFs) and the trend components (RES) of different frequencies. The independent component analysis (ICA) is adopted on the high frequency IMF₁, and the remaining intrinsic mode functions are reconstructed to obtain the IMF_cg. Different methods are applied to predict the IMF₁, IMF_cg, and RES respectively, and the predicted values of the IMF₁, IMF_cg, and RES are summed up as the final predicted value. According to the experimental data, the proposed method is proved to achieve better prediction by making full use of the intrinsic characteristics of the explored load data, a good reference for the short-term load forecasting.