In the wake of the stronger and stronger development of carbon market, the carbon price fluctuation has drawn the attention of researchers, encouraging numerous researchers involved in the carbon price study. Owing to the strongly nonstationary and nonlinear characteristics of carbon price, most of existing approaches failed to forecast the carbon price perfectly. In our study, a novel hybrid forecasting model is presented to forecast the carbon price. Variational mode decomposition (VMD) and independent component analysis (ICA) are utilized to preprocess the chosen data for getting the independent components. Then the independent components are trained by radial basis function neural network (RBFNN) to predict them respectively. Finally, the forecasting result is obtained by linear combination. In addition, the numerical results show that the VMD-ICA-RBFNN model outperforms wavelet-based NN, VMD-RBFNN, EMD-ICA-RBFNN, RBFNN, ARIMA-GARCH and ARIMA models.