In the face of the increasing global challenge of climate change, accurate prediction of carbon emissions is crucial for the development of effective mitigation strategies. To this end, this paper proposes a novel carbon emission prediction model that integrates the Improved Adaptive Empirical Modal Decomposition (ICEEMDAN) and the Improved Marine Predator Algorithm (IMPA). ICEEMDAN can efficiently decompose the raw carbon emission data into a series of easier-to-handle IMFs, whereas the IMPA, through an intelligent optimization mechanism, finely tunes the parameters of the BiLSTM network to achieve the optimal prediction effect. The results show that the proposed model exhibits excellent performance in five indicators, including MSE, and provides a powerful decision support tool for future carbon emission management.