To address the difficulty in accurately predicting the settlement status of transmission line towers, this paper proposes a settlement prediction method for transmission tower based on Complementary Ensemble Empirical Mode Decomposition (CE-EMD). By constructing subcomponent-based settlement prediction models, this method addresses the problem of unstable prediction residuals leading to low prediction accuracy. It achieves accurate prediction of the settlement status of transmission line towers. Case studies show that the EMD model performs well in the decomposition process, with a comprehensive evaluation index ranging from 1.761 to 1.8 with a small variation. The combined Denoising Autoencoder (DAE)- Long Short Term Memory (LSTM) model exhibits a relative error of only 1.45% in prediction, indicating high prediction accuracy. It can accurately forecast the development trend of transmission line tower settlements, which is of great significance in preventing major power outages.