Abstract: Existing TBM parameter point prediction models for stable excavation sections ignore the uncertainty errors during the prediction process and fail to describe the confidence level of the prediction results. This paper proposes a TBM parameter interval prediction model for stable excavation sections based on the Bootstrap-COA-BiGRU algorithm. First, the COA algorithm is used to optimize the hyperparameters of the BiGRU neural network, allowing the model to better autonomously learn the complex nonlinear relationship of the rock-machine interaction in the time and feature dimensions of the TBM data in ascending phase , effectively improving the model's prediction accuracy.Secondly, by analyzing the results of point prediction models, the interval prediction method is introduced to quantify the uncertainty of the model and random uncertainty in the data, obtaining high-quality parameter prediction intervals in TBM stable excavation phase. Finally, the proposed model is applied to the Xinjiang YEGS project for interval prediction of TBM parameters under class Ⅱ~Ⅳ surrounding rock conditions, and the results are compared with BP, GRU, BiGRU, and COA-GRU models to verify the superiority and practicality of the proposed model, promoting the development of TBM intelligent construction.