Abstract
To address the complex deformation of wellbores influenced by surrounding coal mining operations, this study employed an improved modified least-squares ambiguity decorrelation (MLAMBDA) algorithm based on the double-difference model for high-frequency dynamic computation of Bei Dou System and Global Navigation Satellite System (BDS/GNSS) observation data. A quantitative analysis was conducted on the performance of various combinations of BDS/GNSS in wellbore deformation monitoring, and the effects of different baseline lengths on the monitoring results were evaluated. Based on the high-precision deformation monitoring sequences, an intelligent early warning model for wellbore deformation was established using the deep learning Bi-LSTM algorithm. The results indicate that the monitoring accuracy of the BDS/GNSS multi-system combination in the E, N and U directions is within 2 mm, with all three directions outperforming the results obtained from a single Global Position System (GPS) system. As the baseline length increased from 1 km to 6 km, the accuracy in the E, N, and U directions decreased by 15.8%, 16.0%, and 5.6%, respectively. Within a 6 km range, the horizontal accuracy remains better than 3 mm, while the vertical accuracy is better than 6 mm, meeting the requirements for wellbore deformation monitoring. The early warning model can flexibly adapt to the deformation conditions at different sites and the various disturbances encountered, effectively capturing the complex nonlinear time-varying characteristics of the observation time series. The prediction of future results for one month based on one year of observation sequences achieves an accuracy better than mm, providing a safeguard for safe production in mines. This research method can also be extended to use BDS/GNSS for hourly level high-precision deformation monitoring and early warning of major engineering infrastructure such as bridges, dams, and high-speed railway systems.