A Stability Analysis Method for Surrounding Rock of Underground Powerhouses in Pumped-Storage Power Stations Based on the Fusion of SoVMD and Transformer

A hybrid analysis model based on self-optimizing variational modal decomposition ( SoVMD ) and recurrent reconstructed transformer ( RRTransformer ) is proposed for the monitoring and analysis of surrounding rock stability of underground powerhouse of pumped storage power station. Firstly,a reasonable monitoring scheme is designed to collect the deformation and displacement data of rock mass which can effectively characterize the stability of surrounding rock. Secondly, considering the nonlinear and fluctuating characteristics of the deformation displacement sequence, SoVMD is used to decompose it and extract multi-scale frequency components ( MsFCs ). Finally, based on the sliding modeling mechanism, MsFCs are used as the training input of RRTransformer, and the optimal deformation prediction analysis results are obtained through parameter optimization and model reconstruction. The stability analysis process of surrounding rock of an underground powerhouse of a domestic pumped storage power station is selected as a case. The results show that the designed monitoring scheme can effectively obtain the deformation and displacement data of rock mass, and the proposed prediction analysis model has excellent performance. The prediction analysis results are of great significance to the safety management of surrounding rock of underground powerhouse.