Abstract: The GNSS technology, as an important tool for mining subsidence monitoring, is significantly affected by the noise present in its time series. This paper proposes a denoising method that combines an Improved hybrid grey wolf particle swarm optimization (IPSOGWO) and an improved complete ensemble empirical mode decomposition with adaptive noise (ICEEMDAN), coupled with wavelet thresholding (WT). The IPSOGWO optimizes the hyperparameters of the ICEEMDAN algorithm to decompose the GNSS time series and extract the intrinsic mode functions (IMF). The multi-scale permutation entropy is used to select the IMF components containing noise. These components are then secondarily processed using wavelet thresholding and reconstructed with the remaining IMF components to obtain the denoised results. Experiments with simulated signals and actual data from an automated monitoring station in a mining area demonstrate that the proposed method outperforms the wavelet threshold, complete ensemble empirical mode decomposition (CEEMD), and GWO-ICEEMDAN in terms of denoising performance, providing reliable data for subsequent analysis of working face subsidence.