To evaluate the safety performance of the high-rise structure， GNSS-RTK technology is applied for the dynamic moni‐toring of Tianjin radio and television tower. The monitoring errors and noise are analyzed. An ACVMD-WT algorithm is proposed based on variational modal decomposition （VMD） and wavelet threshold （WT）. The simulation signal verifies that VMD has superior signal reorganization ability than the complete ensemble empirical mode decomposition with adaptive noise （CEEMDAN）.The algorithm adopted is used to denoise the horizontal signals and obtain the structural modal parameters. The results show that the water environment is prone to generate background noise， which adversely affects the monitoring result. The information components are selected effectively by ACVMD-WT mixed filter， and the signals before and after noise reduction have a strong correlation. In contrast to the ensemble empirical mode decomposition （EEMD）， CEEMDAN， and CEEMDAN-WT， the method used has a better noise reduction effect. The first three-order frequencies and damping information of the structure are better captured from the noise reduction signal， the relative error of the first-order frequency with the finite element analysis is 0.189%， and the maximum value of the third-order frequency relative error is only 4.054%， improving the recognition and accuracy of the modal frequency.