Abstract:This paper takes Hebei province as the research area, using wavelet transform and RBF neural network methods to carry out GNSS-PWV prediction research. Firstly, wavelet decomposition is performed on the PWV sequence of GNSS stations, and then the high and low frequency signals decomposed by the wavelet are predicted by the RBF neural network. Finally, the appropriate high frequency and low frequency signals are selected through experiments to reconstruct the GNSS-PWV prediction values. Compared with the actual measured GNSS-PWV values and RBF predicted PWV values, we find that the accuracy of GNSS-PWV predicted based on wavelet transform and RBF neural network is higher than that of the RBF neural network, and the accuracy of the prediction results decreases with the increase of the prediction time.