Abstract: Addressing the challenge of predicting sound speed profiles in oceanic internal waves environments, this study introduces a hybrid time series forecasting method that integrates Prophet and Long Short-Term Memory (LSTM) models. Initially, the Prophet model is utilized to detect and correct anomalies caused by internal waves, thereby mitigating the impact of irregular disturbances on the prediction. Subsequently, the LSTM model is employed to forecast the time series of the corrected data, enabling accurate reconstruction of the sound speed profiles. Experimental results demonstrate that this combined approach effectively reduces internal waves interference and enhances prediction accuracy, providing novel insights and technical solutions for sound speed profile forecasting in complex marine environments.