Rising sea levels and an increasing frequency of storm surges due to climate change have intensified disaster risks in coastal areas. This study aims to develop a storm surge prediction algorithm using deep learning techniques based on the Global Forecast System (GFS) of the NOAA and the Japan Meteorological Agency's Meso-Scale Model (JMA-MSM) data and compare the predictive performance of the two models using their atmospheric data as input variables. A model integrating Convolutional Neural Networks, Long Short-Term Memory, and Attention mechanisms was designed, with tidal observation data used for training. The models were validated using four major typhoon cases that directly impacted the Korean Peninsula. The JMA-MSM-based model outperformed the GFS-based model, improving the average RMSEs by 0.34 cm, 0.73 cm, and 1.86 cm, and MAPEs by 0.15%, 0.36%, and 0.68% for the West, South, and East Sea regions, respectively. This improvement is attributed to the higher spatial resolution of JMA-MSM, which better captures localized meteorological changes. This study aims to enhance the efficiency of storm surge predictions for coastal disaster preparedness and provide a foundation for future research utilizing additional meteorological data.