Using the meteorological observations from coastal observing stations in Guangxi from 1 January 2016 to 31 December 2021 and 0.25°×0.25° ERA5 data with a 3-h interval, this study calculates and analyzes the required physical quantity values based on the physical processes of sea fog formation and dissipation as well as forecasters’ experience. A total of 68 physical quantities that may affect sea fog are preliminarily selected as initial forecasting factors, and an objective forecasting model for the sea fog in the Beibu Gulf is established using the ingredient-based method, decision tree method and index method. Verification is conducted using the threat score (TS) for the forecasts of the sea fog processes from January to April 2023. The results are as follows. (1) The three objective forecasting products perform well in predicting the sea fog in the Beibu Gulf, but the performance in the Beihai sea area is inferior to that in other sea areas. (2) Based on the TS of the 24-h forecasts using the three methods, the forecasting performance of the ingredient-based method is significantly better than that of the decision tree method and the index method. (3) Based on the TS of the 48-h and 72-h forecasts using the three methods, the decision tree method outperforms the ingredient-based method and the index method significantly. (4) A comparative analysis of the TS of the ECMWF (European Centre for Medium-Range Weather Forecasts) model products reveals that the average TS (only around 0.30) of the ECMWF model products for the coastal sea fog in the Beibu Gulf is lower than the TSs of the three objective methods, and notably lower than the average TS (0.40) of the ingredient-based method and the decision tree method. Nevertheless, from the perspective of the forecasting results across different sea areas, the ECMWF model products exhibit significantly better performance (with the TS of 0.46) in the Fangchenggang sea area compared to the ingredient-based method (with the TS of 0.38) and the index method (with the TS of 0.36).