Aiming at the scarcity of observational data for extreme storm surge events in the Bohai Sea and the inherent limitations of traditional forecasting methods in terms of physical mechanism constraints and probabilistic result interpretation, this study proposes an integrated framework for storm surge prediction and uncertainty quantification combining numerical simulation and deep learning. A hybrid wind field forcing scheme is constructed to compensate for the lack of observational data via numerical models, on the basis of which various machine learning models are developed to perform deterministic storm surge forecasting and reduce deterministic prediction errors. For uncertainty quantification, a multi-model fusion probabilistic prediction system is established by introducing bidirectional long short-term memory networks, adaptive bandwidth kernel density estimation, and sequential forward feature selection.This study effectively improves the forecasting accuracy of storm surges in the Bohai Sea through the hybrid forcing scheme, and simultaneously realizes the quantitative assessment of forecasting risks, providing important theoretical support and decision-making basis for the early warning and prevention of marine disasters, the scientific management of coastal zone resources, and the protection of coastal ecosystems.