High winds are external hazards that can affect the safety of nuclear power plants (NPPs), primarily through wind-borne debris and loss of offsite power. While Probabilistic Safety Assessment (PSA) methods based on Event Tree (ET) and Fault Tree (FT) models are widely used to evaluate such risks, they face inherent limitations in representing complex component dependencies and cascading failure effects. This study utilizes a Bayesian network (BN)-based approach for high-wind PSA. First, we develop an ET-FT scenario of high-wind-induced NPP. Then, it involves converting ET-FT models into a BN to enable probabilistic inference within a unified framework. Furthermore, to address the limitations of static BN models, this research introduces an intensity-based Dynamic Bayesian Network (DBN) in which system states evolve as a function of hazard intensity. This approach effectively captures component dependencies, accounts for cascading effects such as debris impact and power loss, supports data-driven probability updates, and enables system-level risk assessment across varying hazard intensities. The proposed DBN method is applied to a Korean OPR1000 reactor unit to demonstrate its applicability. The analysis results identify cascading failure scenarios and system-level risk transitions with varying hazard intensity, insights that are difficult to obtain with conventional static ET-FT models.