Abstract: In order to dynamically study the failure risk of LNG storage tanks, the bow-tie model and the Bayesian model are firstly constructed and utilized to identify the key factors leading to leakage and predict the probability of occurrence of each main event, i.e. the leakage and accident; furthermore, the probability of occurrence of each main event is dynamically updated according to the history records of abnormal events; finally, the wind speed correction model based on acyclic flow around cylinder is utilized to simulate the high frequency and risk accidents. Through applying the above methods to the LNG storage tank of a receiving station, the results show that among the 73 basic factors, stress corrosion of auxiliary pipeline, poor contact between valve body and bonnet, external overpressure, typhoon and partly human and management factors are the main causes of leakage of LNG storage tanks; considering the polymorphism and relativity of events, the probability of leakage increased from 8.42×10^-3 to 92.8×10^-3 and accidents increased by 111. 06% yearly according to the history data; of all accidents, jet fire and flash fire are most likely to occur with a maximum impact of 93.7 m and 436.0 m.