In view of the existing problems of confined water risk analysis of deep foundation pit, such as single analysis method, insufficient risk prior assessment theory and lack of dynamic risk prediction during the construction process, this paper proposed a confined water risk analysis method of deep foundation pit based on Bayesian network (BN) to realize the pre-analysis of confined water risk accidents and the dynamic risk evaluation of the whole construction process. From aspects of environment, design, construction and management, the method established a risk evaluation index system of confined water of deep foundation pit, constructed static BN risk analysis model, and completed the risk probability prediction, accident factor diagnosis, disaster factor identification and other risk assessment contents in advance. On this basis, the dynamic risk analysis of confined water in the whole construction process based on monitoring data was realized by setting risk transfer nodes and observation nodes and introducing Noisy-Max hypothesis. Finally, taking the north deep foundation pit of Jingjiang Yangtze River tunnel project in Jiangyin as an example, the risk level of confined water in the project was analyzed, the related risk factors were further defined, and the dynamic risk change was accurately predicted. Results show that the proposed method has high applicability and can provide practical guidance for the safety of deep foundation pit construction.