Abstract: In order to enhance the security protection capability of power information networks, a dynamic assessment problem for multi-source threats is studied, and a situation assessment model based on Bayesian networks is constructed. By deeply analyzing the threat characteristics of the physical layer and information layer in the power information network, combined with the probabilistic inference mechanism of Bayesian networks, a three-layer evaluation framework including the threat source layer, state evaluation layer, and comprehensive situation layer was designed. The dependency relationship and dynamic evolution law between threat factors were quantified, and a solution strategy based on a combination of precise inference and approximate inference was proposed. The research results indicate that the model can effectively identify the cumulative effects of multi-source threats such as equipment failures and network attacks, providing a scientific basis for power system security decision-making, significantly improving the accuracy and real-time perception of threat situations, and laying a theoretical foundation for ensuring the stable operation of the power grid.