Abstract: With the large-scale grid connection of new energy sources and the widespread application of microgrids and distributed power sources, the uncertainty of AC/DC hybrid power grids has significantly increased, making traditional deterministic power flow analysis insufficient to meet the needs of system risk assessment. Addressing the problems of low computational efficiency, insufficient accuracy, and poor risk assessment accuracy of existing probabilistic power flow algorithms, this paper proposes a fault risk assessment and prediction technique for AC/DC systems based on the maximization of fractional moments and information entropy. An optimized model is constructed to minimize the difference between the fractional moments of random input variables and their estimated values ??to generate high-quality sample points. The statistical characteristics of random output variables are calculated, and the probability density function is reconstructed using the information entropy maximization criterion to quantify the probability of exceeding limits and accurately determine the system risk state.