In this paper, an intelligent interaction prediction method (IIPM) for two-level multiobjective optimization of large-scale steady-state industrial processes is proposed based on the structural model of hierarchical intelligent steady-state control systems, and its convergency theorem is given and proved. The outstanding characteristic of IIPM is that an intelligent inference and decision unit ID is introduced in the structure of algorithm, thus its efficiency and practicality are improved. In the end, the effectiveness of the algorithm is illustrated by the simulation of typical industrial process problem.