Feature modeling for complex industrial processes is the basis for studying their optimal control. Complex industrial processes generally have uncertain characteristics such as strong interference, nonlinearity, and large time-varying. Some of the processes involve complex biochemical reactions with strong contamination and high risk, and the detection data is highly dimensional and noisy, which all put forward more urgent needs and higher standards for the establishment of accurate industrial models. This paper summarizes the current modeling ideas and research progress of complex industrial processes, and aims to analyze the applicability and effectiveness of different modeling methods from multiple perspectives, so as to lay the modeling foundation for the advanced optimal control theory to guide the actual industrial production. First, the main industrial modeling methods are divided and summarized from three aspects: mechanism modeling, data-driven modeling and hybrid modeling. Second, the specific design ideas of various modeling methods are described, and the model structure and algorithm characteristics are analyzed. Then, the specific applications of different modeling strategies in solving the problems of index modeling, controlled object modeling, and full-scale modeling in the actual industrial processes are investigated. Finally, combined with the current trend of industrial intelligent construction and its challenging problems, the future research ideas and development directions are pointed out.