Abstract: Based on stochastic optimization strategy, a formulation methodology was proposed for optimal synthesis of thermally coupled complex distillation column systems comprising simple columns, complex columns with side rectifier and/or side stripper as well as partially or fully thermally coupled (Petlyuk) columns with prefractionators.To cope with the uncertainty of the number of column sections and thermally coupled streams for a given multi-component mixture separation problem, a decomposition strategy was proposed so that the original problem could be divided into a series of subproblems with different numbers of column sections.To represent and manipulate the flowsheet structure in the optimization, a coding procedure was developed.The procedure was implemented by tearing the representation into two strings of code: separation sequence and thermal-coupling scheme.The flowsheet structure encoding was performed by adopting a binary sort tree approach that is originally used in data structure theory, so that the problem formulation and solution becomes tractable.Finally, taking the recoveries of key components in the pretractionators and reflux ratios as continuous variables, a conceptual mixed-integer nonlinear programming (MINLP) problem was formulated.The solution to the problem could be achieved with an improved simulated annealing algorithm.