Nested Method For Mixed-integer Nonlinear Programming In Process Synthesis

The efficient solving method of mixed-integer nonlinear programming (MINLP) is of great importance to the process synthesis. Studies upon MINLP solving method for process synthesis are conducted in this paper. The main contributions are summarized as follows:(1) A three layers nested method is proposed for superstructure based MINLP in process synthesis (PS). It combines both the merits of heuristic methods and deterministic algorithms. Moreover, an adaptive model reformulation mechanism is proposed based on the special characteristics of MINLP in process synthesis. Heuristic methods, such as tabu search (TS), genetic algorithm (GA), and simulated annealing (SA) deal with the integer variables in the outer loop. Given the integer variables by the outer loop, the original MINLP degrades to a NLP, which is further solved by deterministic methods, like SQP. But before delivering the outer loop result into the inner loop, an adaptive model reformulation mechamismn is proposed to automatically ameliorate the NLP model by removing its redundance. This adaptive model reformulation module, based on the mechanism of process synthesis, bridges the inner and outer loops and greatly improves the global optimality of the method.(2) Applications of the nested method to heat exchanger network synthesis (HENS) are conducted. The results have demonstrated this method is efficiency for HENS.(3) A superstructure-based MINLP model is proposed for water network synthesis (WNS). The model combines water using processes and water treatment operations into a single network. This model is universal, and easy to expand with the least number of continuous variables. Good performande of the nested method applied to WNS is also achieved.