文摘
A nested method combining tabu search (TS) and sequential quadratic programming (SQP) for the mixed-integer nonlinear programming (MINLP) is proposed. It involves MINLP in a nested inner and outer loop,where TS in the outer loop involves the integer variables, and SQP in the inner loop solves the nonlinearprogramming (NLP) with integers specified from the outer loop. Good performance has been demonstratedby several mathematic examples and a pump system example, but problems with regard to heat exchangernetwork synthesis (HENS) are still encountered. Further analysis shows that even given the optimal structureof a heat exchanger network (HEN), it is difficult to solve the inner loop NLP. Based on the Yee and Grossmannmodel, an adaptive model reformulation method is later proposed, to bridge the inner loop and the outerloop. Given the integer variables by the outer loop, it automatically simplifies the NLP model, in regard toboth dimension and complexity, before delivering it to the inner loop SQP method. The results are presentedand discussed through examples.