As random walk algorithm with compulsive evolution is used to optimize heat exchange network, the structure with lower annual cost in continuous variable evolution stage is destroyed by integer variable evolution operation. As a result, the structure formed after the evolution of continuous variables can not play its full advantages, which restricts the evolution of continuous variables. We propose an optimization strategy of separating continuous variables and integers to protect the effective structure, thereby, protecting structures with development potential and reducing the combined annual cost further. In example study our result is better than that in literature. Precision of the algorithm is improved, and validity of the strategy is verified.

As node-wise non-structural superstructure model is applied to heat exchanger network optimization, heat exchanger crossover may appear. It is found that crossover structure causes an increase of total heat transfer area for heat exchange unit with same heat loads. At the same time, it reduces obviously computational efficiency of the algorithm. A vertical node-wise non-structural superstructure model is established. Compared with node-wise non-structural superstructure, the vertical node-wise non-structural superstructure model reduces computational complexity in the optimization process and eliminates adverse effects of crossover structure, which improves efficiency and accuracy of optimization. Finally, optimization performance of RWCE algorithm based on vertical non-structural superstructure model is analyzed in two examples. The results are superior to those in literature.