Refrigeration system synthesis based on de-redundant model by particle swarm optimization algorithm

Abstract

Simultaneous optimization of refrigeration system (RS) and its heat exchanger network (HEN) leads to a large-scale non-convex mixed-integer non-linear programming (MINLP) problem. Conventionally, researchers usually adopted simplifications to confine problem scale from being too large at the cost of reducing solution space. This study established an optimization framework for the simultaneous optimization of RS and HEN. Firstly, A more comprehensive and compact model was developed to guarantee a relatively complete solution space while reducing model scale as well as its solving difficulty. In this model, a tandem arrangement of connecting sub-coolers and expansion valves was considered in the superstructure; and the pressure/temperature levels were optimized as continuous variables. On this basis, we proposed a “two-step transformation method” to equivalently transform the cross-level structure into a non-cross-level structure, and the de-redundant superstructure was established with ensuring comprehensiveness and rigor. Furthermore, the MINLP model was developed and solved by Particle Swarm Optimization algorithm. Finally, our methodology was validated to get better optimal results with less CPU time in two case studies, an ethylene RS in an existing plant and a reported propylene RS.