دانلود رایگان مقاله روش بهینه سازی سیستماتیک برای شبکه مبدل حرارتی و فرایند طراحی به طور همزمان

Abstract

Distillation units require huge amounts of energy for the separation of the multicomponent mixtures involved in refineries and petrochemical industries. The overall efficiency of the distillation column system is determined from the trade-offs of the Operating Expenditures (OPEX) and Capital investment cost (CAPEX), as there is a strong interaction between the distillation columns and the Heat Exchanger Network (HEN) of the interconnecting streams. In this paper, a systematic Mixed Integer Non-Linear Programming (MINLP) optimization methodology for process integration of distillation column complex is presented. The highly nonlinear rigorous models of the distillation column and phase change are being substituted with simple surrogate models that generate operating responses with adequate accuracy. The methodology is applied on two case studies of the aromatics separation PARAMAX complex. The results illustrate significant reductions on the Total Annualized Cost. With a scope limited to the benzene and toluene columns, the gain reaches about 15%.

6. Conclusions

This paper has presented a general methodology for Process integration optimization that simultaneously takes into account the CAPEX-OPEX trade-offs. The methodology finds the optimum process design and operating conditions in addition to the HEN con- figuration of the interconnecting streams of the units involved. For the HEN synthesis, a continuous monotonic model is included in the formulation in order to accurately take into account the phase changes that might occur during heat exchange between streams. The sequential solution strategy of the methodology benefits from the capacities in differentfields ofthree softwares and reduces the overall computational time needed in large scale optimization problems. The surrogate models efficiently replace the computational demanding rigorous models of the commercial simulators and curries outthe optimization procedure in one software thattargets the solution directly without the computationally costly data interchange procedure that usually takes place when an interface of multiple softwares is considered. For the optimization procedure, the only information required is the surrogate models parameters and the input-output specifications of the process for a given set of initial values of the optimization variables. Following this, the interaction of the process units, represented by the surrogate models, with the HEN of the interconnecting streams enables the formulation to treat the overall process as a ‘black box’ where neither prior heuristics nor engineering experience is needed in order to generate the optimum solution. The dependence ofthe proposed MINLP formulation on initial conditions is addressed with a multistart approach that is used as a screening technique for the identification of the most promising initial values andhas a significantimpact onreducing the overall computational time needed. The methodology is applicable to both grassroot design, when all degrees of freedom are optimized, and retrofit design in the case where a subset of the design variables is fixed to meet the requirements of the application of interest. Finally, the methodology was applied on a distillation column complex with two case studies. Three different objectives were examined for the optimization problem on the first case study to illustrate that the simultaneous consideration of the CAPEX-OPEX trade-off is the best strategy for minimizing the overall cost. Both CAPEX and OPEX sub-objectives are lower in the grassroot case when compared to a retrofit case. The gain reaches 15% between the study sequential approach (operating conditions optimization, then column design with engineering design rules, and finally exchanger network optimization) and grassroots design (simultaneous optimization of all variables).