مدل بهینه سازی شبیه سازی برای یک فرایند ساده LNG

A gradient free optimization-simulation method for processes modelled with the simulator Aspen HYSYS is developed. The tool is based on a Tabu Search (TS) and the Nelder-Mead Downhill Simplex (NMDS) method. The local optima that result from the TS are fine-tuned with NMDS to reduce the required number of simulations. The tool has been applied to find the total refrigerant flow rate, composition and the refrigerant suction and condenser pressures that minimize the energy requirements of a Prico process. The main strength of this method is that it has a high probability of obtaining a better solution with significantly fewer simulation runs than other metaheuristic methods. Also, by changing the TS step size it is possible to influence the initial search pattern, thereby taking advantage of already gained process knowledge to decrease the optimization time. The method is general and can be applied to other processes modelled in Aspen HYSYS.

Energy and petrochemical process plants consist of unit operations such as separators, valves, expanders, compressors and heat exchangers. Each of these unit operations contributes its own set of more or less realistic thermodynamic equations as well as mass and heat balances. Such equation systems normally have a few degrees of freedom. The units are linked to each other by the material and energy streams with another set of process variables, such as flow rate, pressure and temperature. The challenging task is to minimize the investment and operating costs of the plant with respect to these process variables. In general, adjustments in the operation of one of the units will have consequences for other units, and these relationships are often nonlinear. Mathematical programming (MP) and deterministic optimization methods are widely used in process design as these methods have the ability to find the best possible solution for the mathematical model that describes the process (Edgar, Himmelblau, & Lasdon, 2001). A common example of using MP in process design is the synthesis of heat exchanger networks (HEN). Also, some attempts have been made to connect the HEN with the background process. However, only smaller problems have been solved this way. Two thorough reviews of heat exchanger network synthesis (HENS) were published by Gundersen and Naess (1988) and by Jezowski, 1994a and Jezowski, 1994b. Furman and Sahinidis (2002) have contributed with a critical review and annotated bibliography of 461 papers on HENS. Due to physical laws and economic relations, the mathematical model commonly results in a non-convex nonlinear programming (NLP) problem. Furthermore, with discrete decisions, a mixed integer nonlinear programming (MINLP) problem has to be solved. These types of problems can be hard, or even practically impossible to solve using deterministic global optimization algorithms without further simplification of the model. The main advantage of using equation based programs and global solvers, is that it can guarantee that the global optimum of the model is found (Floudas, 1999). However, if the equation based model cannot be solved unless it is made too unrealistic, then the proven optimum may not be the best possible solution in the real world.

An optimization-simulation tool based on a global Tabu-Search and a local Nelder-Mead search for optimization of an LNG process modelled in the sequential modular process simulator HYSYS is developed. The tool has been successfully applied to optimize the Prico LNG process with 7 independent variables applying three different methods for selecting the heat exchanger area. The objective function value is improved from the initial feasible solution with 23–36% for the investigated cases. The first 100 iterations give the largest improvement. A theoretical analysis of the shape of the CCs and a comparison of the results from the tool indicate that the results obtained are optimal or very close to optimal. It is also shown that the optimization-simulation tool has the ability to escape one local optimum to search for others. It can therefore be concluded that the optimization-simulation tool shows great potential for optimization of energy and petrochemical processes. The main strength with the optimization-simulation model is that local searches are performed from several points in promising areas, whereas only a global search is performed in regions that are not promising. The optimization tool can be implemented to optimize other processes modelled in Aspen HYSYS.