Session: 08-03-01: 4E Analysis and Optimization of Thermodynamic Systems

Paper Number: 71428

Start Time: Wednesday, 05:00 PM

71428 - Mixed-Integer Nonlinear Programming (MINLP) Based Optimization of Refrigeration Systems Integer Nonlinear Programming (MINLP)-Based Optimization of Refrigeration Systems 

Due to the increased role of the refrigeration systems in the global economy, the study on different refrigeration systems gained renewed interest. Particular attention shoul be given to the heat-driven systems, i.e. ad- and adsoption, ejection and compression heat-driven systems. Cascade and hybrid absorption-compression refrigeration cycles are being investigated by several authors. Such cycles attractively combine the advantages of each standalone cycle. This paper adresses the optimization of refrigeration systems by proposing a deterministic mixed integer nonlinear programmng (MINLP) model. The superstructure-based representation of vapor compression-absorption cascaded refrigeration system recently presented by authors is extended in order to include more candidate configurations as well as a cost model considering capital and operation expenditures (CAPEX and OPEX). Then, a MINLP model is derived from the resulting superstructure and solved using an optimization algorithm based on the generalized reduced gradient method. A simple solution strategy but flexible and robust enough to facilitate the numerical convergence of the model is proposed. The optimal configuration is supposed to be selected from the superstructure-based model considering the following candidate configurations: 1) single effect vapour absorption refrigeration system (S-ARS), 2) double effect vapour absorption refrigeration system (D-ARS), 3) compression refrigeration system (CRS), 4) single effect evaporation–vapor compression system (S-ARS/CRS), 5) double effect evaporation–vapor compression system (D-ARS/CRS). The condenser operating at the high temperature in the double effect VAR as well as the heat exchangers operating at the low and high temperatures are also considered as candidate process-units and they can be optionally selected or removed by the optimization algorithm. In the absorption refrigeration system, if it is selected by the algorithm, water will be used as refrigerant whereas Li-Br as absorbent. However, the refrigerant to be used in the compression refrigeration system it is not known at priori and represents an optimization variable – discrete decision –. If this system is selected, the refrigerant will be selected from the list of candidates. The optimization model is inherently combinatorial in nature and involves more than 20 candidate combinations. Given the desired design specifications – temperature and cooling capacity – the optimization problem consists in finding the optimal configuration, sizes of all process-units and operating conditions that minimizes the total annual cost. Also, the maximization of the coefficient of performance (COP) and the minimization of the total heat transfer area of the system are investigated. Different design specifications are investigated. Reference cases taken from the literature are used to compare the optmal solutions obtained from the proposed MINLP model. The proposed model is fully implemented in GAMS – general algebraic modeling system – which is a high level modeling environment widely used in Process System Engineering PSE.  DICOPT is used as MINLP solver, CONOPT as NLP solver, and CPLEX as MIP solver. A sensitivity analysis is performed in order to investigated the influence of the main model parameters on the optimal solutions. To the best of the authors’ knowledge, there are not publications that addressed the minimization of the total heat transfer areas and total annual cost of vapor compression-absorption cascade refrigeration systems by proposing a superstructure-based model with the possibility to select refrigerants and remove system components, and using determinist optimization algorithms.

Presenting Author: Tatiana Morosuk Technische Universitat Berlin

Authors:

Sergio F. Mussati INGAR (CONICET-UTN)
Tatiana Morosuk TU Berlin
Miguel C. Mussati INGAR (CONICET-UTN)