Page 230 - Sustainability in the Process Industry Integration and Optimization

S o f t wa r e T o o l s 207

The system models problems in a natural and highly compact
way. The package includes an integrated development environment
and a group of integrated solvers. GAMS was the first algebraic
modeling language, and it is formally similar to several common
programming languages. Models are described in algebraic
statements that are easy for humans and machines to read. The
system is capable of handling models of many different types, so
switching between model types can be done with a minimum of
effort. For instance, the same data, variables, and equations can be
reused for a linear and a nonlinear model by simply converting a
small number of parameters to variables. GAMS also includes a
variety of solvers for different classes of models.
9.6.2 MIPSYN
The MIPSYN (short for Mixed Integer Process SYNthesizer) is a
user-friendly computer package for the integrated synthesis of new
plants and for the innovative reconstruction of existing plants at
different levels of complexity. The tasks to which it can be applied
range from simple nonlinear programming (NLP) solutions for
plant optimization problems to the mixed integer nonlinear
programming (MINLP) optimization of heat-integrated, flexible
plants. MIPSYN is the successor to the PROSYN synthesizer
(Kravanja and Grossmann, 1990; Kravanja and Grossmann, 1994).
As such, it is based on the most advanced modeling and optimization
techniques: those rooted in disjunctive MINLP. The MIPSYN
software can simultaneously address both discrete optimization
(e.g., selection of process units, their operating status, their
connectivity, and ranges of operation) and continuous optimization
(of temperatures, flows, pressures, etc.). The package integrates the
following methods and related components: (1) GAMS with a variety
of different NLP and MILP solvers; (2) different versions of the outer
approximation (OA) algorithm—including the modified OA/ER
algorithm (“ER” denotes “equality relaxation”) and a new logic-
based OA/ER algorithm—which are supervised by MIPSYN
command files; (3) a simple simulator that serves as an initializer to
provide NLP subproblems with feasible (or nearly feasible) starting
points; (4) a library of models pertaining to process units,
interconnected nodes, and the simultaneous integration of heat and
mass; (5) a database of the physical properties of the most common
chemical components; and (6) a hybrid modeling environment, with
a link to external FORTRAN routines, for solving the implicit part of
synthesis models.
Execution of the NLP and MILP steps in OA algorithms is
performed through the use of GAMS saving and restart capabilities,
which enable the user to execute MIPSYN in automated or interactive
modes of operation. The synthesizer features many important