Page 189 - Sustainability in the Process Industry Integration and Optimization

166 Cha p te r E i g h t

The issues tackled by PI are essentially complex optimization
problems. As a result, optimization is used by PI to answer the
question of “how should the task be performed?” The general goals
and specific targets are usually achieved by employing optimization
tools at various stages. For instance, process performance targets are
typically evaluated by employing a numerical technique that involves
a cascade of some sort—a heat cascade in the case of heat recovery
targeting. One way to implement such cascades is by using the
transshipment optimization formulations, where the external utility
use, resource intake, or emissions rate is set as the objective function
to be minimized. Once the PI goals are established, engineers strive
to achieve the best possible performance. In the case of grassroots
design or network synthesis, the criterion is minimization of the total
annualized cost; in the case of a retrofit, the main criterion may be
minimizing the investments necessary to achieve a certain
performance improvement or minimizing the payback period for a
given investment. For operational improvements, the criteria include
minimizing operating costs or maximizing marginal financial or
performance gains. In all cases, a certain system model—including
the appropriate objective function—is formulated. The model is then
subjected to optimization toward the end of achieving (or maximally
approaching) the PI targets. Another function of targets is to partition
complex optimization problems into sets of simpler problems that are
easier to solve. This approach exemplifies the problem decomposition
principle, applied for decades in the world of software development,
also known as the “divide and conquer” strategy.

8.2 Optimization Tools for Efficient Implementation of PI

For optimizing process models, a wide variety of linear programming
(LPR), nonlinear programming (NLP), and mixed-integer programm-
ing (MIP) methods can be used, depending on the nature of the
problem being solved. Some of these methods were described in
Chapters 3 and 7. Special tools and software (see Chapter 9)
incorporating optimization methods have been developed to exploit
PI possibilities when performing process synthesis, accounting for
the interactions between process operating conditions and the
networks for resource recovery (energy and water). There are four
main groups of optimization tools applied for PI. First, the Pinch
Analysis (Linnhoff et al., 1982) enabled industrial engineers to
obtain better results with the simple Pinch Design Method than
with Mathematical Programming methods in applications to
industrial Heat Integration; see Chapter 4. Second, the graph-
theoretic method is based on process graphs (P-graphs), which were
originally developed for Process Network Synthesis (PNS) (Friedler
et al., 1992b; Friedler, Fan, and Imreh, 1998); see Chapter 7. Third,
Papoulias and Grossmann (1983) introduced linear constraints in