This paper discusses simulated annealing (SA) as a technique for finding the optimum configuration and power settings for multiple compressors where the number of compressors is large, their arrangement is both series and parallel, and their behavior is non-linear. Statistical mechanics is used in the study of condensed matter physics. This discipline utilizes statistical techniques to estimate the aggregate or macroscopic behavior of a ' natural system from consideration of its microscopic features. The molecular behavior of a liquified metal as it cools and solidifies may be described using statistical mechanics. This natural process is known as annealing. The known physical laws that support statistical mechanics have been used many times in the past to describe the aggregate behavior of some simulated system. In this paper, compressors and their operation have been cast in the role of the system to be simulated, and a technique, SA, derived from statistical mechanical considerations, has been used to find the "best use" mode of their operation. The paper discusses tests of SA against the more ubiquitous mixed integer non-linear (MINLP) and heuristic techniques. The heuristics tests were carried out at the Texas Eastern Transmission Corporation (TETCO), and some discussion of that system is included. Strictly speaking, optimization is the practice of the doctrine of optimism. Philosophically, it means to hold the view that everything in the world is naturally ordered in the best possible way. Probing the natural continuum with tools designed to reveal more and more of the natural order and its structure, it is easy to become increasingly more impressed at the perfection with which the pieces of the natural jig-saw puzzle fit together. How is this jig-saw maintained? How does nature search an infinite solution space and instantly come up with an answer that satisfies all conditions in any given situation? For example, when a liquid metal solidifies, how does nature solve the problem of instantaneously positioning billions of molecules in their proper places? Investigation doesn't yield much of an answer to this kind of metaphysical question. it does however, reveal what appear to be links and similarities between natural systems' architectures and the structures of very large and complex problems, and further shows that the emulation of natural processes helps solve these problems. It seems that nature deals only in probabilities. So it is with the problem described in this paper; where enumerative and calculus methods have failed to handle large and discontinuous problems, statistical techniques that emulate natural processes have enabled the inquiry to push on into previously unexplored space (for this particular problem, at least), yielding answers that are more consistently optimal than those obtained using earlier methods. Right or wrong, the metaphysical view that nature is a perfect whole with a structure that is consistent throughout is appropriate to this paper, because it deals with a complex system problem whose solution is given by direct comparison with a highly structured natural process.

This content is only available via PDF.
You can access this article if you purchase or spend a download.