Article 1:
BRILL ED (1979) USE OF OPTIMIZATION MODELS IN PUBLIC-SECTOR PLANNING, MANAGEMENT SCIENCE, 25
Summary:
This article summarizes and elaborates on major problems in optimization algorithms. These problems include many local optima (meaning a true optimal point is difficult to determine), similar optimal solution values that have great differences with respect to the input parameters, and a lack of ability to characterize all potential objectives in the system (meaning some solutions are left out). Also, adding new objectives to a system can cause inferior points in the original analysis to become optimal points, thus meaning that a new solution must be considered. The author then suggests that the means around these problems is a shift in focus by analysts where, rather than trying to find "the answer" (or the most optimized point), sets of solutions should be generated and put into the form of alternatives for the planners to select designs from. He then elaborates on the potential methods of implementing optimization models in a planning process. Included in this are joint use of models (such as simulation/optimization, analytical/optimization, or a toolbox of many methods) and the use of models to generate alternatives and/or facilitate evaluations. The author says that model runs can allow people to look at various parameters of the outlook and then try to determine which aspects they like...and then build upon those suggested outputs. Ultimately, the paper suggests that modelers should tend to put emphasis on potential alternatives rather than a singular optimal point as this will allow model flexibility and possibly allow the user to create models that generate more effective outputs.
Summary:
I really appreciated this article, because I have often wondered about the validity of various solutions in an optimization attempts. Obviously, there are multitudes of directions to take for further analysis of the planning and alternative approaches. Traditionally, I have seen optimization models skeptically in the sense that I know that often political or personal reasons can completely nullify a model's output when making final selections.
Article #2
Pan TC, Kao JJ (2009) GA-QP Model to Optimize Sewer System Design, JOURNAL OF ENVIRONMENTAL ENGINEERING, 135(1) 17-24
Summary:In this article, the authors used a Genetic Algorithm and quadratic programming model to determine optimum designs and alternatives for a sewer system design. When designing sewer systems, the primary decisions involve pipe diameters, and pipe slope decisions, which are dictated by topography of the area. Both of these factors then contribute to the cost of the system. Because of the large number of decisions to make in a system, the researchers noted that the GA analysis would have difficulty generating a solution, and thus they used the QP model to assist in the search.
Discussion:
This article serves as an extension and application to the initial article. In this case, the modelers used multiple models to help refine the searches, and then attempted to generate better guesses to plug into the QP program. I still do have trouble following the approach, so I am looking forward to class to determine more information.
