Using Multiple Objective Optimisation to Generate PolicyInsights for System Dynamics Models

摘要

Multiple objective optimisation (MOO) is an optimisation approach that has been widelyused to solve optimisation problems with more than one objective function. The benefitof this approach is that it generates – using genetic algorithms - a set of non-dominatedsolutions which a policy maker can explore and evaluate before making a final optimalselection. This paper demonstrates that MOO can be used to assist policy makers explorea richer set of alternatives when deciding on a range of values for key parameters in theirsystem dynamics model. In order to demonstrate the approach, a well-known case study –The Domestic Manufacturing Company – is used, and a stock and flow model and amultiple objective optimiser are designed and coded. The results show that validsolutions are generated, and that each of these solutions can be examined independently –and hence give greater insight into the problem at hand - before a decision is made as tothe most appropriate solution.