期刊名称：International Journal of Computer, Information, and Systems Science, and Engineering
印刷版ISSN：1307-2331
出版年度：2008
卷号：02
期号：03
页码：156-156
出版社：World Academy of Science, Engineering and Technology
摘要：This paper describes the optimization of a complex
dairy farm simulation model using two quite different methods of
optimization, the Genetic algorithm (GA) and the Lipschitz
Branch-and-Bound (LBB) algorithm. These techniques have been
used to improve an agricultural system model developed by Dexcel
Limited, New Zealand, which describes a detailed representation of
pastoral dairying scenarios and contains an 8-dimensional parameter
space. The model incorporates the sub-models of pasture growth and
animal metabolism, which are themselves complex in many cases.
Each evaluation of the objective function, a composite 'Farm
Performance Index (FPI)', requires simulation of at least a one-year
period of farm operation with a daily time-step, and is therefore
computationally expensive. The problem of visualization of the
objective function (response surface) in high-dimensional spaces is
also considered in the context of the farm optimization problem.
Adaptations of the sammon mapping and parallel coordinates
visualization are described which help visualize some important
properties of the model¡¯s output topography. From this study, it is
found that GA requires fewer function evaluations in optimization
than the LBB algorithm.
关键词：Genetic Algorithm, Linux Cluster, Lipschitz
Branch-and-Bound, Optimization
Loading...
