用遗传算法求解物流运输中多级中转站定位优化问题

A Genetic Algorithms Approach to Optimum Locating of Multiple Stations in Logistics Transportation

文章建立了物流运输中多级定位优化大规模非线性混合整数规划模型。由于该模型用传统方法直接求解相当困难,文章应用遗传算法对该模型进行了求解。在建模过程中,对模型中的连续实型变量进行离散化处理,从而使整个优化模型变成纯0-1非线性整数规划优化模型;在求解过程中,应用自适应α绝断-指数比例变换适应度法,提高了快速搜寻该模型全局最优值的能力;应用自适应概率指数比例变换法,改进了交叉概率计算方法;应用基于基因权重对基因位置进行动态排序的方法,使优良基因变得集中,从而克服了交叉算子容易破坏长度很长的优良模式的弱点,并依此改进变异概率的计算方法。仿真表明,应用文章提出的遗传算法计算模型,可在微机上稳定地获取该模型的最优解。

This paper establishes the large-scale MIP model of optimum locating of multiple stations in logistics transportation. Because the model is very difficult to solve by the traditional methods,a synthetic solution is presented by genetic algorithms. In establishing the optimization model, the real continuous variables are changed into discrete 0-1 variables so that the nonlinear MIP model is transferred into a pure 0-1 nonlinear IP model. In evolution simulating process, the adaptive αcutoff-pewer law scaling method is applied to improve fitness so that the search ability of global optimum solution increases largely; the adaptive power law sealing method is used to improve crossover possibility; the method of dynamic ranking of gene loci by gene weights is presented to make excellent genes concentrated so that the shortcomings of crossovers to destroy excellent schemata with great encoding lengths are overcome, the method is also used to improve mutation possibility. The empirical results show that the optimum solution of the optimum model can be gained on microcomputers.