Multi-objective Optimization for Location/Allocation Planning of Networked Facilities Using Genetic Algorithm

考虑到开发灵活的优化方法管理冲突解决的重要性，我们研究了多目标下的场地选址问题。正如“邻避”（Not In My backyard）一词所述，危险废物的场地选址问题是研究这种情况的一个合格案例，也就是说，它与所有人类、环境和经济问题有关。本文首先对场地选址问题的研究背景进行了广泛的研究，并给出了几种典型的场地选址公式和几个实例。然后将该问题概括为多目标混合整数规划（MOMIP）。为此，我们提出了一种智能支持的方法，结合遗传算法（GA）与数学程序的混合使用，以及用于价值函数建模的神经网络（NN）。为了提高目前线性规划中的数学程序的性能，在程序实现阶段，我们采用了作者开发的渐进式线性程序。此外，针对多目标遗传算法的求解问题，我们引入了一种新的遗传操作——修复操作，以提高算法的性能。从实际的角度来看，我们也认为提供一种反映现实世界问题解决中嵌入的某些定性方面的方法非常重要。关于这一点，我们已经表明，目前的方法可以很好地处理这种情况。以垃圾处理场规划问题为例，将其表述为混合整数线性规划（MILP），从涉及的各个方面对所提出方法的有效性进行了数值检验。

Taking importance of developing flexible optimization methods managing conflict resolution in mind, we have concerned with site location problems under multi-objectives. As known from the term NIMBY (Not In My Back Yard), the site location problem of hazardous wastes is an eligible case study for such a situation, that is to say, associated with all human, environmental and economic concerns.First backgrounds of the site-location problems have been investigated broadly, and presented some kinds of typical formulation with a few examples. Then we have formulated the problem generally as multi-objective mixed-integer programs (MOMIP). For this, we have proposed an intelligence supported approach incorporating a hybrid use of genetic algorithm (GA) with mathematical program, and neural networks (NN) for value function modeling. To improve the performance of the mathematical program referred to the linear programming presently, at the stage of program implementation, we applied the progressive linear program developed by the author.Furthermore, concerned with the solution of multi-objective GA particularly, we have introduced a new genetic operation named repair operation to improve performance. From practical points of view, we also considered it very important to provide an approach reflecting certain qualitative aspects imbedded in real-world problem-solving. Concerning with this, we have shown the present approach can deal with such like situation very suitably.By taking an example in waste disposal site planning problem, which was formulated as a mixed-integer linear program (MILP), we have examined the effectiveness of proposed approach from every aspect concerned here numerically.