模糊多目標組合規劃基因演算法應用於提昇運輸系統災後應變效率之研究

Abstract

自然或人為的災禍，如颱風、空難、車禍等對運輸系統常造成極大之負面影響。若運輸系統災後效率低落，則救災資源如民防團、消防隊、醫療小組、土木工程隊等將無法及時趕赴災區提供必要支援，民眾則因此遭受生命財產之巨額損失。以往運輸安全之研究多著重於災前、災時及災後應變復舊觀念之文字宣導，但應如何合理化、科學化及邏輯化地落實這些應變原則卻付之闕如。為合理化、系統化暨科學化地進行大規模救災及復舊，因此，本運輸安全研究不同於以往許多根據統計分析或建立標準應變程序之相關研究方法，而是創新地提出應用模糊多目標規劃以求解此類問題。復因大規模救災及復舊問題亦為一組合最佳化問題(Combinatorial Optimization Problems），而國際上風行之基因演算法 (GAs)具有平行搜尋及自我演化的特性，已在理論基礎上被證實在求解組合最佳化問題上最具效益且特別適合應用到大規模的實務問題上。故本研究應用基因演算法求解大規模救災及復舊問題，以提昇運輸系統災後應變效率。研究中並提出二個案進行探討，第一個案為機場最適消防站之區位規劃，冀使機場內事故損失最小化;第二個案為震災後路網復舊之最佳化排程，欲使破損路網修復時間最小化，且同時修復期間之路網交通擁塞最少。研究結果顯示，利用基因演算法於此二個案可簡便地求出其近似最佳解，而本論文將可提供決策單位研擬相關應變策略之科學基礎。

Natural disasters or man-error often cause negative impacts to a transportation system. People injured or died in these disasters because of the inefficient reaction in a post-disaster transportation system. Past transportation-safety researches have focused almost exclusively or identifying the concepts of disaster prevention, disaster response and disaster recovery of a transportation system. But no further attention has been paid to the question of how reasonably, scientifically and logically secure the efficient reaction of a post-disaster transportation system. Since many activities of rescuing lives, e.g., fire-fighting, restoration, refugees' evacuation, etc. are mainly supported by transportation systems (networks), if the reaction efficiency of a post-disaster transportation system is unacceptable, mass travelers, e.g., doctors, fire-fighters, policemen, work-troops for reconstruction, patients, etc. can't be efficiently conducted via the destroyed transportation network. Thus, this study is far from many traditional transportation-safety researches, which heavily relied on conceptual introduction, statistical analysis or standards establishment. Thus, to promote the reaction efficiency of a post-disaster transportation system, we creatively apply the fuzzy multi-objective programming in such a study so as to implement appropriate strategies for promoting the reaction efficiency of a post-disaster transportation system. Two practical transportation-safety issues: airport firestation location and network restoration problem are considered in this study and resolved. Study results show that satisfying solutions can be easily derived by our modified genetic algorithms (GAs). Moreover, we confirm that our efforts devoted in this study can be a powerful decision support basis for pre-simulation against disasters.