考量環境限制下之物流路網設計問題

Abstract

永續發展及環境保護的議題逐漸被各國各界重視，從最初溫室效應、聖嬰現象開始，到近幾年的極端氣候，使各界警覺到人類對大自然產生的破壞已逐漸衍生為對人類生活的迫害。極端事件的發生與大氣中二氧化碳濃度逐年升高有關，為減緩極端事件的現象，世界各地皆有制定不同的因應政策，以減少溫室氣體或汙染物質的排放，因此許多相關研究也因此應運而生，如相關環境政策之評析、綠色運輸之績效評估與指標建立、減排策略之擬訂，以及減排技術及運具之研發等。本研究則提出一個降低路網碳排放量之減排模式供政府決策單位參考，當政府有特定的減排目標時，可以此模式檢視能否以改善政策達成之及其可達成程度，並在擬定相關改善政策的同時，兼顧下層承攬業者對政策的反應，使路網發展方向符合減排目標。 本研究以雙層規劃方法(bi-level programming)建構物流路網設計問題之減排模式，上層問題為減碳政策模式，下層問題為運輸路網指派模式。上層政府決策者期望透過改善政策降低路網碳排放量，改善政策包含投資港口處理容量之投資政策、補貼業者使用特定港口之補貼政策及外部成本內部化政策，而下層物流業者則在成本最小化的前提下完成運送服務。上層政府決策者在做決策時會參考下層物流業者對改善政策的反應，而下層業者也會將其對政策的反應回饋給上層決策者，兩者間之關係為主從賽局。 下層模式為線性問題，故可以單行法求解；上層模式起初以Sensitivity Analysis-based algorithm (SAB法)求解，但此法只在下層均衡解對上層變數之敏感度為連續且可微分之情況下可行，而本研究之下層均衡解對補貼變數之敏感度為連續但不可微分之函數，故發展Branch-SAB法來處理此問題。Branch-SAB法是先以求解離散型路網設計問題之方式處理補貼變數後，再以求解連續型路網設計問題之SAB法求解其餘決策變數。 最後以一範例說明求解過程及結果並進行數值分析，結果顯示改善政策可降低一部份的碳排放量，但有其上限。而在不同減排目標下，改善政策內容亦有所不同。整體而言，當減排目標越嚴苛，上層目標值呈現增加的趨勢，路網碳排放量、環境成本、運輸成本及下層目標值則呈現下降的趨勢。

It is believed that global warming and extreme weathers is a result of the increasing amount of emissions and greenhouse gases. With the rising environmental concerns in the recent years, it is important to take the related emissions into account when designing the freight transportation network. Policy issues for carbon emissions reduction are widely studied, such as, the evaluation of environmental policies, the strategies and techniques for reducing emissions.How to lower the emissions to meet a mandatory cap is also an issue for freight network design problem. This study proposesa bi-level programming formulation to model the policies of emission reduction for freight transport,whereas the lower-level problem is a transportation network assignment problem modeling the network flows, and the upper-level problem determines the most optimal improvement policies for minimizing the improvement costs. Three improvement policies are considered, including enhancing port capacity, subsidy for using a port, and internalization of external cost. Under the bi-level framework, the leader at the upper-level would consider the reaction function of the follower at the lower-level, and the follower would give its feedback to the leader as well, under the Stackelberg equilibrium. A Branch Sensitivity Analysis-Based (Branch-SAB) algorithm is proposed to solve the upper-level problem, in which the sensitivity of the lower-level equilibrium solution respect to the upper-level decision variables is continuous but non-differentiable at certain points, whereas the lower-level model is a linear programming problem and can be solved by the Simplex Method. Numerical example with a hypothetical network is used to demonstrate the performance of the model, and the results show that the improvement policies can reduce the network emissions, subject to a reduction upper bound. A higher improvement cost is needed for a stricter reduction target, but the corresponding network emissions, environmental cost, as well as transportation cost would be decreased.