中期貨櫃碼頭營收管理問題

Abstract

在過去的幾十年中，由於國際貿易的盛行，對於海運的依賴也愈來愈大，隨著船隻的大型化以及需求增加，在現在碼頭不斷地面臨擁擠的問題，因此如何進行碼頭營運效率的安排是一項關鍵性之決策，本研究引入營收管理(Revenue Management, RM)的策略，核心為中期的營收管理問題結合船席模板問題，將港口的營運週期分為尖峰時刻(peak hour)以及離峰時刻(off-peak hour)，透過不同的定價可以反應出船舶業者在不同時間點的願付價格，並且會產生出每個不同船舶業者所對應尖離峰的價格的時間窗(time window)，在依照決定好的時間窗、尖離峰時間以及訂價之後，可以將這些資訊套入進行策略性中期船席模板問題(Berth Template Problem, BTP)的求解，船席模板問題能夠在一個固定的循環週期長度內重複使用，其目的為降低有大量船舶停靠需求之貨櫃碼頭的擁擠程度並且同時能夠提升港口營運業者的利潤，讓港資源做最適當的安排。因此本研究為一個二階規劃(Bi-level Programming)的決策問題，上層為運用營收管理所進行之差別訂價決策，下層為中期的船席模板問題，上下層之間存在關聯性，因此本研究於下層BTP問題中找到上層的時間切割點與尖離峰價格之搜尋機制，同時於下層BTP使用路徑生成法(Path Growing Algorithm, PGA)進行有效率的近似求解。在數值測試的部分，與最佳解的求解時間相比，本研究的演算法設計能夠大幅降低求解時間，並且有良好的求解績效，增加實務上能夠運用的可能性。

In the past decades, world trade have been growing so that we are much more rely on the sea transportation. The resulting cargo traffic growth and the increase of demand lead the ports into congested problem. Therefore, how to manage the ports efficiently is a key point. This study introduces revenue management (RM) strategy to do the price discrimination. This study divides the port handling cycle into peak hour and off-peak hour. Because each shipping line has their own preferences time slot and the willingness-to-pay (WTP), we can get the time window of each shipping line. And then we have to know if the ports can service all the ships or not. So the research developed an integer programming (IP) model for the midterm berth template problem. With the objective of revenue maximization, the ports manager can have more revenue and use the set of price to reduce the congested problem. This study develops a bi-level modeling framework to link the two decisions: pricing and berthing. The pricing decision in the upper-level problem determines the rate of the berthing service with respect to time. In the lower-level problem, the shipping lines make the berthing requests based on the pricing decision of the terminal operator, and the berth planning decision is made accordingly. This study use the feedback information from the lower-level to search price in the upper-level. And in the lower level uses the path growing algorithm to find the solution efficiently.