- 學位論文
城際客運高速鐵路列車停站與服務規劃之最佳化研究
Optimization of Train Stopping Patterns and Service Plans for High Speed Railways
摘要
城際客運高速鐵路業者為滿足不同旅客之需求，通常會透過「列車停站規劃」來製定如每站皆停、區間停車、越站停車或直達快車等停站型態組合，接著再以「列車服務規劃」來擬訂各停站型態在各時段之發車頻率。實務上，此兩項規劃作業通常是依據經驗法則或是在考量地方公平需求下妥協產製，其中經驗法則可能無法顧及整體資源與需求之最佳配置，而經公平訴求所產製之結果可能會犧牲多數旅客的權益。 基於上述原因，本研究延續過去的成果致力構建「列車停站規劃」與「列車服務規劃」之最佳化整合模式。研究中「列車停站規劃模式」精進為可求解大規模的0-1整數規劃問題，以最短旅客車內旅行時間為目標，同時將階級化停站方式納入考量，並設計數種情境以求得最佳停站型態組合；至於「列車服務規劃模式」則是採混合整數規劃予以構建，模式之目標為業者利潤最大及旅客旅行時間成本加權值最小，除了將列車與各站旅客時空遞移之動態需求納入考量外，並將上一階段得出之列車停站組合作為輸入，以在不同情境中找出各時段下之最適發車頻率。 本研究以台灣高鐵2008年之背景資料進行案例探討，在列車停站規劃之眾多案例中，以4種「台北-左營」為起迄之最佳停站型態組合而言，旅客車內旅行時間相對於高鐵既有停站型態可降低6.2%之幅度，每日約可節省6,734旅客小時；在列車服務規劃方面，若將上述最佳停站型態組合作為列車服務規劃之輸入，則相對於高鐵公司之停站型態組合及發車頻率而言，本研究所得最佳解之目標值可顯著增進12.57%之幅度，每日約可增加業者利潤$2,419,201，透過敏感度分析可知旅行時間成本之權重值越大、或是未服務旅客轉移比例越低，則整體服務規劃的目標值越低。 本研究之具體貢獻係透過所構建之模式操作，可以避免傳統經驗法則或試誤法所須耗費的大量時間與成本，在合理時間內即可進行多種不同情境的列車停站與服務頻率規劃以供決策者評估；另一方面，若面對營運環境改變或者公平訴求希望變動特定的停站型態或服務頻率時，亦可運用所建最佳化模式評估各方案的績效，作為決策權衡的參考。整體而言，本研究構建之模式不但能滿足旅客需求，亦能協助業者創造更多利潤，可達成雙贏的規劃目標。
並列摘要
To provide satisfactory services to passengers, intercity high speed railway operators usually design various stopping pattern combinations through “train stop planning”. The stopping patterns may include all-stop service, zonal service, skip-stop service, and express service. The operators then assign adequate service frequency for each pattern to meet passengers’ needs through “train service planning”. In practice, stopping patterns and service frequencies are typically determined by empirical rules or the needs of equity proposed by local individuals. The former empirical rules and individual needs may not consider the optimal allocation of resources while the local equity purpose may even sacrifice all passengers’ benefits. Thus, this research, based on previous studies, aims to develop the optimal models for integration of these “train stop planning” and “train service planning”. Train stop planning model is refined to deal with large-scale binary integer programming problem. The objective is the minimization of passenger in-vehicle time while the classification of stopping mode is also taken in account. The study then designs several scenarios to generate the best combination of stopping patterns. Train service planning model is constructed in a mixed integer programming framework while the objective is the maximization of operator profit and the weighed passenger traveling time cost. Furthermore, the train service planning model considers dynamic demand feature of trains and passengers over time and space. Additionally, the optimal combination of stopping patterns obtained from the stop planning model is used as the input for the train service planning model to determine the service frequencies of each stopping pattern in different periods. This research takes Taiwan High Speed Rail as a case study. It is shown that if taking the optimal stopping pattern of “Taipei-Zuoying” into account, the objective value of optimal stopping patterns combination has 6.2% better than the existing one, which can save 6,734 passenger hours a day. It is also shown that the objective value of optimal train service plan is 12.57% better than the original one, which increases $2,419,201 profit a day. Furthermore, the sensibility analysis shows that objective value of train service planning will decrease when weighting factor increases or un-serviced passenger’s transfer rate decreases. The main contribution of this research could then be verified. The mathematical model developed can help of saving time and cost resources based on research results comparing to the empirical rules or trial and error practices. Furthermore, the model can effectively and efficiently find the optimal solutions in different scenarios to enhance quality of decision making. On the other hand, the model can provide performance evaluations for both pros and cons of specific cases while responding to environment changes or the requirements of equity. In summary, the proposed models can help of not only meeting passengers’ needs but also considering operator’s profitability.