國內航線旅次需求型態推估與班次起飛時間之訂定

Abstract

規劃良好的班表是航空公司達成較高營收或較低成本的基礎，其中班次時間之訂定則是規劃班表的最主要工作。班次時間之訂定須以需求型態為主要依據，但由班次承載情形所顯現之需求型態(市場需求型態)為旅客囿於既有班表並考慮各種因素後選擇的結果，與旅客真正想使用航空運具之需求型態(實際需求型態)已有差異；實際需求型態雖可利用對旅客的實際調查得出，但此將須耗用龐大的人力與時間，所得之結果亦可能隨時間或空間的變化而無法適用，如何利用簡易可得的資料以合理推估航線上旅客的實際需求型態，實為班表規劃研究之重要課題。旅客在選擇班次時，通常仍會考慮各班次對其之吸引等因素，故即使航空公司能掌握各時間點旅客數，亦可能因其他班次之競爭而未能使其班次有較高的承載情形，合理的班次承載人數估計才是航空公司據以排定班次起飛時間的依據，而此班次起飛時間選擇的方法，由於相關研究不足且未能配合國內短程航線之特性，更有發展的需要。 依此，本研究先由市場之發展演進、供需現況，以及對旅客問卷調查資料之分析等，歸納說明市場各項特性，並進一步分析整理國內旅客選擇班次之行為與考慮因素，假設旅客選擇班次時會有一考慮的時間範圍，構建旅客對班次之選擇模式，以計算旅客選擇各班次之機率，再配合各時間點意欲使用航空運具之人數(實際需求型態)，以收集各時間點選擇班次的人數便可估計出班次承載人數的想法，建立班次承載推估模式。雖然各時間點意欲使用航空的人數未知，但因班次實際承載人數可觀測得出，因此，本研究採動態遞迴推估方式，以估計之實際需求型態，依已知的班次選擇機率估計各班次承載人數，經與實際觀測班次承載人數比較得出誤差，用以修正之前之估計，反覆進行以逐漸修正得出較低估計誤差之實際需求型態。又因利用班次承載推估模式可估計出各時點之可能承載人數，乃進一步依現有機隊規模，在受限的服務班次數下，以總服務旅客數最多為目標，構建一班機起飛時點選擇之二元整數規劃模式，以規劃草擬班表之班次起飛時間點，並依問題所具有特殊結構，發展一適用之演算法求解。 研究結果顯示，推估得到之實際需求型態經班次承載人數推估模式所估計的各班次承載人數，多數班次之估計誤差在20人以下，且近50%之班次其估計誤差在10人以內，顯示出相當的準確性，此估計誤差之產生主要是因旅客選擇行為模式無法完全說明旅客實際行為，以及各班次之承載人數逐日變化所致；而由不同起始值設定仍可收斂至相近需求型態的結果，顯示此遞迴推估模式之適用性，亦即推估所得之實際需求型態已有相當的準確性。另由敏感度分析之結果，顯示票價變動對班次承載之影響明顯，而變動班次起飛時間對班次承載影響之分析，亦可作為航空公司調整其班次起飛時間之參考；此外，由旅客選擇效用函數中班次時間差之參數，早到與晚到校估值的差異，以及採早到與晚到對旅客吸引時間範圍不同時，其估計誤差較採早到與晚到時間範圍相同下小，此等結果皆顯示國內旅客對於早到與晚到的感受不同。

The flights' departure time setting is the critical work of the flight timetable construction, and a well-planned flight timetable can provide higher revenues or lower costs for airlines. The flight timetable construction should be based on demand pattern. However the market demand pattern viewing from flights load is a result of the passengers choice under a well-known flight timetable. There is a difference from the actual air travel demand pattern that means the number of desired departure passengers in each time point. Although actual travel demand may be obtained through a survey conducted on passengers, such a survey takes up enormous manpower and time; moreover time and space changes may render the findings useless. The acquisition of a reasonable estimation of the actual passenger travel demand by simplified data is an important issue for flight timetable construction. Flights remain to be a very important attraction in the flight choice consideration of passengers. Hence the flight load is influenced from competition. Flight load estimation should be the basis for the airline to set the flight departure time. Lack of sufficient studies in flight departure time setting and the inability to coordinate schedules with short-distance domestic flights make it imperative to develop studies in this area. This study traces the market development, supply-and-demand conditions, as well as the passengers questionnaire date survey analysis to classify market characteristics. We analyzed the domestic air passenger choice behavior and consideration factors, and assumed that passengers would choose flights under an acceptable time period. This constitutes the air passenger choice model and the probability to choose flights in each time point. Then the flight load estimation may be gathered from the choice passenger of each time point. A flight load estimation model is developed from this exercise. Although the actual desired departure passenger of each time point is known, the actual passenger load of each flight is observable. Hence our study used the dynamic recursive estimation method to estimate the actual travel demand through the errors of estimation flight load and observation flight load and to revise them iterated. Otherwise, since the flight load estimation model used can estimate the probable passenger load of each time point, a flight departure time setting concept is formulated as a binary integer programming model subject to the fleet size and service flight frequency. Since the flight departure time problem has some good characteristics available, with the help of the very useful information, an efficient algorithm which can solve the problem without too much difficulty is thus proposed. The results from the analyses have revealed that the estimation load errors is less than 20-passengers in majority of the flights, and almost 50% of the flight estimation load error is under 10 passengers. This shows a very high accuracy rate. The estimation error is mainly due to the inability to fully explain the actual passengers behavior in the passenger choice model, as well as the every day regular flight load changes. The varying initial value setting can still provide the closest demand results, a manifestation of the applicability of the recursive estimation model and the high accuracy rate of the actual travel demand obtained from the estimation. The sensitivity analysis results show that ticket price changes have a significant effect on flight load. Analysis of the influence of different flight departure time alteration on flight load may also serve as a reference for the flight departure time adjustment considerations of airlines. Furthermore, the estimation parameter value in passenger choice utility function for passengers to early arrival is less then to late arrivals. This shows that domestic flight passengers prefer the early arrival flight to late arrival flight under the same time difference between passenger desired departure time and flight departure time.