Einführung in Verkehr und Logistik [1.15ex] (Bachelor) [1.15ex
Transcription
Einführung in Verkehr und Logistik [1.15ex] (Bachelor) [1.15ex
WS 13/14 Einführung in Verkehr und Logistik 1 / 46 Einführung in Verkehr und Logistik (Bachelor) Revenue Management and Fleet Assignment Univ.-Prof. Dr. Knut Haase Institut für Verkehrswirtschaft Wintersemester 2013/2014, Dienstag 10:15-11:45 Uhr, Phil E WS 13/14 Einführung in Verkehr und Logistik 3 / 46 2. Konzeptionelle Grundlagen Airline Planning Process1 Langfrist Mittelfrist Implementierung Kurzfrist Kontrolle 3 J. 6 M. 1 J. 8 W. 4 W. 2 W. 0 Flottenplanung Netzwerkplanung Flugnetz Marktmodellierung Fleet Assignment Wochenplanung | fully dated Aircraft Rotation Revenue Management Operations Control Pricing Personal Revenue Management Crew Pairing Crew Rostering Abbildung 1: Allgemeiner Flugplanungsprozess einer Fluggesellschaft Quelle: in Anlehnung an Grothklags 2006, S. 3. 1 Seedie Über weiteren Teilprobleme der Flugplanerstellung besteht in der Literatur, trotz [Gro06]. vielfach divergierender Terminologie, ein breiter Konsens. Im Rahmen der Netzwerk- WS 13/14 Einführung in Verkehr und Logistik 5 / 46 Introduction to Revenue Management Origin of Airline Revenue Management I I I I Airline Deregulation Act of 1978 (prices, schedule) Low Cost Carrier (LCC) entered the market ! price pressure American Airlines first launched special discount fares Instruments of RM: I Price Discrimination I Overbooking I Capacity Controls I Forecasting systems (not covered here) WS 13/14 Einführung in Verkehr und Logistik Basics I Revenue Management (RM) is also known as Yield Management I RM can be either quantity- or price-based depending on the control variable I most airlines commit to fixed prices and tactically allocate quantity I LCC use price as the primary tactical variable I Objective: utilization of additional yield according to individual willingness to pay (WTP) I Successful application also in other service areas: hotels, car rentals 6 / 46 WS 13/14 Einführung in Verkehr und Logistik 7 / 46 Conditions conducive to RM2 I I I I I I Demand variability and uncertainty Production inflexibility (variations in supply difficult) Advance sales Perishable inventory High fixed costs/low marginal costs Customer heterogeneity (preferences) 2 Talluri and van Ryzin (2005, pp. 13). WS 13/14 Einführung in Verkehr und Logistik Price Discrimination (PD) First-degree/perfect price discrimination I Each customer is charged according to its WTP I Requires information on each customers WTP & ability to vary price by customer and unit ! Theoretical abstraction because of lack of information Second-degree price discrimination I Discrimination by offering various possible purchase contracts I Customers decide which contract to purchase ! Self-selection by customers Third-degree price discrimination I Customers are divided into groups based on identifiable characteristics (students, children, retiree) I Prices differ for different groups, all members of a group pay the same amount 8 / 46 WS 13/14 Einführung in Verkehr und Logistik I RM ! 2nd degree PD I Segmentation of customers q I self-selection q(r) I Prices for almost similar goods differ I Different ticket fares within the same class q* r* of service I Economy, Business, First Class I Restrictions apply r q I Optimal strategy ! prices according to q1 q(r) q2 individual WTP I Not feasible I Instead several fare classes (FC) for each q3 r1 WS 13/14 r2 r3 service class r Einführung in Verkehr und Logistik Fare Classes I I I Varying ticket prices for identical product (flight from A to B) Up to 20 different fare classes on a single flight FC are represented by letters I full fares: F - first class, C - business class, Y - economy class I discount fares economy class: M, B, K, H, Q, S, W I Restrictions apply for cheaper tickets: I I I I I Rebooking/Cancellation fees Advance purchase requirements Trip length and length of stay Saturday-night stay Restrictions provide necessary Fencing I to separate demand of business and leisure travellers I to prevent that high-value customers buy-down to cheaper tickets I 9 / 46 Objective: additional demand through cheap tickets and better utilization of capacity 10 / 46 WS 13/14 Einführung in Verkehr und Logistik 11 / 46 Deterministic 1-Fare-Class Problem Price-Sales-Function (PSF) r Offer price q Demand q(r ) Price Sales Function q(r ) is continuously differentiable and exhibits an also continuously differentiable and strictly monotonic decreasing inverse function: 1 r (q) = q (r ) Example: Linear PSF q(r ) = a r (q) = a=b WS 13/14 br 1=b q 0r a=b 0qa Einführung in Verkehr und Logistik Revenue function u(q) = q r (q) Marginal revenue function u 0 (q) = r (q) + q r 0 (q) Assumption: The marginal revenue function is strictly monotonic decreasing within its domain. Example: Linear PSF u(q) u 0 (q) u 00 (q) = a=b q = a=b = 2=b 1=b q 2 2=b q ! concave function 12 / 46 WS 13/14 Einführung in Verkehr und Logistik 13 / 46 Objective Function: Maximizes Revenue max u(q) = q r (q) Constraints q q C 0 Optimal Solution Extreme value with disregard of the capacity constraint: u 0 (q) = 0 ! q0 q = minfC ; q0 g und r = r (q ) WS 13/14 Einführung in Verkehr und Logistik Example q(r ) = 400 0:5r C = 120 Solution: r (q) = 800 u(q) = 800 q u 0 (q) = 800 2q 2 q2 4q ) q0 = 200 ) q = minf120; 200g = 120 ) r = r (120) = 560 u = 800 120 2 1202 = 67 200 (= 120 560) : 14 / 46 WS 13/14 Einführung in Verkehr und Logistik 15 / 46 Capacity Controlling I I I I I Revenue maximizing control of sales processes Decision about acceptance/rejection of booking inquiries Single-leg or network flights Supports price discrimination and erroneous fencing Types of controls: I Booking limits I Protection Levels I Bid prices I Principles: I Capacity is allocated to a request if its revenue is greater than the value of the capacity required to satisfy it I The value of capacity is measured by its expected opportunity cost (displacement cost) I Optimization problem regarding the acceptence of booking requests WS 13/14 Einführung in Verkehr und Logistik Booking Limit3 bj I Limits the amount of capacity C that can be sold to any fare class j I Partitioned: available capacity is divided into seperate blocks I Nested: available capacity overlaps in a hierarchical manner, more expensive fare classes have access to all the capacity reserved for cheaper fare classes Protection Level yj I Specifies an amount of capacity to reserve (protect) for a particular fare class j Formal Relationship bj = C 3 Talluri yj 1 and van Ryzin (2005, pp. 28-30). j = 2; : : : ; n 16 / 46 WS 13/14 Einführung in Verkehr und Logistik 17 / 46 Relationship between bj and yj WS 13/14 Einführung in Verkehr und Logistik Example An airline is operating the flight legs A-B and B-C as well as the combination of both A-B-C. The utilized aircrafts have a capacity of 100 seats. For the purpose of price discrimination two fare classes are assigned to each flight leg. Overbookings are not taken into account. The following table shows the expected demand per flight leg and fare class: Route A-B B-C A-B-C Price (EUR) Demand Price (EUR) Demand Price (EUR) Demand Class W 250 100 220 80 460 80 Class K 490 40 400 60 880 50 How many seats per flight leg in combination with a certain booking class should be sold? 18 / 46 WS 13/14 Einführung in Verkehr und Logistik 19 / 46 Linear Optimization Model xij number of seats in class W on flight leg i-j (i.e., A-B, B-C or A-C) yij number of seats in class K on flight leg i-j max F (x ; y ) xAB xBC 250xAB + 220xBC + 460xAC + 490yAB + 400yBC + 880yAC = +xAC +xAC +yAB +yBC +yAC +yAC xAB xBC xAC yAB yBC yAC xAB ; xBC ; xAC ; yAB ; yBC ; yAC Optimal Solution: (F (x ; y ) = 86100 EUR): 100 100 100 80 80 40 60 50 (A (B B) C) 0 and integer xBC = xAC = 0; xAB = 10; yAB = 40; yBC = 50; yAC = 50 WS 13/14 Einführung in Verkehr und Logistik Overbookings I 50% of all reservations result in cancellations and no-shows I Objective: increasing the total volume of sales in the presence of cancellations I Contrary to RM overbooking does not optimize the customer mix (i.e., best allocation of price or capacity) I Amount of sold tickets exceeds the amount of seats actually available I Airlines are obliged to compensate passengers for denied boarding 20 / 46 WS 13/14 Einführung in Verkehr und Logistik 22 / 46 Terminology Fleet (aircraft) type: A certain model of aircraft (e.g., Boeing B767-300) Fleet (aircraft) family: A set of aircraft types with the same cockpit configuration and crew qualification requirements Flight leg: An airport-to-airport flight segment Through-flight: Two consecutive flight legs that are flown by the same aircraft Itinerary: A sequence of one or more flight legs Fare class: A particular type of fare restriction Turn time: Minimum time an aircraft needs between its landing and the next take-off WS 13/14 Einführung in Verkehr und Logistik Decision Problem I Assigning aircraft types with different capacity to scheduled flights I Depends on potential revenues, availability, operational costs, equipment capabilities I Too small aircraft = spilled (lost) customers (insufficient capacity) I Too large aircraft = spoiled (unsold) seats (high operational costs per seat) I Challenging task ! complexity of airline schedules and depending airline processes (crew scheduling, maintenance, RM) 23 / 46 WS 13/14 Einführung in Verkehr und Logistik 24 / 46 Fleet Assignment Model (FAM) I Foundation of analytical work in this field I Usually formulated as a mixed integer program (MIP) I Construction as time-space-network Basic FAM4 main constraints 1. Cover constraints - Each flight leg is assigned to exactly one fleet type 2. Balance constraints - Ensure conservation of flow 3. Aircraft availability constraints - The number of available aircrafts of each type bounds their usage 4 Hane et al. (1995). WS 13/14 Einführung in Verkehr und Logistik 25 / 46 5 Time-Space-Network H.D. Sherali et al. / European Journal of Operational Research 172 (2006) 1–30 Station A Station B Stations A2 9:00am Arcs for Type 1 A1 C1 10:00am E2 C2 11:00am E1 Wrap-around arcs for Type 1 B1 12:00pm F1 B2 D1 D1 Time 5 See Arcs for Type 2 Sherali et al. (2005). F2 7 WS 13/14 Einführung in Verkehr und Logistik 26 / 46 Properties I I I I I Focuses on representing flight legs (arcs) Model decides on feasible connections Fleet type-dependent flight/turn-times ! network for each fleet type A directed flight arc belongs to the movement of an aircraft type 3 types of arcs: (1) Ground arcs: represent aircrafts staying at the same station (2) Flight arcs: represent flight legs (3) Wrap-around arcs: connects the last event of the day with the first event of the day I A network time-line is associated with each station I Nodes represent arrivals and departures of a flight leg at a station I Same-every-day fleet assignment ! moderate computational complexity WS 13/14 Einführung in Verkehr und Logistik Sets L set of flight legs indexed i S set of stations indexed s F set of fleet types indexed f Ot̃ set of legs whose time span contains the time point t̃ (e.g. 3 am, for counting purposes) Ts temporal ordered set of departure and arrival time points at station s and time point t̃ Ast leg(s) with arrival time t Dst Pst 2 Ts at station s leg(s) with departure time t 2 Ts at station s the predeccessor time point of t 2 Ts where the predeccessor time point of the first time point in Ts is the last time point in Ts (wrap-around); j Pst j= 1 27 / 46 WS 13/14 Einführung in Verkehr und Logistik 28 / 46 Parameters cfi costs of leg i if covered by aircraft of fleet type f t̃ time point for measuring the number of aircrafts Qf available number of aircrafts of fleet type f Variables xfi = 1 if fleet type f covers leg i (0, otherwise) yfst number of aircraft of fleet type f on the ground of station s at time point t 2 Ts WS 13/14 Einführung in Verkehr und Logistik XX 29 / 46 Objective function: Minimizes total costs min 2 2 cfi xfi f F i L X Cover constraints 8i 2L xfi = 1 f X X Balance constraints yfsp + 2 i Ast xfi 2 i Dst xfi = yfst 8 f 2 F s 2 S t 2 Ts p 2 Pst ; ; ; WS 13/14 Einführung in Verkehr und Logistik X X 30 / 46 Aircraft availability constraint 2 xfi + 2 yfs t̃ Qf 8f 2F s S i Ot̃ Domains of variables xfi yfst 2 f0 1g 0 and integer ; WS 13/14 8f 2 F i 2 L 8f 2 F s 2 S t 2 Ts ; ; ; Einführung in Verkehr und Logistik 31 / 46 Small Example Leg specific data leg i 1 2 3 4 origin o HAM HAM FRA MUC destination d FRA MUC HAM HAM arrival t 410 500 550 620 flight time 50 80 50 80 Cost cfi Fleet fleet f 1 2 3 departure t 360 420 500 540 name A320-200 B737-500 E195 quantity 1 1 1 fleet f 1 2 3 i=1 2700 2800 1900 i=2 3200 3100 2600 i=3 2700 2800 1900 i=4 3200 3100 2600 WS 13/14 Einführung in Verkehr und Logistik 32 / 46 Solution Variable values xfi : I x22 = x24 = x31 = x33 = 1 I x11 = x12 = x13 = x14 = x21 = x23 = x32 = x34 = 0 Objective value Z : I c22 x22 + c24 x24 + c31 x31 + c33 x33 = 10000 Results: I HAM - FRA - HAM: B737-500 ! costs: 2 3100 I HAM - MUC - HAM: E195 ! costs: 2 1900 WS 13/14 Einführung in Verkehr und Logistik Why a simultaneous approach? I Planning processes are usually examined separately from one another I Combining two ore more processes might lead to superior solutions ! cost savings, revenue increase I Integrating RM in FA: Provides the possibility to account for demand variability I Idea: Enable airlines to fit the fleet type to demand on a certain leg I But: Planning horizons differ for both problems (see previous slide) I Difficult to adjust fleet types due to demand changes at short notice ! swapping = short run interchange of already assigned aircrafts ! between two flight legs For simplification we disregard differing planning horizons in the following case study 34 / 46 WS 13/14 Einführung in Verkehr und Logistik 35 / 46 Model assumptions I The model is solved for a predefined (short) period of time ! usually for 1 day (same daily schedule) ! Here: analytic solution according to slides 10-11 I Demand is independent between flight legs I Network revenue Rfi is the solution of a RM model I Aircraft availability is neglected to reflect a strategic decision process I The model decides on the optimal fleet composition for a given network I Result: Optimal assignment of fleet types to legs such that the contributed profit minus fixed costs is maximized WS 13/14 Einführung in Verkehr und Logistik Mathematical formulation Parameters CAf Fixed cost per aircraft of fleet type f CFf Fixed cost of fleet type f B Large number Cf capacity of fleet type f (number of seats) qfi0 optimal quantity of sold seats for fleet type f on flight leg i according to revenue function u(q) qfi Rfi = minfCf ; qfi0 g Profit contribution for fleet type f on flight leg i Rfi = Rfi (qfi ) 37 / 46 WS 13/14 Einführung in Verkehr und Logistik 38 / 46 Variables zf = 1, if fleet type f is chosen (0, otherwise) XX Objective function max 2 2 Rfi xfi f F i L WS 13/14 XX 2 2 f F s S CAf yfs t̃ X 2 CFf zf f F Einführung in Verkehr und Logistik Cover constraints X 2 xfi = 1 39 / 46 8i 2 L f F X X Balance constraints yfsp + 2 i Ast xfi 2 i Dst xfi = yfst 8 f 2 F s 2 S t 2 Ts p 2 Pst ; ; ; WS 13/14 Einführung in Verkehr und Logistik X 40 / 46 Linking constraint yfs t̃ B zf 8f 2 F s Domains of variables xfi zf yfst WS 13/14 2 f0 1g 2 f0 1g 0 and integer ; ; 8f 2 F i 2 L 8f 2 F 8f 2 F s 2 S t 2 Ts ; ; ; Einführung in Verkehr und Logistik Case Study I I I I I Airline: Air Berlin (AB) Period: Monday (Summer schedule 2011) Airports S considered: 4 - HAM, DUS, NUE, STR Number of flight legs L: 42 Fleet-types available: 15 (Do-328 TP, Q-400, CRJ-700, CRJ-900, E175, E195, B737-500, B737-300, A319-100, B737-700, A320-200, B737-800, A321-200, B777-200 BL, B777-200H GW) I Maximum quantity of aircrafts per fleet-type: not restricted I linear PSF: rfi (qfi ) = qmaxi 2 qfi I qmaxi = 200 + " with " a normal distributed random variable with = 0 and = 256 6 To account for variability in demand (Listes and Dekker 2005). 42 / 46 WS 13/14 Einführung in Verkehr und Logistik 43 / 46 Results Airline AB Legs 42 Aircrafts 8 Surplus 76.301 Revenue 205.340 Costs 129.040 Passengers 3.345 I Assumption: all possible fleet types are available and can be chosen ! Results do not reflect the actual AB fleeting I Fleet type 1: 6 aircrafts of CRJ-700 I Fleet type 2: 2 aircrafts of CRJ-900 WS 13/14 Resulting network Einführung in Verkehr und Logistik 44 / 46 WS 13/14 Einführung in Verkehr und Logistik 45 / 46 Literature I Grothklags, S.: Flottenzuweisung in der Flugplanung: Modelle, Komplexität und Lösungsverfahren. Doktorarbeit, Universität Paderborn, Paderborn, 2006. Hane, C.A., C. Barnhart, E.L. Johnson, R.E. Marsten, G.L. Nemhauser und G. Sigismondi: The fleet assignment problem: solving a large-scale integer program. Mathematical Programming, 70(1):211–232, 1995. Klein, R. und C. Steinhardt: Revenue Management, Band 1. Springer Verlag, Heidelberg, 2008. Listes, O. und R. Dekker: A scenario aggregation–based approach for determining a robust airline fleet composition for dynamic capacity allocation. Transportation science, 39(3):367–382, 2005. Sherali, H.D., E.K. Bish und X. Zhu: Airline fleet assignment concepts, models, and algorithms. European Journal of Operational Research, 172(1):1–30, 2006. WS 13/14 Einführung in Verkehr und Logistik Literature II Talluri, K.T. und G. Van Ryzin: The Theory and Practice of Revenue Management. Springer Verlag, 2005. 46 / 46