=Paper=
{{Paper
|id=None
|storemode=property
|title=Airline Fuel Savings Estimation Based on Segmented Fuel Consumption Profiles
|pdfUrl=https://ceur-ws.org/Vol-923/paper06.pdf
|volume=Vol-923
}}
==Airline Fuel Savings Estimation Based on Segmented Fuel Consumption Profiles==
https://ceur-ws.org/Vol-923/paper06.pdf
Airline Fuel Savings Estimation Based on Segmented
Fuel Consumption Profiles
Bruno Marques1, Nuno Leal1
TAP Portugal, Lisbon, Portugal
{bmarques, nleal}@tap.pt
Abstract. Fuel conservation programs are instruments used by airlines to im-
prove operational efficiency and trim fuel related costs. Identifying fuel savings
to accurately manage these programs has always been an issue due to the opera-
tion volatility and lack of reliable data. Advanced data management systems
were developed to support these programs, but having means to identify fuel
savings is still compelling. A new methodology based on segmented fuel con-
sumption profiles is proposed as a tool to accurately identify fuel savings across
periods. This approach allows for a detailed fuel consumption analysis with full
operation coverage.
Keywords: airlines, fuel efficiency, fuel conservation, fuel consumption pro-
files
1 Introduction
Airlines today struggle to survive in a highly competitive market, facing high operat-
ing costs and being seriously affected by the global economic and financial crisis.
Market deregulation and growth of low-cost carriers have since late 90's reinforced
the need to improve and follow operational costs and finding means to reduce them.
From the airlines’ cash operating costs, fuel represents the highest block of direct
operating costs, having had a dramatic increase in the latest years [1, 2]. Additionally,
fuel prices constant fluctuation represents a challenge to the airlines.
Besides the high fuel costs, there are emerging global environmental concerns to
reduce carbon emissions from the aviation industry. Despite the fact that aviation
transport system energy intensity continues to decline due to more efficient aircraft,
engines and general procedures, air travel continues to experience the fastest growth
of all transport modes. Air transport industry was responsible in 2005 for approxi-
mately 2.5% of total anthropogenic carbon emissions [3]. However, aviation’s relative
contribution to climate change is presumably higher due to the fact that the majority
of emissions are produced at high altitudes [4]. If no additional fuel efficiency
measures are adopted, this contribution may grow up to 15% if this industry keeps the
growth pace at around 5% per year.
Due to these environmental concerns and the need to reduce aviation carbon emis-
sions, there were several commitments from different entities to approach the global
warming issue and define strategies to reduce carbon footprint. IATA (International
Air Transport Association) proposed a four-pillar strategy for carbon neutral growth
from 2020 [5]. Despite the fact that today’s aircraft are 70% more efficient than first
jet-era aircraft [6], Technology, mainly related to aircraft and engine manufacturers
new solution developments, as well as biofuels, has the best prospects for reducing
aviation carbon emissions. Infrastructure, through improvements on air traffic man-
agement and airport infrastructure, is a major opportunity and may provide 4% emis-
sion reduction by 2020. A more efficient combination of air traffic management and
airline procedures can reduce by 30% the typical descent fuel burn [7]. Economic
Measures can prove to be another mechanism that can contribute to reduce aviation
carbon emissions. The fourth and last pillar, Efficient Operations, mainly airline’s
responsibility, is potentially the one that can result in immediate carbon emissions
reductions.
Airlines, aiming at mitigating fuel cost and carbon emissions have developed for
several years extensive fuel conservation programs with numerous initiatives covering
areas as flight and airport operations procedures, aircraft weight reduction or engine
and aircraft washing [8]. These initiatives, all together, have already contributed to a
general operational efficiency improvement. As some of these initiatives come at cost,
it is critical to properly analyze the impact of such implementation. It is also vital to
monitor across time the adherence to these initiatives, as well as perform periodic
assessments of fuel burn reductions to continuously evaluate program’s operational
impact. Airlines soon realized that up-to-date, reliable operational data is imperative
to complete these tasks.
2 Problem description
Fuel conservation programs developed by airlines typically integrate all key areas of
operations that affect fuel usage having as main objectives trimming annual fuel costs
and increasing operational efficiency. Many airlines’ fuel conservation programs have
failed because of the lack of automated data feeding fuel analysis and dashboards to
cross-departmental stakeholders [9]. On cross-departmental programs like these it is
fundamental to monitor the implemented initiatives’ performances, as well as the
overall performance.
In terms of fuel conservation, the lack of useful, readily available and reliable rec-
orded data is a crucial issue. The airline operation, highly dependent on procedures
and checklists, usually generates large amounts of data that is typically spread out
through the departments. Therefore, data availability has always been one of the big-
gest issues in the fuel conservation programs.
To solve this data availability and quality issues, airlines, such as TAP Portugal,
invested substantial effort in developing complex database systems that could concen-
trate information from different sources, and could provide means to perform smart
fuel consumption analysis.
Despite the significant improvements in data availability, historically, the evalua-
tion of fuel savings within the airlines has always been a tough task due to the con-
stant changes in the airline’s operation. Airline’s operation is highly dynamical not
only in terms of operated routes, but also in operated aircraft or aircraft types. Differ-
ent aircraft within a fleet have different fuel consumption profiles due to distinct aer-
odynamic and weight characteristics, but also fuel performance degradation. Also,
different aircraft operating in different routes have distinct fuel consumption profiles.
As the operation constantly changes between periods, one needs to look at various
aggregation levels to properly compare the fuel consumption between periods, ensur-
ing that all the flights are being taken into account.
2.1 Fuel Efficiency parameter
To identify airlines’ fuel consumption reductions, one needs to compare typical fuel
consumption profiles between periods. As there isn’t a single fuel efficiency parame-
ter that is suitable to all applications, airlines need to identify the one that best suits its
aircraft operations and available data. Aircraft fuel consumption varies significantly
with flight time and aircraft performance degradation, but is also influenced by car-
ried weight, en-route winds, flown route profiles or en-route and airport congestion.
Generically commercial aviation fuel burn is function of two key factors: aircraft fuel
efficiency – which stands for the amount of productivity delivered by the aircraft
through the usage of fuel energy and; operational factors – that comprises mass load
factors, airline and air traffic control inefficiencies [10]. While this could be translat-
ed as Fuel Burned per ASK (Available Seat Kilometer) or Fuel Burned per PKU (Pas-
senger Kilometer Used) for passenger airlines, Fuel Burned per TK (Tonne Kilome-
ter) may be most appropriate to cargo carriers. On aircraft design stage one of the
most popular parameters if Fuel Burned per ATK (Available Tonne Kilometer) [11].
However, when the objective, more than representing an efficiency parameter, is to
quantify the fuel savings obtained by the initiatives under the fuel conservation pro-
gram, one needs to identify a variable or set of variables that corrects fuel burn from
quantifiable effects that influence fuel consumption and that are not linked with fuel
conservation program initiatives. The fuel efficiency parameters above mentioned
depend on the great circle distance (GCD) between two airports that is constant
throughout time. Therefore the effect of different routes flown by aircraft between
two airports is not properly taken into account. To adequately evaluate the fuel sav-
ings between periods, instead of using GCD, it is recommended to use flown hours to
normalize fuel consumption. Additionally, as a flag carrier’s typical operation is not
only carrying passengers but also cargo, and as carried weight also plays an important
role in the aircraft’s fuel consumption, it is necessary to correct fuel burn from differ-
ences on carried weight between periods. Therefore, on top of normalizing the fuel
consumed by flown hours, one can use the payload variable to additionally correct
fuel burn.
In order to adequately compare consumed fuel differences between periods it is re-
quired to identify the quantifiable parameters that can be used to define a fuel effi-
ciency profile illustrative of the operation.
2.2 Information architecture
The data analyzed in the fuel consumption analysis has the granularity of a single
flight. For every single flight, numerous parameters are recorded and collected, each
having its own source. Flight duration can be reported automatically by the aircraft
systems or by airport ground handling agents but can be also reported by flight crews
in their debriefing procedure. Additional flight information, as carried passengers or
payload are reported by crew or ground handling agents in different formats and tim-
ings. Fuel figures are logged by the flight crew members in the debriefing procedure.
All these figures may be reported by a totally different process when the flights are
operated in a wet-lease basis from another operator.
The diversity of data sources available (associated with a multiplicity of processes
to obtain the same data) can cause gaps in each individual measure, making that flight
unusable (despite the fact that several other measures associated with that same flight
can be available). This means that it is understandable and acceptable that the infor-
mation produced from this data uses a sample size usually above 90%, but away from
the complete set.
The operational systems are the core data source, where all processes, being auto-
matic or manual, end up providing values. Some values are critical for the operational
system (precise weight estimates are crucial to feed the flight plan generator) and
others are just statistical (fuel consumption).
An important part of transforming data into information is related to the building of
a standard data warehouse (DW), where data from several sources is collected and
made available as a single record, like schematically described in Fig. 1.
Automatic Manual
System #1 System #2 System #3 System #4 System #5
Flight
Duration
Weight
DW
Fuel Distance
Burned
Fig. 1. Data flow, from users and systems to corporative data warehouse
The diversity of processes also implies that information will be produced based on
data with different quality levels. Automatic processes are usually more reliable than
manual ones, since manually recording (in a paper document) and further reading
might more likely lead to insertion errors (when no feedback is given to the user about
the correctness of the value) or interpretation errors (namely calligraphy issues or
unclear values). Data quality can be checked while loading data into the DW, search-
ing physically impossible values, or values that fail expected correlations between
several measures (for instance, fuel consumption per flown hour on a single flight
should be coherent with the used aircraft average value). These checks can lead to two
distinct consequences: the identification of flights that need to be further examined, or
identified as having “no good” information that are rejected on the analysis process.
The data quality process should mainly focus on the data input processes of the op-
erational systems, both in manual and automated scenarios. This continuous evolu-
tion, involves migrating manual processes into automatic ones, providing users im-
mediate feedback on entered values, creating simple boundaries for data input and
providing systems or forms that are clear and less error prone.
These systems continuous improvement is mandatory to proficiently analyze fuel
consumption, identify improvements on fuel savings and pinpoint areas that require
additional work.
3 Fuel savings model
As previously stated, quantifying fuel savings within airline’s operations has histori-
cally been a demanding task, not only due to the dynamic airlines’ operations, but
also to the lack of required data, both in quantity and quality. While this latter hurdle
has been addressed by investments in capable information systems, the comparison of
operation and fuel consumption profiles still represents a challenge. The proposed
solution attempts to minimize the potentially misleading effects of computing fuel
savings between two periods that have distinct sets of flights.
The proposed model is based on the identification of fuel consumption profiles that
are representative of consumed fuel in distinct periods. These period-characteristic
profiles are the basis to recognize equivalent fuel consumption amongst periods.
When pinpointing the fuel savings for a period compared with a reference, pre-
computed reference period profiles are used to extrapolate what would be the fuel
consumption if the operational reference period characteristics would still rule.
In a nutshell, the solution explores the process described here briefly. Error! Ref-
erence source not found. provides an example of operation data, for two consecutive
years. The different values obtained in the variation of each parameters shows distinct
change rates, meaning that changes from one year to the other are not linear.
Table 1. Basic example data, set with two years for comparison
Avg Flight Fuel Burned
Distance Fuel Burned
Year # Flights Distance per Distance
Flown (km) (ton)
(km) (kg/km)
N–1 100 350,000 3,500 1,100 3.14
N 150 585,000 3,900 1,500 2.56
Variation(%) +50% +67% +11% +36% -18%
In a simplified approach, as an example, if the fuel profile function is defined as
fuel burned per distance, it would produce the values presented in the last column of
Error! Reference source not found.. When the year N fuel profile is applied to the
Year N flown miles, as expected, the result is 1,500 ton. If the year N – 1 fuel profile
is applied to the Year N operation flown miles, the value 1,838.5 ton is obtained. This
value can be read as the Year N fuel consumption, if the consumption profile had not
change from year N – 1. The delta between both values (+338.5 Kg) represents the
additionally spent fuel from one year to the other. This methodology enables the
comparison of equivalent fuel consumption profiles characteristic of different periods
and by using this, allows the identification of differences in fuel efficiency. This ex-
ample tries to set the path for the two major improvements that this process can bene-
fit from:
• Fuel Profile Function – In the example, the distance was used, but other
flight information can be used to define a profiling function that reflects the
fuel consumption behavior in the compared periods.
• Flight Segmentation – In the example, an overall profile was used, applied
to both periods. But knowing for instance, that long haul flights have differ-
ent fuel consumption behavior when compared with medium haul flights,
can lead to flight segmentation, originating two segments with distinct fuel
profile functions.
The fuel consumption calculation using the generated fuel profile functions is a
two-step process: firstly identify for each flight the applicable segment and fuel pro-
file function for the considered periods; secondly, with flight information as flight
time, carried weight, aircraft performance, compute fuel consumptions for each period
using the applicable fuel profile functions. The difference between the extrapolated
reference period fuel consumption and the actual period calculated fuel consumption
represents the amount saved or additionally burned compared to the reference period.
A schematic representation of the process is found on Figure 2.
Flight
Flight
Reference Period
Flight Time Profiles
FlightProfiles
Profiles
Weight
Route Fuel Consumption
Used Aircraft Considering Reference
Period Profile
….
Flight
Δ= fuel saved
Information
or additionally
Current Fuel
Consumption burned
Flight
CurrentFlight
Period
Profiles
FlightProfiles
Profiles
Fig. 2. Fuel savings calculation process
3.1 Fuel Profile Function
Flight fuel consumption varies with the aircraft used, but also depends on physical
variables like flight duration, carried weight, wind speed and direction, and other
daily features like weather, used routes and both airport and en-route congestion.
Although all these variables have an impact on the aircraft fuel consumption, their
contribution to the final figures is different. Flight duration, for instance, has a much
larger contribution than aircraft performance. On top of this there are variables that
are harder to estimate and to quantify their impact on fuel consumption.
In order to properly evaluate the fuel consumption savings, it is mandatory to iden-
tify a fuel efficiency parameter that can best represent the operation. The selection of
variables used in this parameter should be the ones that have larger impact on fuel
consumption and have available data.
Considering the profile function as the ratio between fuel consumed on a flight and
a variable, or a set of variables, the accuracy of several different considered alterna-
tives is presented on Table 2. In this table the absolute error is the difference between
the real fuel consumption and the estimated fuel consumption, calculated on a flight-
by-flight basis, using the calculated fuel profile functions.
Table 2. Average absolute error using several possible parameters
% Absolute Error
Variables Used
(Average Flight Fuel)
Carried Weight 22.34%
Airport Distance 7.17%
Flown Hours 5.47%
Flown Hours × Carried Weight 3.49%
Flown Hours × Carried Weight × Aircraft Performance 3.35%
As expected, when considering using one variable profile function, the smallest er-
ror is obtained when using flown hours, since it is the parameter with larger impact on
fuel consumption. The combination of flown hours and carried weight reduces further
the error, as the weight also has also a significant impact on fuel consumption. On top
of this, when considering aircraft performance, the error is minimized, as the actual
aircraft and engine performance is taken into account.
When quality data is available, the selected fuel profile function should be the one
that provides the minimum error in estimating aircraft fuel consumption. Although the
model is flexible to use these, or any other parameters, as on the current model there
is reliable data available to calculate the fuel burned per flown hours per carried
weight per aircraft performance, this will be the fuel profile function used since it is
the one with minimum error. This fuel profile function will be then computed for each
considered flight segments representative of the period’s operation.
Flight Segmentation
The amount of segments to be used is a difficult decision. When the number of seg-
ments used is increased, the profile calculation quality is also enhanced, as it de-
scribes a more specific type of flights. On the other hand, a higher number of seg-
ments mean a smaller sample of flights used for each profile calculation, leading to a
greater impact caused by outlier flights. Table 3 provides a comparative analysis for
several segmentation approaches taking into consideration one year of operation.
Table 3. Comparative analysis of flight segmentation
Aircraft
Model Aircraft
All Haul Model Aircraft Route
Route Route
Quarter
# Segments 1 2 12 90 900 6,000 17,000
# Elements
100,000 53,000 9,000 1,200 120 20 6
/Segments
Outlier Weight 3% 3.6% 3.9% 4.1% 4.7% 4.9% 6.3%
On segments with a lower sample size, outlier’s weight can be more misleading, so
defining a minimum required sample size can help to exclude segments that may
cause distortion on the final figures. On the current model a minimum of 10 elements
per segment is required.
One of the biggest issues concerning this subject is the fact that there are many var-
iables, as aircraft or routes changing from one period to another. When lower aggre-
gation levels are used, finding common segments in the analyzed period can be chal-
lenging. For instance, when an aircraft-route aggregation level is selected, if a route is
operated in one period, but not on the other, a comparison at the same level is not
feasible. In such circumstance a higher aggregation level that enables a comparison
between the two periods must be used. Table 4 provides a comparative analysis of the
flight coverage rate between two distinct periods. For example, when using a model
segmentation approach, a flight is considered uncovered when the aircraft model used
in one period was not operated in the other one.
Table 4. Flight segmentation comparative coverage rate
Aircraft
Model Aircraft
All Haul Model Aircraft Route
Route Route
Quarter
Profile
100% 100% 99.7% 97% 93% 89% 85%
Coverage
Since the solution should provide a comparison between all the operated flights
(and not only the ones that share a similar profile with the reference period being
used), an extra qualifying step is needed on the segment selection. This extra step
requires the computation of more profiles, not only on the considered detail level, but
also on upper aggregation levels. This way, when a profile is missing on a more de-
tailed level, a less detailed profile can be used to ensure that all the flights are cov-
ered, like shown on Figure 3.
Yes
All Profile exists at this level? Profile
. No
.
Yes
Aircraft
. Profile exists at this level? Profile
. No
.
Yes
Aircraft + Route
. + Weekday Profile exists at this level? Profile
Fig. 3. Flowchart for profile match
On the current model, the lowest segmentation used is obtained by splitting the set
of flights by aircraft and route.
4 Information Analysis and Visualization
One of the main purposes of this fuel savings calculation method is to be able to cal-
culate fuel savings on a per-flight basis. This means that, at the bottom line, it is pos-
sible to identify how much fuel would have been consumed on the same flight if ref-
erence period’s conditions were still valid, being the difference between actual and
estimated reference period values, the amount of fuel saved or additionally burned.
Having information at this level, provides the ability to give a greater insight on how,
when and why fuel consumption changes are happening. Table shows some of the
potential analysis that can be performed from the generated information.
Table 5. Possible analysis for data visualization
• Changes in airline fleet configuration
Aircraft
• Aircraft / engine performance degradation
• Changes in airline network configuration
Route • Flight planning routing (planned and flown)
• En-route and airport congestion
• Low season / High season operation
• Operational fuel saving measures (before and after)
Calendar
• Operational unexpected events (namely, weather condition, volcan-
ic eruption, strikes)
5 Conclusions and Future Work
Quantifying fuel savings has been a challenge that airlines face as they seek new ways
to improve fuel efficiency. The proposed methodology ensures that an adequate com-
parison between periods is achievable through the usage of a multi-stage aggregation
levels approach. Defining a suitable fuel efficiency profile varies from airline to air-
line and greatly depends on the available data. This methodology provides total flexi-
bility on the fuel efficiency profiles used as well as the aggregation levels considered
in the calculation. Obtained results demonstrate that full coverage of operation is fea-
sible, allowing a complete fuel efficiency comparison across periods with distinct
operation. The generation of fuel savings data on an aircraft, or aircraft-route basis
provides a step change in the typical fuel savings analysis, giving room to identifying
trends and spotting changes in the airline’s operations.
The described solution is generic enough to be easily adapted to other domains
where the problem of comparing performance needs to be calculated over changing
operational scenarios.
The solution is sensible to data volume and quality. As described, when segmenta-
tion goes to a more detailed level the average size of each of the segments drops,
making outlier records more relevant in the profile function definition, causing larger
deviations. All the improvements that benefit data quality will also benefit the quality
of the fuel savings estimates.
A systematic and reusable analysis process still needs to be defined over the ob-
tained set of information, in order to increase the visibility of emerging problems and
provide correct savings for specific measures decided by the company.
6 Bibliography
1. Morrison, J.K.D. Sgouridis, S., Hansman R.J.: Game Theory Analysis of Aircraft
Manufacturer Innovation Strategies in the Face of Increasing Airline Fuel Costs. Mas-
sachusetts Institute of Technology Internation Center of Air Transport, United States
of America (2011).
2. Srivastava, A.N.: Greener Aviation with Virtual Sensors: a case study. Data Mining
and Knowledge Discovery. 24, 443-471 (2012)
3. Lee, D.S., Fahey, D.W., Forster, P.M., Newton, P.J., Wit, R.C.N., Lim, L., Owen,
B., Sausen, R.: Aviation and global climate change in the 21st century. Atmospheric
Environment. 43, 3520-3537 (2009).
4. Solomon, S., Qin, D., Manning, M., Chen, Z., Marquis, M., Averyt, K.B., Tignor,
M., Miller, H.L.: Contribution of Working Group I to the Fourth Assessment Report
of Intergovernmental Panel on Climate Change. Cambridge University Press, Cam-
brige (2007).
5. IATA: A global approach to reducing aviation emissions. Switzerland (2009).
6. Lee, J.J.: Historical and Future Trends in Aircraft Performance, Cost and Emis-
sions. Massachusetts Institute of Technology, United States of America (2000).
7. Knorr, D., Chen, X., Rose, M., Gulding, J., Enaud, P., Hegendoerfer, H.: Estimat-
ing ATM Efficiency Pools in Descent Phase of Flight – Potential Savings in both
Time and Fuel. 9th USA/Europe Air Traffic Management Research and Development
Seminar. Berlin, Germany (2011).
8. Lemelle, Y., Docus, M., Fueri, M.: Operating procedures for cost savings. 16th
Airbus Performance and Operations Conference. Paris, France (2009)
9. Falise, F.: From emissions management to fuel conservation. ERA Regional Inter-
national. August 2011, pp. 10. United Kingdom (2011).
10. Yutko, B.: Approaches to Representing Aircraft Fuel Efficiency Performance for
the Purpose of a Commercial Aircraft Certification Standard. Massachusetts Institute
of Technology International Center of Air Transport (2011).
11. Economon, T.D., Copeland, S.R., Alonso, J.J., Zeinali, M., Rutherford, D.: Design
and Optimization of Future Aircraft for Assessing the Fuel Burn Trends of Commer-
cial Aviation. 49th AIAA Aerospace Sciences Meeting. Orlando, United States of
America (2011).