Академический Документы
Профессиональный Документы
Культура Документы
a r t i c l e i n f o a b s t r a c t
Keywords: Reducing greenhouse gas emissions in freight transportation requires using appropriate
Fuel consumption emission models in the planning process. This paper reviews and numerically compares
Emission models several available freight transportation vehicle emission models and also considers their
Vehicle routing outputs in relations to field studies.
Freight transportation modeling
Ó 2011 Elsevier Ltd. All rights reserved.
1. Introduction
Green logistics, aiming at minimizing the harmful effects of transportation on the environment, has gained in importance.
In particular, an explicit consideration is given to reducing levels of CO2 through better operational level planning. Measuring
and reducing emissions requires good estimations to be fed into planning activities, which in turn require estimation models
to be incorporated into the planning methods.1 The choice of the type and the nature of emission functions are important for
deriving accurate estimates in the planning of transportation activities.
There exists a variety of analytical emission models that differ in the ways they estimate fuel consumption or emis-
sions, or in the parameters they take into account in the estimations. In this context, Ardekani et al. (1996), for example,
divide models into urban (vehicle speed is less than 55 km/h) and highway (vehicle speed is at least 55 km/h) fuel con-
sumption models, whereas Esteves-Booth et al. (2002) consider three types of emission models based on emission fac-
tors, average speeds, and mode. Here we compare a number of such models, and assess their respective strengths and
weaknesses.
An instantaneous fuel consumption model, or instantaneous model for short, is developed by Bowyer et al. (1985) as an
extension of Kent et al.’s (1982) power model. It uses vehicle characteristics such as mass, energy, efficiency parameters,
drags force and fuel consumption components associated with aerodynamic drag and rolling resistance, and approximates
⇑ Corresponding author.
E-mail address: T.Bektas@soton.ac.uk (T. Bektasß).
1
For example, Jabali et al. (2009) investigate reducing CO2 emissions in road-based freight transportation, Bauer et al. (2010) consider emissions in rail
transportation, and Fagerholt et al. (2010) look at reducing emissions in shipping.
1361-9209/$ - see front matter Ó 2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.trd.2011.01.011
348 E. Demir et al. / Transportation Research Part D 16 (2011) 347–357
the fuel consumption per second. The model assumes that changes in acceleration and deceleration levels occur within 1 s
time interval and takes the form
(
a þ b1 Rt v þ ðb2 Ma2 v =1000Þ for Rt > 0
ft ¼ ; ð1Þ
a for Rt 6 0
where ft is the fuel consumption per unit time (mL/s), Rt is the tractive force (kN = kilonewtons) required to move the vehicle
and calculated as the sum of drag force, inertia force and grade force as Rt = b1 + b2v2 + Ma/1000 + gMx/100,000. Further-
more, a is the constant idle fuel rate (in mL/s, typically between 0.375 and 0.556), b1 is the fuel consumption per unit of
energy (in mL/kJ, typically between 0.09 and 0.08), b2 is the fuel consumption per unit of energy-acceleration (in mL/
(kJ m/s2), typically between 0.03 and 0.02), b1 is the rolling drag force (in kN, typically between 0.1 and 0.7), and b2 is the
rolling aerodynamic force (in kN/(m/s2), typically between 0.00003 and 0.0015). In addition, x is the percent grade, a is
the instantaneous acceleration (m/s2), M is the weight (kg), and v is the speed (m/s).
Using Model 1, the amount of fuel consumption Ft(mL/s) for a journey of duration t0 can be calculated as:
Z t0
Ft ¼ ft dt: ð2Þ
0
The instantaneous model operates at a microscale level and is better suited for short trip emission estimations. The model
does not make use of macro-level data such as the number of stops but is able to take into account acceleration, deceleration,
cruise and idle phases. Using data from special on-road experiments in Melbourne, Bowyer et al. (1985) showed it is able to
approximate fuel consumption of individual vehicles within a 5% error margin for short trips, and dynamometer tests sug-
gested that its accuracy is within 10% for a variety of on-road experiments (Esteves-Booth et al., 2002).
A four-mode elemental model is described by Bowyer et al., in a refinement of Akçelik (1982) estimates fuel consumption
for idle, cruise, acceleration and deceleration. The model includes the same parameters as Model 1 but introduces new con-
siderations such as initial speed, final speed and energy-related parameters. It requires data related to the distance, cruise
speed, idle time and average road grade as inputs. A vehicle is said to be in an idle mode when the engine is running but
the speed is below 5 km/h. More accurate estimations can be made if the initial and final speeds for each acceleration
and deceleration cycles are known. The model consists of four functions, Fa, Fd, Fe and Fi, corresponding to fuel consumption
estimations (mL) for acceleration, deceleration, cruise and idle modes.
where Ek denotes the change in kinetic energy per unit distance during acceleration and is calculated as
p
Ek ¼ 0:3858104
v 2f v 2fi =xa . The integration coefficients are k1 = 0.616 + 0.000544vf 0.0171 vi and k1 = 1.376 +
0.00205vf 0.0053vi. When the travel distance xa and the travel time ta are not known, they can be estimated as xa = ma(-
p
vi + vf)ta/3600 where ma = 0.467 + 0.00200vf 0.00210vi and ta = (vf vi)/(2.08 + 0.127 (vf vi) 0.0182vi. C is the function
parameter (in mL/km, typically between 21 and 100), and B is the function parameter – in (mL/km)/(km/h)2 and typically
between 0.0055 and 0.018).
where fi denotes the idle fuel rate (mL/h), vc is the average cruise speed (km/h), and xc is the travel distance (km). The change
in positive kinetic energy per unit distance during the cruise mode is calculated as Ek+ = max{0.258–0.0018vc, 0.10} and the
other parameters are set to kE1 ¼ maxf12:5=v c ; þ0:000013v 2c ; 0:63g, kE2 = 3.17, and kG = 1–2.1 Ek+ for x < 0, and 1–0.3 Ek+ for
x > 0 with kE1, kE2 and kG the calibration parameters.
The elemental model assumes minimum loss of driving information, and hence minimum loss of accuracy in fuel con-
sumption estimates. It is better suited for estimation of fuel consumption for short distance trips, but its large number of
parameters and the existence of four functions can make it difficult to implement. Bowyer et al. experimented with Model
2 and compared it against the instantaneous model. Their results suggest that the elemental model can predict fuel con-
sumption within a 1% error margin. If the initial and final speeds are known, the model yields more accurate estimates
for fuel consumption, and provides results very similar to those of the instantaneous model.
The running speed fuel consumption model is an aggregated form of the elemental model and was introduced by Bowyer
et al. (1985) to calculate fuel consumption during periods when a vehicle is running and is in an idle mode. The model is:
n o
F s ¼ max ati þ ðfi =v r þ c þ Bv 2r þ kE1 b1 MEkþ þ kE2 b2 ME2kþ þ 0:0981kG b1 MxÞxs ; ats ; ð8Þ
where Fs is the fuel consumption (mL), xs is the total distance, vr is the average running speed (km/h), ts and ti the travel and
idle time. Average speed is calculated as ts = 3600xs/(ts ti). Furthermore, Ek+ = max{0.35–0.0025vr, 0.15}, k1 = max{0.675–
1.22/vr, 0.5}, k2 = 2.78 + 0.0178vr.
The model is an extension of the instantaneous model and can be viewed as an aggregation of the elemental model. Accel-
eration, deceleration and cruise modes are considered together within a single function. It does not take into account the idle
mode of a vehicle, but it can be used to estimate fuel consumption in a variety of traffic situations, ranging from short to long
distance trips, although it is more useful for the latter.
A comprehensive emissions model for heavy-good vehicles was developed in Barth et al. (2000, 2005) and Barth and
Boriboonsomsin (2008). It follows, the model of Ross (1994), consisting of three modules; engine power, engine speed
and fuel rate.
Ptract ¼ ðMa þ Mg sin h þ 0:5C d qAv 2 þ MgC r cos hÞv =1000; ð9Þ
3
where v is the speed (m/s), and M is the weight (kg), with q is the air density in kg/m (typically 1.2041), A is the frontal
surface area in m2 (typically between 2.1 and 5.6), and g is the gravitational constant in m/s2 (typically 9.81). In addition,
Cd is the coefficient of aerodynamic drag (typically 0.7), and Cr the coefficient of rolling resistance (typically 0.01). To trans-
late the tractive requirement into engine power requirement, the following is used:
where P is the second-by-second engine power output (kW), gtf is the vehicle drive train efficiency (typically 0.4), and Pacc the
engine power demand associated with running losses of the engine and the operation of vehicle accessories such as usage of
air conditioning (typically 0).
where N is the engine speed (in rpm, typically between 16 and 48), S is the engine-speed/vehicle-speed ratio in top gear Lg,
R(L) is the gear ratio in gear L = 1, . . . , Lg, v is the vehicle speed (m/s), and g is the efficiency parameter for diesel engines
(typically 0.4).
2.5. Model 5: Methodology for calculating transportation emissions and energy consumption (MEET)
Hickman et al. (1999) work on emission factors for road transportation (INFRAS, 1995) describes a methodology called
MEET, used for calculating transportation emissions and energy consumption for heavy-good vehicles. This methodology in-
cludes a variety of estimating functions, which are primarily dependent on speed and a number of fixed and predefined
parameters for vehicles of weights ranging from 3.5 to 32 tonnes. For vehicles weighing less than 3.5 tonnes, the fuel con-
sumption is estimated using a speed dependent function of the form e = 0.0617v2 7.8227v + 429.51. For other classes of
vehicles, MEET suggests the use of:
GC ¼ A6 v 6 þ A5 v 5 þ A4 v 3 þ A2 v 2 þ A1 v þ A0 ; ð14Þ
where GC is the road gradient correction factor. The following is used to take the load factor into account:
2.6. Model 6: Computer programme to calculate emissions from road transportation (COPERT) model
COPERT was developed by Ntziachristos and Samaras (2000) and estimates emissions for all major air pollutants as well
as greenhouse gases produced by various vehicle categories (e.g., passenger cars, light duty vehicles, heavy duty vehicles,
mopeds and motorcycles). Similar to Model 5, it uses a number of functions, which are specific to vehicles of different
weights, to estimate fuel consumption. For example, the function for a vehicle with weight less than is 0.0198v2 2.506
v + 137.42.
The model is also based on on-road measurements, like Model 5 but does not take road gradient and acceleration into
account, but t can differentiate between two speed ranges for each vehicle class.
3. Simulations
Fuel consumption depends on a number of factors that can be grouped into four categories: vehicle, driver, environmental
conditions and traffic conditions. Using three of these four categories, Table 1 compares comparison of the six models.
Driver-related factors are difficult, if not impossible, to integrate into estimation models. The table shows that all models
consider vehicle load, speed and acceleration, although the way in which they incorporate them varies, especially for vehicle
load. Models 1–4 are similar in their consideration of detailed and technical vehicle-specific parameters, such as vehicle
shape (frontal area), and road conditions (gradient, surface resistance), while 5 and 6 present simpler estimations through
a predefined set of parameters for a number of vehicle classes. Model 5 is, to some extent, able to take into account factors of
E. Demir et al. / Transportation Research Part D 16 (2011) 347–357 351
Table 1
Comparison of Models 1–6 regarding factors affecting fuel consumption.
load and gradient through the correction factors, but this is not so for 6. No model explicitly considers driver-related factors
or some vehicle related factors such as transmission type, or tire pressure, largely because quantifying such them is difficult.
Numerically experiments are used to compare the models under different scenarios. In all the experiments, we assume a
single vehicle driven on a 100 km road, and vary vehicle speed, load, acceleration and road gradient.
Vehicle speed: Countries impose different restrictions on driving speed and here the lower and upper speed limits are set
to 20 km/h and 110 km/h.
Vehicle load: The gross vehicle weight rating (GVWR) is the maximum allowable mass of a road vehicle or trailer when
loaded, including the weight of the vehicle itself plus fuel, passengers, cargo, and trailer weight. Commercial trucks are
usually classified; Classes 1 and 2 or referred light duty, 3–5 as medium duty, and 6–8 as heavy duty. Here we consider
a vehicle from each group. The load factors used for light duty vehicles are 0% (unloaded), 10% and 20%. The load factors
used for medium duty vehicles are 0%, 15% and 30%. Finally, the load factors used for heavy duty vehicles are 0%, 30%, 60%
and 90%.
Acceleration: While there are two types of acceleration: average acceleration which denotes the change in velocity divided
by the change in time and instantaneous acceleration which corresponds to the acceleration at a specific point in time, i.e.
only consider the latter.
Road slope: The gradient of a road affects the resistance of a vehicle to traction, as the power employed during the driving
operation determines the amount of fuel consumption. Road gradient factors are set to ±0.57 and ±1.15 degrees for the
road.
The experiments are based on scenarios generated by varying values of these four parameters and are summarized in
Table 2. For scenarios 1–14, there are ten possible speed values, ranging from 20 km/h to 110 km/h in increments of
10 km/h, as well as the three types of vehicle (i.e., light, medium, heavy) giving 432 possibilities. Only selections from this
set are presented. In scenarios 15–18, vehicle speed is kept constant at either 50 km/h or 70 km/h, but load is gradually
changed from 0% to 10% in Scenarios 15 and 17 and from 0% to 30% in the other two.
Models 1–3 give fuel consumption in mL per time or distance. Model 4 estimates fuel consumption in gram fuel per time
or distance, and Models 5 and 6, CO2 emissions in grams per unit of distance. For comparative purposes, these outputs are
converted to the estimated fuel usage (in L) for the 100 km road segment.
3.1. Results
Tables 3–5, present results for scenarios 1–14 in Table 3 for three levels of speed: 50 km/h, 70 km/h and 100 km/h and
provide, for each scenario, the fuel consumption (L) estimated by each model.
From Tables 3–5, it can be seen that there is a considerable increase in fuel consumption with respect to the changes in
vehicle speed, with Model 1 the most sensitive among all those tested. With this model, the difference in fuel requirements is
about 146% when speed is increased from 50 km/h to 100 km/h; whereas Models 2 and 4 show very similar results for each
of three speed levels. The models based on on-road measurements, Model 5 and 6 yield similar fuel consumption.
Scenarios 1–3 show that fuel consumption depends on vehicle load with all models sensitive to changes in load and accel-
eration. Models 1–4 are also sensitive to changes in deceleration rates, but this less so for Models 5 and 6. Similar conclusions
can be made for changes in road gradient; all models, with the exception of 6, show an increase in fuel consumption when
there is an increase in gradient.
352 E. Demir et al. / Transportation Research Part D 16 (2011) 347–357
Table 2
Setting of parameters in the 18 predefined scenarios.
Scenario Speed (km/h) Load (kg) Acceleration (km/h/s) Road gradient (degrees)
1 20–110 0% 0 0
2 20–110 15% 0 0
3 20–110 30% 0 0
4 20–110 15% 0.01 0
5 20–110 15% 0.02 0
6 20–110 15% 0.01 0
7 20–110 15% 0.02 0
8 20–110 15% 0 0.57
9 20–110 15% 0 1.15
10 20–110 15% 0 0.57
11 20–110 15% 0 1.15
12 20–110 15% 0.01 0.57
13 20–110 15% 0.01 0.57
14 20–110 15% 0.01 0.57
15 50 0–10% 0 0
16 50 0–30% 0 0
17 70 0–10% 0 0
18 70 0–30% 0 0
Table 3
Fuel consumption with speed of 50 km/h for scenarios 1–14.
Table 4
Fuel consumption with speed of 70 km/h for scenarios 1–14.
Table 5
Fuel consumption with speed of 100 km/h for scenarios 1–14.
Table 6
Fuel consumption for scenarios 15–18.
Fig. 1. Fuel consumption for three types of vehicle under various speed levels estimated in Model 4.
354 E. Demir et al. / Transportation Research Part D 16 (2011) 347–357
consumption for a medium duty vehicle, light and heavy vehicles are not considered because they exhibit similar patterns in
terms of fuel consumption, with only actual consumption values being different, indicated that vehicle weight has a signif-
icant the effect on fuel consumption are similar to that seen in Fig. 1. Comparing Models 2 and 4, the latter is more sensitive
to the changes in load. Optimal vehicle speed turns out to be around 55 km/h for an unloaded medium duty vehicle using
Models 2 and 4.
Fig. 2. Fuel consumption with a 0.01 m/s2 acceleration as estimated by Models 2 and 4.
Fig. 3. Fuel consumption with a 0.01 m/s2 deceleration as estimated by Models 2 and 4.
E. Demir et al. / Transportation Research Part D 16 (2011) 347–357 355
2
Model 2 does not allow a direct input of resistance and drag as parameters in the estimation.
356 E. Demir et al. / Transportation Research Part D 16 (2011) 347–357
Table 7
Effect of changes in rolling resistance and aerodynamic drag consumption.
Table 8
Comparison of the fuel consumption measured with on-road fuel consumption.
On road Vehicle weight (kg) Average speed Model 1 Model 2 Model 3 Model 1 Model 5 Model 6
30.3 15,000 38.8 37.73 32.26 51.12 34.93 19.18 24.10
(25%) (6%) (69%) (15%) (37%) (21%)
43.6 50,000 64.2 76.58 65.75 85.61 61.73 33.85 41.17
(76%) (51%) (96%) (42%) (22%) (6%)
53.0 60,000 53.7 61.42 73.75 96.27 70.79 36.44 44.21
(16%) (39%) (82%) (34%) (31%) (17%)
3.2. Discussion
Measurements of on-road fuel consumption of vehicles are typically carried out using such methods as engine and chassis
dynamometer tests, tunnel studies, remote censoring and on-board instrumentation readings. Here we compare our results
analysis from the simulations entailing Models 1–6 with measurements carried out by Erlandsson et al. (2008) who
conducted on-road measurements of 15, 50 and 60 tonnes, heavy-goods vehicles. In these experiments, the average speeds
were set to 38.8, 64.2 and 53.7 km/h for the three classes of vehicles. The tests were executed on a highway segment of
100 km, with the other parameters those used in Scenario 1. Table 8 shows the results obtained using Models 1–6 in absolute
terms (L) as well as the percentage difference from on-road fuel consumption measurements.
As seen there are large discrepancies between the results yielded by the models and those of the on-road measures. Mod-
el 4 seems to provide the best estimation for a vehicle with weight of around 15,000 kg. However, for heavier vehicles Model
6 yields better estimations. Models 5 and 6 underestimate emissions for this particular data set in all cases, whereas the
remaining models overestimate them; Model 3’s results, for example, quite far off to the on-road measurements.
4. Conclusions
The study has compared a number of models that have been developed to look at the fuel consumption and greenhouse
gas emissions associated with road freight transportation. The models produce somewhat different results in simulations
using broadly realistic assumptions, but overall are consistent with expectations; e.g. fuel consumption varies with size of
vehicle, the gradient of the road track, and speed. When comparing the modeled results with comparable data from actual,
road use data, the models vary in their performance.
Acknowledgements
The authors thank two anonymous referees for their constructive comments. This work was partially supported by a
Pump-Priming grant from the School of Management at the University of Southampton and by the Canadian Natural Sciences
and Engineering Research Council under grant 39682-10. This support is gratefully acknowledged.
References
Akçelik, R., 1982. Progress in Fuel Consumption Modelling for Urban Traffic Management. Australian Road Research Board Report 124, Canberra.
Ardekani, S., Hauer, E., Jamei, B., 1996. Traffic impact models. In: Traffic Flow Theory. US Federal Highway Administration, Washington, DC, pp. 1–7.
Barth, M., Boriboonsomsin, K., 2008. Real-world CO2 impacts of traffic congestion. Transportation Research Record 2058, 63–71.
Barth, M., An, F., Younglove, T., Scora, G., Levine, C., Ross, M., Wenzel, T., 2000. Comprehensive Modal Emission Model (CMEM), Version 2.0 Users Guide.
Technical Report. <http://www.cert.ucr.edu/cmem/docs/CMEM_User_Guide_v3.01d.pdf> (accessed on 20.01.11).
Barth, M., Younglove, T., Scora, G., 2005. Development of a Heavy-Duty Diesel Modal Emissions and Fuel Consumption Model. Technical Report. UC
Berkeley: California Partners for Advanced Transit and Highways, San Francisco.
Bauer, J., Bektasß, T., Crainic, T.G., 2010. Minimizing greenhouse gas emissions in intermodal freight transport: an application to rail service design. Journal of
the Operational Research Society 61, 530–542.
Bowyer, D.P., Biggs, D.C., Akçelik, R., 1985. Guide to fuel consumption analysis for urban traffic management. Australian Road Research Board Transport
Research 32.
E. Demir et al. / Transportation Research Part D 16 (2011) 347–357 357
Erlandsson, L., Almen, J., Johansson, H., 2008. Measurement of emissions from heavy duty vehicles meeting Euro IV/V emission levels by using onboard
measurement in real life operation. In: 16th International Symposium Transport and Air Pollution, Graz.
Esteves-Booth, A., Muneer, T., Kubie, J., Kirby, H., 2002. A review of vehicular emission models and driving cycles. Proceedings of the Institution of
Mechanical Engineers, Part C 216, 777–797.
Fagerholt, K., Laporte, G., Norstad, I., 2010. Reducing fuel emissions by optimizing speed on shipping routes. Journal of the Operational Research Society 61,
523–529.
Genta, G., 1997. Motor Vehicle Dynamics: Modelling and Simulation. World Scientific Publishing, Singapore.
Hickman, J., Hassel, D., Joumard, R., Samaras, Z., Sorenson, S., 1999. MEET-Methodology for Calculating Transport Emissions and Energy Consumption.
European Commission, DG VII. Technical Report. <http://www.transport-research.info/Upload/Documents/200310/meet.pdf> (accessed on 20.01.11).
INFRAS, 1995. Workbook on Emission Factors for Road Transport: Explanatory Notes. Technical Report. Bern.
Jabali, O., Van Woensel, T., de Kok, A.G., 2009. Analysis of Travel Times and CO2 Emissions in Time-dependent Vehicle Routing. Technical Report. Eindhoven
University of Technology, Eindhoven.
Kent, J.H., Post, K., Tomlin, J.A., 1982. Fuel consumption and emission modelling in traffic links. In: Proceedings of SAE-A/ARRB 2nd Conference on Traffic
Energy and Emissions, Melbourne.
Ntziachristos, L., Samaras, Z., 2000. COPERT III Computer Programme to Calculate Emissions from Road Transport: Methodology and Emission Factors
(Version 2.1). Technical Report. European Environment Agency, Copenhagen, Denmark. <http://lat.eng.auth.gr/copert/files/tech50.pdf> (accessed on
20.01.11).
Ross, M., 1994. Automobile fuel consumption and emissions: effects of vehicle and driving characteristics. Annual Review of Energy and the Environment
19, 75–112.