Evaluation of the influence of toll systems on energy consumption and CO₂ emissions: A case study of a Spanish highway

2.1 Mechanical model for calculating the energy consumption

Energy consumption rates from “top-down” models are normally based on transportation demand and depend on several factors at a national scale: occupation rate, speed and journey length. These rates measure the current energy efficiencies given actual occupation rates at a national level. Comparing energy consumption between modes based on these rates may lead to serious errors when the circumstances are not comparable, and makes it difficult to obtain valid conclusions at a local scale. The use of energy consumption “bottom-up” mechanical model improves the quality of the assessments, and enables evaluation of the energy consumption impacts of new policy measures regarding infrastructure and modal shift, and permits the differential reasons to be identified. The mechanical model is important for enabling a better understanding of – and thus an improvement in – energy consumption. Energy consumption calculations derived with this mechanical model can be used as reference levels for adjusting public subsidies in order to encourage energy efficiency. Tests are required to ensure a better calibration of the model. The mechanical model used in this paper to estimate the energy consumption of a vehicle type i (passenger cars, vans, coaches, articulated trucks) and motor technology j (mostly gasoline and diesel) can be expressed, in mega-joules per vehicle-kilometer (MJ/veh-km), as follows:

In the equation above, the energy consumption U_i,j consists of five groups of external forces:

2.2 Assumptions, vehicle configurations and toll scenarios

In order to apply Eq. (1) properly, energy consumption parameters were assumed, as shown in Table 1. In this paper, no cornering forces or gravitational losses are considered based on the assumption that most toll stations are located in straight sections of highways without slope. No wind was taken into consideration based on the hypothesis that the dominant wind exposures, northeast and northwest, are coincident with the axis of the AP-41 highway and the balance between the positive and negative wind exposures is null (Fig. 1).

Using the equation of energy consumption, we can estimate the amount of energy units consumed per vehicle kilometer travelled (MJ/veh-km) by different vehicle categories, fuel sources and toll scenarios. According to different studies, the engine efficiency is 0.27 for gasoline engines and 0.4 for Otto diesel ones (Table 1). In this paper, vehicles are generally classified into four categories: passenger cars, vans, tourist coaches and articulated trucks. Each vehicle category has individual average vehicle mass (m, kg) and frontal area (A_f, m²). The configurations of researched vehicles are shown in Table 1.

The aim of this paper is to compare the influence of toll systems applied on highways, on vehicle energy consumption and CO₂ emissions. To reach this objective, different case scenarios related to different toll systems are set. For all the scenarios, the area of modeling includes a 316–347 m road section and the tolling station is located roughly in the middle. Table 2 shows the assumptions for each scenario. In case B, a 3-min-pause at the toll was assumed, for people to pay fees and communicate with the toll staff. As in case C, because of the adopting of Electronic toll collection (ETC) technology, commuters just have to slow down their cars to a certain speed (8.3 m/s) for tolling instead of completely stop the vehicle. The section total length differs between scenarios because of different deceleration and acceleration distances. These distances are bigger in case scenario B than in scenario C; in the later case, the vehicle does not stop totally at the toll station. We can consider the distance of case scenario A equal to the distance of scenario B. However and according to model equation (1), the length of road section does not influence the results on energy consumption rates expressed in mega-joules per vehicle kilometer. Effective distances travelled are both in the numerator and denominator of model equation (1). In scenarios B and C, acceleration and deceleration energy consumption, before and after passing by the toll station, are taken into account. In all scenarios we consider no traffic congestion.

2.3 CO₂ emissions and traffic measurements

The CO₂ emissions are estimated from the energy consumption through the carbon emission factor (CEF). According to ADEME (2007), the grams of CO₂ emitted per mega-joule of energy (g CO₂/MJ) are 81 and 86 for diesel and gasoline fuels, respectively. Therefore, the carbon emission of a vehicle type i (passenger cars, vans, coaches, articulated trucks) and motor technology j (gasoline and diesel) can be expressed, in grams of CO₂ equivalent per vehicle-kilometer (g CO₂/veh-km), as follows:

In this paper, we applied the energy and CO₂ emissions model to a road section on the highway AP-41 from Toledo to Madrid (Spain) including a toll station (located at kilometer point 14.3). The section length (L) depends on the case scenario considered (Table 2). The length is 0.347 and 0.316 km, for case scenarios B and C, respectively, including the toll station roughly in the middle and connected highways on both ways. The monthly average daily traffic (MADT) of the four categories of vehicles considered at this point is given by the Spanish Road Traffic Survey (SRTS) from the Spanish Ministry of Public Works (2009a) (Table 3). The SRTS provides a comprehensive measure of the distance travelled by vehicles.

By multiplying the MADT by the CO₂ emissions per vehicle-km and the section length, the daily CO₂ emissions on highway AP-41 can be obtained at this section. Consequently, based on CO₂ emissions, the comparison among different scenarios is made. The daily carbon emissions of a vehicle type i and motor technology j were computed using the following expression:

3.1 Scenario analysis

The results of CO₂ emissions per vehicle kilometer by vehicle and fuel categories for all three scenarios are shown in Fig. 2. Note that in case B, a 3-min-pause at the toll is included, during which the vehicles continue to burn fuel and emit CO₂. For diesel and gasoline engines, the emission rate of CO₂ is 0.05 and 0.06 kg/min (Fuzzi et al., 2006). As a result, the pause step in case B contributes with 0.15 kg (0.17 kg) to the total CO₂ emissions.

Close

Figure 2

CO₂ emissions per veh-km by vehicle type and engine category and case scenarios: free flow, traditional toll and tele-toll. Note: ^aCase B consists of three steps of vehicle movement: deceleration, stop and acceleration. Data shown in the graph are the summary of fuel consumptions of all three steps. ^bCase C consists of two steps of vehicle movement: deceleration and acceleration. Data shown in the graph are the summary of fuel consumptions of both two steps.

Export to PPT

For each scenario, the comparison among different vehicles is illustrated in Fig. 2. It is evident that the vehicle mass affects CO₂ emissions significantly, since heavy vehicles like articulated trucks emit 8.4 times as that of passenger cars. Besides, there are major differences between case A and case B (case B counts up to 20 times of case A), whereas case B and case C are similar. So far, there are no big advantages of using ETC toll system instead of traditional one in terms of energy consumption and CO₂ emissions. In addition, the type of engine fuel (gasoline or diesel) can also influence CO₂ emissions significantly, especially in articulated trucks and buses. For this reason in Spain there are only diesel trucks and buses.

Fig. 3 shows the emission share of different movement steps, namely the deceleration, stop and acceleration steps in case scenario B, and deceleration and acceleration steps in case scenario C. The main emission shares go together with the acceleration step (note the value of the labels of Fig. 3 around 90%). Also in case B, the share of the stop step decreases with the vehicle mass because it seems that heavy vehicles consume much more energy with the deceleration and acceleration steps that light vehicles, while the consumption during stop step is relatively constant.

Close

Figure 3

Emission percentage of each movement steps in case B and case C.

Export to PPT

3.2 Total emission on road section of highway AP-41

By multiplying the monthly average daily traffic of vehicles (MADT) by the corresponding CO₂ emission rates (Fig. 2) and the length of the section (constant equal to 0.347 km), the CO₂ emissions on the road section of highway AP-41 are calculated. In the estimation of the emissions, we consider that buses and trucks use only diesel fuel. Diesel vans and cars represent, respectively, 68% and 64% of the total light vehicle traffic in Spain, according with the interurban national traffic measurements and records of fuel purchases (Perez-Martinez and Monzón, 2010). Consequently, the MADT was weighted up according to these shares and applied the respective fuel emission rates. Fig. 4 shows the amount of emissions released per day for each case scenario and vehicle category. The figure shows year-through significant differences of CO₂ emissions among the three case scenarios and vehicle categories during the period January (2008) to March (2009). It is evident that either case B or case C produces much more CO₂ emissions than case A in any type of vehicles, but the variation between case B and case C is much smaller, especially regarding heavy vehicles. The technology of ETC does not seem to reduce the CO₂ emissions significantly, but more to shorten the time for drivers to pass the toll area. Free flow is beneficial for reducing CO₂ emissions, but ETC seems of no help in this aspect.

Close

Figure 4

CO₂ emissions per day on road section of AP-41: comparison among case scenarios and vehicle categories, January 2008–March 2009.

Export to PPT

Across the year, passenger cars have a bigger responsibility for emitting CO₂, in spite of their low CO₂ emission rate per kilometer. All vehicles, except buses, present lower daily emissions during summer time; whereas buses have the highest CO₂ emissions during summer time. Throughout the analyzed period, the total emissions of the vehicles present a downward trend due to the negative growth of the daily traffic between Toledo and Madrid. From May to July, more CO₂ is produced due to the peak season of tourism. CO₂ emissions in 2008 were estimated to be 80.1, 1087.7 and 1049.3 tones under cases A, B and C, respectively. Throughout the year, cases B and C emit more than 13 times of the emission under free flow. Based on the data in case B, the difference between A and B is 93% while the difference between B and C only accounts to 4%, which further weakened the importance of ETC technologies in highways.

4.1 Road slope, wind exposure factor and parameters depending on service type

This paper shows that energy consumption is most sensitive to slope (Fig. 5). Increasing road slope by 10%, the energy demand increased significantly from 21.56 to 29.18 MJ/veh-km (35%, Fig. 5), as vehicles consume more energy on unfavorable slopes during the acceleration phase under the case study B. While it is critical to vehicle fuel consumption and CO₂ emissions, the slope of road seems unlikely to change. Topography, construction and economy have to be considered in building a highway, which sometimes make the chosen of steep slope inevitable. However, tolls are located in segments of highways with small slopes and curves.

Close

Figure 5

Sensitivity analyses: energy consumption and CO₂ emissions of gasoline cars upon parameter changes and under case B scenario. Note: Case B consists of three steps of vehicle movement: deceleration, stop and acceleration.

Export to PPT

Wind exposure factor (e_v) is the second most influential parameter concerning vehicle energy consumption (Fig. 5). A 10% reduction in e_v can decrease total energy demand by 10%. While in the scenarios of this paper the e_v is set to 1.0 to ignore its effect, it is influential either positively or negatively to the energy consumption in many practical situations. If the wind is from the backside of vehicles, the e_v parameter is below 1.0 to consider wind push during cruising. In contrast, vehicles suffer from wind resistance when driving toward the wind direction. Fig. 1 shows the annual wind distribution of Madrid and map of highway AP-41. From the figure, it is obvious that either driving from Madrid to Toledo or reversely, wind effect is considerable, since the annual wind direction is N/NNE–SW/SSW, same as the general direction of highway AP-41.

Mass correction factor for rotational inertia acceleration (C_i) and acceleration rate (a) contribute to the inertial resistance (U_i) when vehicle is accelerating. In the scenarios discussed in this paper, namely case B and case C, vehicles are assumed to decelerate (to 0 or to a low speed) and then accelerate. Therefore, the fuel consumed by rotational inertia acceleration is remarkable. A 10% reduction in C_i and a can decrease total energy demand by 9.5%.

The average vehicle speed (vv) is not as detrimental factor as e_v and a even though to the square effect on the aerodynamic drag and the kinetic energy losses. The sensitivity analyses in this paper show that a 10% reduction in the average vehicle speed can decrease total energy demand by only 0.3%. In this case, kinetic energy losses are more dependent on parameters such as, vehicle acceleration a, engine size and vehicle mass (m). Other factors involved in this case study include traffic conditions at the toll station, the number of stops during the phase 2, driver behavior, braking coefficient and highway profile (i.e. presence or absence of slope at the station). This range of factors, some of them depending directly on service type, makes it difficult to model fuel efficiency on a per-vehicle basis due to the great amount of variables involved in the calculations. In this study we consider acceleration–deceleration rates according to the gear used and for all vehicle types (Table 2). In general, reducing the number of stops can decrease inertial acceleration losses. In this regard, case study B (traditional toll) makes more stops than alternative toll systems (free flow and ETC), and consume more energy due to inertial losses.

4.2 Parameters not depending on service type

Vehicle mass (m) is one of the most significant parameters affecting the energy demand of a vehicle, since energy demand is directly proportional to rolling resistance (U_r), inertial acceleration resistance (U_i) and gravitational losses (U_g), and indirectly proportional to cornering resistance (U_c) according to Eq. (1). Therefore, reducing the vehicle real mass is one of the most effective ways of decreasing the energy demand of passenger land transport modes. By reducing mass by 10%, using lighter construction components, the vehicle energy consumption would decrease by 9.8%, as shown in Fig. 5. Despite the importance of mass on energy consumption, cars have actually been increasing in mass in the last few years due to several factors such as improved safety, pollution control devices and additional auxiliary services (i.e. air conditioning, electronic windows and mirrors, etc.) (Van den Brink and Van Wee, 2001; Van Wee et al., 2005). In the future, the ways of producing new materials with high strength to mass ratio can be useful for reducing vehicle weight.

Apart from the deceleration and acceleration processes, other parameters show sensitivity to energy consumption in a constant speed cruising. Losses due to rolling resistance are proportional to the friction coefficient of the tires (C_r). A 10% reduction in rolling resistance coefficient could reduce energy demand by 0.4%. Similarly, the main factors affecting aerodynamic losses are the aerodynamic drag resistance (C_d) and the vehicle frontal area (A_f). A 10% reduction in drag resistance coefficient and frontal area could reduce energy demand by 0.2%. The frontal area of the vehicle is bigger in collective transport modes than in private cars and measures the space available inside the car (Table 1). The frontal area has increased in recent years, especially in private cars, due to consumer demand for greater comfort (Van den Brink and Van Wee, 2001; Advenier et al., 2002). Conversely, the aerodynamic drag resistance coefficient has decreased in all systems due to better design parameters, and this has a major influence on energy consumption, particularly in windy conditions.

4.3 Engine efficiency

In Fig. 5, engine efficiency (η_motor) is the only parameter inverse-proportional to the energy consumption. This is because the improvement of engine efficiency would actually cut down energy wasted, thus decreases the total energy demand. According to heat engine theory, modern gasoline engine has an average efficiency of about 30%, because most of the energy produced is consumed by friction, mechanical sound and turbulence. In the scenarios set in this paper, the η_motor is even lower; since at slow speed the efficiency is lower than average, due to a larger percentage of the available heat being absorbed by the metal parts of the engine, instead of being used to perform useful work. In this sensitivity analyses, if the η_motor is improved by 10%, the energy consumption could be reduced by 11%. However, the raise of engine efficiency has been a problem to engineers for years, which is depending on varied factors such as compression ratio, oxygen, fuel quality and manufacturing level. Future improvement could be aiming on new engine technologies and better fuels (Kaul and Edinger, 2004; Niedzballa and Schmitt, 2001).

The final result of Fig. 5 is a combination of total “tank to wheel” energy and associated CO₂ emissions. We must add to these values the energy losses and emissions that occur between the primary sources and the vehicle (“well to tank”). In the case of gasoline oil in Spain these losses are significant and are around 19%. The calculation of energy losses and emissions before reaching the vehicle for different fuel used in transport and in the local circumstances of Spain can be seen in López Martínez et al. (2008).

4.4 Uncertainty of estimates

The traffic measurements and the energy consumption and emission rates calculated using model from Eq. (4) are subject to uncertainties mainly because of traffic measuring errors, speed distribution, fleet composition and road gradient. Kühlwein and Friedrich (2005) performed a detailed error analysis to quantify the margins of error of the traffic-related input parameters on the total energy consumption and CO₂ emissions in a German highway. These authors estimated that total traffic flow data are subject to a variation coefficient (VC) of less than 1% for automatic counting results and their effects on the energy consumption and CO₂ emission rates were quantified to be less than 1.2%. Similarly, the effects of speed distribution, fleet composition and road gradient were quantified to range from 0.2% to 5%. Total error of the energy consumption and CO₂ emission rates can be estimated from the quantified uncertainties of the traffic-related model input parameters from the sum of effects of the individual single errors on the total emission. In our case study, automatic counting of traffic flows on a highway segment with no slope, the mean total error could be in the range of 7.5–10%, mainly due to fleet composition and simplification of model to only four types of vehicles (cars, vans, buses and trucks).