Models of energy and emissions for the materials lifecycle of several vehicle technologies

Transcription

1 Models of energy and emissions for the materials lifecycle of several vehicle technologies Gonçalo Oliveira Nero Correia Instituto Superior Técnico, Technical University of Lisbon, Av. Rovisco Pais, 1 Pavilhão de Minas, 4º andar, Lisboa Portugal Abstract A set of simplified estimation models for energy consumption and emissions of Greenhouse Gases (GHG) and pollutants (NO x, VOC, CO, SO x, PM 10 and PM 2.5) was developed, related to the vehicle materials lifecycle. Five vehicle technologies were analysed: internal combustion engine vehicle (), hybrid electric vehicle (HEV), plug-in hybrid electric vehicle (PHEV), battery electric vehicle (BEV) and fuel cell hybrid electric vehicle (FCHEV). The scope is to obtain models that are simple and fast to use, saving user time and supporting the development of a vehicle environmental ranking, related to its materials lifecycle. A weighted pollutant emissions category was developed, based on the pollutants damage costs, for a easier technology comparison. The energy and emission models receive as inputs variables of public access as the weight of the vehicle and of its batteries, electricity production mix, recycled materials content and the fuel cell power. Data from the GREET 2012 database was subjected to multiple linear regressions and a selection of the relevant variables to include in the models was performed, using a statistical hypothesis test. Forty five energy and emission estimation models were obtained for the vehicle materials lifecycle: one for each technology and impact category. After the model validation was concluded that the models have a good performance, achieving average errors lower than 12% for the VOC, NO x and SO x emissions and lower than 34% for the other categories. Keywords Vehicle materials lifecycle, energy and emissions, electric vehicle, fuel cell vehicle, hybrid vehicle, multiple linear regression. Introduction The issue of the energy efficiency, applied to several industry and domestic sectors, is currently seen as a priority in the development of a society desired to be increasingly more sustainable. More specifically, the automotive sector has developed energy efficiency measures that have allowed reductions on fuel consumption per kilometer. According to a study published in 2006, the specific consumption per kilometer of cars sold in the European Union fell by 36% between 1975 and 2002 (Zachariadis, 2006). These measures were significantly boosted by the European directive which seeks to achieve better energy performances in this sector: the goal of maximum fuel consumption for gasoline passenger cars is about 5.6 L/100km and 4.1 L/100km in 2015 and 2020, respectively. As for the diesel passenger car, the goals are 4.9L/100km and 3.6 L/100km for the two time horizons (European Commission, 2012a). In addition to energy efficiency, the issue of air pollution and its impacts caused either referring to Greenhouse Gases (GHG) or other pollutants, is often a central issue in the political agenda of many countries and discussed in numerous global conferences. As such, the European Union has set targets to reduce emissions of carbon dioxide (CO 2), limiting the release of this gas at 130 g/km for vehicles circulating in 2015, and 95 g/km for those travelling in 2020 (European Parliament and Council, 2009). Pollutants are released in a daily basis by the transportation sector and due to the effects they have on public health and on the environment, their emissions are being increasingly curtailed. The European Union produces one of the most restrictive laws in the world in relation to pollutants emission, known as the Euro standards. 1

2 These standards were primarily implemented in 1992 and have undergone constant updates, either by the introduction of pollutants that were not previously considered, whether by imposing more ambitious goals. The standard currently in force (Euro 5), that applies to the categories of passenger cars and light commercial vehicles (up to 2610 kg) circulating in Europe (European Commission, 2012b), differentiates the emission limits by category and fuel. It includes exhaust emissions, evaporative emissions and from the motor s crankcase. For the category of light commercial vehicles there s also a differentiation by the vehicle mass, allowing higher emissions for heavier cars. The regulated pollutants are carbon monoxide (CO), hydrocarbons (HC), nitrogen oxides (NO x) and particulate matter (PM) 1 (DieselNet, 2012). Thus, as an example, currently the gasoline passenger car must have emissions below 1000 mg CO/km, 100 mg THC 2 /km, 60 mg NO x/km and 5 mg PM/km. The Euro 6 standard, expected to enter into force in September 2014, contains even more significant emission restrictions, especially for NO x emissions from diesel vehicles, with reductions of more than 50%. It is also expected to introduce a category that limits the number of particles per kilometer (European Parliament and Council, 2007). Consequently, the existing measures to reduce GHG emissions and pollutants as well as to reduce energy consumption currently focus in the car use phase, i.e. during the time period in which the vehicle is actually being used. Thus, the other lifecycle stages of the automobile remain unregulated, particularly the materials lifecycle. This study is expected to give a contribution to the vehicle rating scores and consumer information about the vehicles performance, concerning the lifecycle of the embodied materials, facilitating comparison of different propulsion technologies. To this end, were developed simplified models based on multiple linear regression, applied to the GREET 2012 database for the technologies of, HEV, PHEV, BEV and FCHEV. Literature review Based on the reviewed studies about the vehicle materials lifecycle an attempt was made to draw conclusions about the characteristics of the vehicles studied, the lifetime and the energy mix, and its relationship with emissions and energy consumption. The Table 1 summarizes the most relevant information about each study, including the main features of the vehicles. Has it can be seen, the reviewed studies have very specific methodologies by using for example different vehicle lifetimes or by covering different Lifecycle Analysis (LCA) phases. In Table 2 are presented the results obtained for each study, in terms of energy consumption and emissions for several technologies. In relation to the studies covering the, the results of the study (Hu, Pu, Fang, & Wang, 2004) are substantially higher than other studies results, since the entire lifecycle is accounted, including the fuel and the use cycles. The study (Liu, Wang, & Yang, 2007) in turn, aims to compare transportation modes depending on the technologies used and therefore the results are reported in g/passenger.km, not allowing comparison with other studies. Another study that compares different car technologies (Granovskii, Dincer, & Rosen, 2006), contains only results for emissions and groups them into two categories: Air Pollutants (AP) and Greenhouse Gases (GHG). The first category is the sum of NO x and CO, weighted with coefficients that took into account marginal costs (of public health), related to these pollutants emissions. The second category is related to GHG, and was weighted with the coefficients of global warming potential (Houghton, LG Meira Filho, BA Callander, & Maskell, 1996). 1 The particulate matter is considered, independently from its diameter. 2 Total Hidrocarbons, including the Non Methane Hidrocarbons (NMHC). 2

3 Table 1: Summary-table for the reviewed studies (P Production, M Maintenance, EOL End of Life, T Transportation (of vehicles), U Utilization, TLC Total Lifecycle; ICE Internal Combustion Engine, EE Electrical Engine) Electricity Vehicle Vehicle Engine Propulsion Fuel Cell LCA Study Technology Software Mix Lifetime (km) Weight (kg) Power (kw) Battery Power (kw) phases P, M, EOL CMLCA (Leduc et al., 2010) n.a Diesel P, M, EOL CMLCA (Hu et al., 2004) China (Wu C, Ma X, Chen Y, Zhao Z, 2006) (Liu, R. Wang, and Yang 2007) China China (Delucchi, 2001) USA (Boureima et al., 2009) (Granovskii, Dincer, and Rosen, 2006) Belgium n.a (Baudoin, 2007) EU (P. Baptista, Tomás, & Silva, 2010) (P. C. Baptista, Silva, Farias, & Heywood, 2012) TLC GREET P, EOL GREET P, U P, T LEM P, T, EOL CLEVER Diesel P, T, EOL CLEVER HEV /60 27 kw, 50 kg (ICE/EE) (NiMH) - P, T, EOL CLEVER BEV kw, 300 kg (Li-ion) - P, T, EOL CLEVER P, EOL - HEV /50 27 kw, 53 kg (ICE/EE) (NiMH) - P, EOL - BEV kwh, 430 kg (NiMH) - P, EOL - FCHEV P, EOL P, EOL GREET Diesel P, EOL GREET HEV kg (NiMH) - P, EOL GREET FCHEV kg (Li-ion) 70 P, EOL GREET BEV kg (NiMH) - P, EOL GREET EU PHEV kg (NiMH) - P, EOL GREET Portugal PHEV kg (NiMH) - P,EOL GREET 3 n.a. not available 3

4 Table 2: Energy consumption (kj/km) and Emissions (g/km) for each study Study Technology Energy CO 2 CO NO x SO x COV HC PM AP Observations (Leduc, Mongelli, Uihlein, & Nemry, 2010) Diesel CO 2eq, Primary energy consumption, SO 2eq, PM 2.5 (Hu, Pu, Fang, & Wang, 2004) Total Lifecycle (Wu C, Ma X, Chen Y, Zhao Z, 2006) SO 2, CO 2eq (Liu, Wang, & Yang, 2007) * 30.6* 1.27* * g/(passanger.km); includes utilization phase (Delucchi, 2001) CO 2eq, NO 2+N 2O, SO (Boureima, Vincent, Nele, Heijke, & Messagie, 2009) (Baudoin, 2007) (Granovskii, Dincer, & Rosen, 2006) Diesel HEV BEV Diesel HEV Parallel-Series HEV Parallel HEV Series BEV FCHEV HEV BEV FCHEV CO 2eq, SO 2eq PM 10 CO 2eq, AP = x CO + NO x (P. C. Baptista et al., 2012) PHEV (P. Baptista et al., 2010) PHEV

5 The results shown for each study are very different, however it can be seen that for CO 2 emissions the studies (Wu C, Ma X, Chen Y, Zhao Z, 2006), (Boureima, Vincent, Him, Heijke, & Messagie, 2009), (Baudoin, 2007) and (Granovskii, Dincer, & Rosen, 2006) present similar values, despite the differences between the mixes of electricity and the masses of the vehicle used. More information is needed in order to understand how these factors affect emissions and energy consumption. As can be seen, the results for Diesel are very similar to those obtained for, especially for the same study, which was expected since these technologies have few differences. For the (Baudoin, 2007) study, having just HEV Parallel-Series technology as a reference, the results of energy consumption and emissions have increased compared to the, namely for CO 2 that have risen 26%. Also for the study (Boureima, Messagie, et al. 2009), the power consumption obtained for hybrid technology is more than 50% higher than the conventional technology () as well as for CO 2 emissions. The HEV's results for the Granovskii analysis indicate a 16% increase in GHG emissions and 17% for Air Pollutants category, compared with 's. The differences observed between the hybrid and conventional technologies can be explained by the use of components in the vehicle that require higher energy consumption for their manufacture and emit more pollutants (Baudoin, 2007), such as the batteries included in HEVs. The studies found for PHEV show results of energy consumption and pollutant emissions highly above those obtained for previous technologies. Note that the vehicle lifetime considered in these analyses is very important on the evaluation of materials lifecycle results since they are a function of the first. Thus, for small lifetimes, emissions and energy consumption will be higher, even for the same technology. For these studies, the lifetime (150,000 km) is lower than that found for other technologies (over 200,000 km), so this may be an explanation for the high values obtained. Regarding the BEV studies, all the results show an increase comparing to s, being the biggest related to the Sulphur Oxides (SO x) emissions (+596%) for the (Baudoin, 2007) study. This variation, despite being very high in relative terms only represents a difference between g/km () and g /km (BEV), which in absolute terms, represents a 0.31 g/km increase. Moreover, significant increases have been shown for the energy consumption (more than 100% for the 2009 and 2007 studies) and for CO 2 emissions (+117%) (Baudoin, 2007). The battery mass, even more significant in the case of the electric vehicle, appears to be determinant on emissions and energy consumption. With respect to FCHEV studies, the 2007 study shows that this technology consumes approximately 48% more energy, emits more 55% CO 2 and 132% SO x than the. For the 2006 study, there was a dramatic increase in both categories analyzed, comparing with conventional technology. CO 2 emissions increased to g/km and combined emissions to g/km. This significant increase is most likely due to the fuel cell included in this technology, which contains components, like the electrodes, that require high energy consumptions during manufacturing (Granovskii, Dincer, & Rosen, 2006). 5

6 Methodology 1.1. Literature review From the technologies analysed (Table 3), it was not possible to identify clear trends between the vehicle characteristics studied and the results, given the different methodologies used and the lack of studies about the latest technologies, so a case study of the five different vehicles was performed, one by each technology. Technology Diesel Table 3: Technologies analysed from the literature review Study (Leduc et al., 2010) (Hu et al., 2004) (Wu C, Ma X, Chen Y, Zhao Z, 2006) (Liu et al., 2007) (Leduc, Mongelli, Uihlein, & Nemry, 2010) (Baudoin, 2007) (Baudoin, 2007) HEV (Granovskii, Dincer, & Rosen, 2006) (P. Baptista, Tomás, & Silva, 2010) PHEV (P. C. Baptista, Silva, Farias, & Heywood, 2012) (Baudoin, 2007) BEV (Granovskii, Dincer, & Rosen, 2006) (Delucchi, 2001) (Boureima, Vincent, et al., 2009) (Baudoin, 2007) (Granovskii et al., 2006) (Boureima, Vincent, Nele, Heijke, & Messagie, 2009) (Boureima, Messagie, et al., 2009) (Boureima, Messagie, et al., 2009) FCHEV (Baudoin, 2007) (Granovskii, Dincer, & Rosen, 2006) 1.2. Case studies Assumptions The five vehicles simulated on the GREET 2012 program were: Volkswagen Golf 1.2 TSI (), Toyota Prius 1.8 (HEV), Toyota Prius Plug-in 1.8 (PHEV), Nissan Leaf (BEV) and Honda FCX Clarity (FCHEV). Thus, the automobile materials lifecycle was subdivided into four distinct parts, which were separately analyzed: the vehicle components, such as the chassis and the propulsion system; the small and the propulsion batteries; the vehicle fluids, such as the motor oil; and the processes of Assembly, Disposal and Recycling (ADR), such as the car s painting. In this analysis were considered a lifetime of the vehicle of km, the European electricity production mix for the year 2009 (OECD / IEA, 2011) and pre-defined values for the fractions of recycled materials: recycled steel (70%), recycled wrought aluminum (11%), recycled cast aluminum (73%), recycled lead (73%) and recycled nickel (44%). The masses of the vehicle and of the batteries are shown in Table 4. Thus, each vehicle is comprised of two batteries, except the containing only the small battery of Lead-Acid (Pb-Ac). While this provides energy at the time of ignition and for the entertainment systems such as radio, the propulsion battery stores the energy needed to power the electric motor present in other technologies. As can be seen, the battery mass increases with the need for storage, which is generally greater in the case of the BEV battery. Also the battery type used depends on the technology, considering the nickel metal hydride (NiMH) battery for HEV and the battery Lithium-Ion (Li-ion), in the other cases. Table 4: Vehicle and batteries mass by technology, kg HEV PHEV BEV FCHEV Vehicle mass Small battery mass Propulsion battery mass The is assumed to have a four cylinder explosion engine with 78 kw of power, the HEV and the PHEV has a four cylinder explosion engine with 73 kw of power and an electric motor of 60 kw, the BEV has an 80 kw electric motor and the FCHEV has an 100 kw electric motor and a 100 kw fuel cell. 6

7 Referring to the vehicle materials, steel is the one used in larger amount, regardless of the propulsion technology (over 55 wt %), showing no significant variations. Cast iron is about 10% of the total mass of the components and is progressively less incorporated into other technologies, achieving only 2% of the total weight of the components of the fuel cell vehicle. The wrought aluminum is embedded in about 2% of the composition of all technologies, except for FCHEV where reaches 7% of the total mass of components. The cast aluminum constitutes about 5% of the total weight of the automobiles, HEV and PHEV, and about 6% and 3% of the BEV and FCHEV, respectively. The carbon fiber-reinforced plastic is not used in any technology, except in FCHEV where is incorporated in large quantities in the fuel cell, etc. The remaining materials represent approximately 20% of the total mass of components and include, for example, plastic, glass, copper and brass. In addition to the outputs provided by GREET 2012, a variable of combined emissions (AP) was introduced, which aggregates all the pollutants emitted, by using a weighting method based on the damages costs (de Bruyn et al., 2010a, 2010b). Thus, each pollutant obtained a weighting factor, with reference to the nitrogen oxides: NO x (1), CO (0.003), VOC (0.240), SO x (1.038), PM 10 (6.113). Unlike other emission parameters, this only considers urban emissions, where most of the world's population is found (United Nations, 2012) and where the greatest damage to public health occurs Energy and emission results The energy consumption for the materials cycle increases considerably with the technological evolution of vehicles. In fact, while the consumes about 406 kj per kilometer, the FCHEV consumes more than 670 kj/km, which corresponds to a 66% increase. This variation can be explained by the high energy requirements of manufacturing the fuel cell, which uses carbon fiber-reinforced plastic and wrought aluminum in large quantities. The increase in energy consumption for the battery electric and hybrid technology is mainly due to NiMH and Li-ion batteries, which require more energy to produce than lead batteries (Baudoin, 2007). There is also an increase in greenhouse gas emissions compared with the conventional technology (), which is about 14% to HEV, 19% to PHEV, 25% to BEV and 70% to FCHEV, reaching 44 g/km. Again, the majority of GHG emissions are due to vehicle components (Baudoin, 2007). The emission of Volatile Organic Compounds (VOC) is approximately 0.15 g/km, with a slight tendency to increase with the evolution of technology. In fact, the biggest increase compared to the is only 1% (FCHEV), and therefore the kind of technology isn t a determining factor on the variation of the pollutants emissions. In contrast to other gases, the emission of VOC is due to the fluids comprised in each car in more than 80%, such as the windshield fluid, which during its use emits the most. Since the amount of fluids used doesn t differ significantly between propulsion technologies (Burnham, 2012), it is understood that the increase of VOC emissions is negligible. The emissions of CO are circa 0.07 g/km, apart from the conventional technology () that takes the value of g/km. A clear trend on the variation of this pollutant emission was not identified, for all technologies. This pollutant is mostly emitted by the production of automobile components, in particular those containing virgin steel which emits around 120 g of CO per kg of material produced (Baudoin, 2007). Being the steel the major material included in cars structure, the absence of an expressive increase in CO emissions can be explained by the fact that the steel mass doesn t vary significantly in the vehicle, for all technologies. The NO x emission is mainly due to ADR processes, more specifically to the activities of painting and dismantling of the vehicle and also to the production of some components (Baudoin, 2007). The most significant increase in NO x emissions occur for FCHEV, which incorporates about 12% of carbon fiber-reinforced plastic and whose emissions of NO x are high (about 16 g / kg of material) (Baudoin, 2007). The emission of PM 10 particles increases from g/km for technology, to g/km (BEV) and to g/km (FCHEV). In fact, also the PM 2.5 particulate emissions increase identically, since they only differ in terms of the diameter. The production of the components is the part of the materials lifecycle that emits the most particles, particularly those with high content of virgin aluminum material which has one of the major emission factors for this pollutant (about 30g/kg of material). The progressive increase of these emissions can be explained by increasing mass of this metal in the structure of the car, following the technological development. SO x emissions are mainly due to the production of vehicle components but also of propulsion batteries, especially those of NiMH and Li-ion batteries that pollute more than the Pb-Ac batteries (Baudoin, 2007). This explains the significant increase in the emissions of SO x, as the capacity and weight of the batteries increases for HEV, PHEV and BEV. Relating to FCHEV, greater emissions are explained because of the average of 1.5 kg Nickel comprehended in its constitution, unlike other types of vehicles, and whose emissions of SO x are extremely high (about 690 g/nickel kg produced). 7

8 The combined emissions vary between approximately 0.4 g/km and 0.6 g/km for the and the FCHEV, respectively. Since this category includes all emissions described above, and taking into account that the largest weighting factor is assigned to the particles (which increased considerably), the increase in these emissions is easily justified with the evolving technology propulsion Correlation analysis After the case studies on the vehicle materials lifecycle, the energy and emissions models were developed based on a sample of simulations done using the latest version of the Greenhouse Gases, Regulated Emissions, and Energy Use in Transportation (GREET) database, which is dedicated to the LCA of the road transport sector (Argonne National Laboratory, 2012). This program consists of two modules: the GREET 1, which deals with the fuel cycle and use phase and the GREET 2, which is associated with the cycle of the vehicle materials. The fuel and use cycles deal with the processes from the production of the fuels to the use phase of the vehicle, included in the Well-to-Wheel (WTW) analysis, and the materials cycle, included in the Cradle-to-Grave (CTG) analysis includes all processes required for the production of the vehicle and its end-of-life. GREET calculates the energy consumption, GHG emissions and six pollutants emissions, more specifically of VOC, CO, NO x, SO x and of particles with diameters smaller than 10 μm and 2.5 μm, respectively PM 10 and PM 2.5. A set of simulations for the materials cycle was done, for five vehicle propulsion technologies:, with an internal combustion engine (ICE) and a small battery (Pb-Ac); HEV, comprising an ICE, an electric motor, a small battery (Pb-Ac) and a NiMH propulsion battery; PHEV, consisting of an ICE and an electric motor, a small battery (Pb-Ac) and a Li-ion propulsion battery; BEV, comprising an electric motor, a small battery (Pb-Ac) and a propulsion battery (Li-ion); FCHEV, comprising a fuel cell, an electric motor, a small battery (Pb-Ac) and a propulsion battery (Li-ion). Table 5: Analysed input/output data Input Output Mass of the vehicle, kg Mass of the propulsion battery 4, kg Lifetime of the vehicle, km Power of fuel cell 5, kw Electricity production mix 6 (%) Fractions of recycled materials 7 (%) Energy consumption, kj/km GHGs emission, g/km Pollutant emissions (CO, VOC, NO x, SO x, PM 10 and PM 2.5), g /km Combined emissions (AP), g/km For each technology, the created input and output data was stored, to subsequently develop multiple linear regression models. In Table 5 are the inputs and outputs analyzed. It is expected that by varying the inputs, also will change the contribution of processes which are directly related to them, such as the production of materials and components, and hence modifying the results of energy and emissions. The vehicle and battery masses and power of the fuel cell are good examples of that, and it s expected that these variables would be directly proportional to the outputs. On contrary, the lifetime of the vehicle is expected to be inversely proportional to the results since these are expressed as a function of the distance travelled along the lifecycle. The electricity production mix is an important element to be included, due to the variation of the emission factors and energy efficiency associated with different energy sources used. Finally, the information contained in the fractions of the recycled materials allows introducing in the model production ways with environmental advantages, because of the lower energy consumption and emissions associated, as are the manufacturing processes of recycled materials (Baudoin, 2007). Note that it is not intended to be a set of simulations of real cars, but rather a sample which covers all possible variations in the composition of the vehicle, within the universe of each parameter input (Table 6). In fact, in the case of the fully electric or fuel cell technologies, there are very few car models available, so limiting the sample to their numbers could imply obtaining statistically insignificant results. Moreover, by using a sample which is representative of simulations of characteristics plausible to take a vehicle, it is possible to obtain a model that is actually useful in the estimation of the energy consumption and emissions, related to the vehicle materials lifecycle of a particular car. Depending on the technology, were made between 900 and simulations that were subsequently analyzed. 4 In the case of the HEV, PHEV, BEV and FCHEV. 5 In FCHEV s case 6 Energy sources used: Oil (%), Natural Gas (%), Coal (%), Nuclear (%), Biomass (%), Renewables (%). 7 Fractions of recycled materials used: Steel (%), Wrought Aluminum (%), Cast Aluminum (%) Nickel (%), Lead (%) 8

9 In Table 6 are shown the simulation ranges for each variable, and that should be respected in the use of models. Table 6: Simulation ranges for each variable Variable Technology Minimum Value Maximum value Mass of the vehicle, kg All Lifetime of the vehicle, km All Electricity production mix (%) All 0% 100% Fractions of recycled materials (%) All 0% 100% HEV 20 (NiMH) 80 (NiMH) Mass of the propulsion battery, kg PHEV 80 (Li-ion) 240 (Li-ion) BEV 100 (Li-ion) 500 (Li-ion) FCHEV 13 (Li-ion) 153 (Li-ion) Power of fuel cell, kw FCHEV 70 (Li-ion) 110 (Li-ion) In these simulations was included the maintenance phase of the vehicle and the intervals of replacement of the tires, batteries and fluids are described as follows: tires are replaced every 65000km, the Pb-Ac batteries each 75000km and the Li-ion and NiMH batteries every km. Relating to the fluids, the motor oil is replaced every 15000km, the brake and the powertrain cooling fluids every 65000km, the transmission fluid every km and the windshield liquid every 12000km. The power steering fluids and the adhesives are not replaced in the vehicle lifetime (Baudoin, 2007; Burnham, 2012) Multiple Linear Regression Once stored the simulations, the model was obtained through multiple linear regression with the method of Linear Least Squares (LLS). By using this method it is assumed that the independent or self-explanatory variables have a linear relationship to each other, and consequently the model takes the following form: where Y is the dependent variable, e.g. GHG emissions, x n are the explanatory variables such as the mass of the vehicle, and c 1, c 2, c 3,..., c n are the coefficients with a total number equal to the number of independent variables that exist. This method makes the estimation of coefficients of the model and seeks to minimize the sum of squared estimation errors. The coefficients obtained are subsequently subjected to statistical analysis to verify their statistical significance. To this end, a statistical hypothesis test was done to accept or reject the zero hypothesis of the estimated coefficients being significant, depending on the value of the resulting t. Thus, for a confidence level of 95%, if t value, in modulus, is greater than 2.96, then the coefficient is significant and should be considered. Otherwise, is void and should be disregarded. Since the independent variables were found to present some non-linearity between them, it was decided to apply firstly the (neperian) logarithm function to the dependent variables. The logarithmic transformation of the dependent variables allows not only to ensure the positivity of all estimated values (essential, since in the lifecycle of materials there is always energy consumption and emissions), but also attempts to make the model completely linear. Thus, the model takes the following form: Therefore, after application of the logarithmic function to the dependent variables, it was used the LLS method. Final Models After obtaining the model coefficients of emissions and energy consumption and selecting the statistically significant variables, the final models were grouped by technology, whose coefficients are presented in Appendix (Tables 8 to 12). There were thus obtained 45 models, one for each independent variable and automobile technology. In order to obtain models that produce results in emission units (g/km) and energy units (kj/km), it was necessary to apply the exponential function previously, being each model an exponential function of the independent variables. Thus, the final models are given generically by: 8 For the FCHEV the minimum mass value was 1250 kg because of the use of heavy components like the fuel cell and the tank. 9

10 where Y is a dependent variable, the independent variable x n and c n its coefficient. Note that these models are constrained to be used under certain conditions that are related to the inputs range and type of battery used (Table 6), the maintenance periods assumed for each component and for conventional material based vehicles, rather than lightweight. Validation The models validation was done in three phases: 1) Comparison with the results generated by GREET 2012, 2) Analysis of residues of each model, 3) Comparison with the results of the literature review. 1) To validate the models, the values generated by the program GREET 2012 were compared with the values estimated by the models and the errors were calculated for each estimate according to the following formula, where the subscript i represents the ith estimate of a sample of n estimations. Subsequently, the average error was calculated by the formula, This analysis was performed not only with estimations of the initial sample used for the construction of the models, based on Table 6, but also used a test sample which had different input values from the first. The average errors are shown in Figure 1, for the initial and test samples. As can be seen, for the initial sample, all models have an average estimation error lower than 15%, except for the GHG emission model of the (20%). The SO x average errors seem to be systematically higher for the HEV, PHEV and BEV technologies. For the test sample, the errors show to be higher to all technologies, being the maximum error achieved (34%) for PM 2.5 and AP emissions and BEV technology. In fact these models show clearly a lower performance for all technologies. A higher average error (30%) is also identified for the GHG emissions of the. In general, the smallest errors are achieved for emissions of VOCs, NO x and SO x. Average estimation errors - Initial sample Energy GEE COV CO NOx PM10 PM2.5 SOx AP 30% 20% 10% 0% HEV PHEV BEV FCHEV 10

11 Residues Average estimation errors - Test sample Energy GEE COV CO NOx PM10 PM2.5 SOx AP 40% 30% 20% 10% 0% HEV PHEV BEV FCHEV Figure 1: Average estimation errors for the initial and test variables, for all models BEV log[g/km] 2) Next, a residues analysis was made for each model to see if they were independent, with constant variance and if followed a normal distribution of zero mean Estimated Values Figure 2: Example of a Residue vs. Estimated values plot for NO x emissions The Figure 2 shows an example of the plots of residues vs. estimated values analyzed. In this case, the NO x estimated values range from -4 to -1.5 log(g/km) and the residues from -0.2 to 0.2 log(g/km), where the biggest concentration of points (estimations) is near the line y=0, associated to null residual values. The distribution of the residues doesn t have a clear pattern and the variance is relatively constant. Additionally to that information, the histograms of the residues of all the models were analysed and in general, has been concluded that all models met the criteria, except the estimation models of VOC and AP emissions, which showed certain nonlinearity. This fact can be explained because it is known that more than 80% of the VOC emissions are related to vehicle s fluids (Baudoin, 2007) and since the models have no fluid related inputs, the errors tend to be higher. In relation to the AP models, a more complex estimation must be done since it is a combined pollutant emission category, thus the models adequacy is weaker. 3) In this analysis the scope was to use the developed models to calculate the results of energy and emissions, giving the inputs used for each study of the literature review and to compare them. For example, in the Table 7 the results between the PHEV studies and models are shown and the relative estimation error (having has a reference the study value) is calculated. For the CO 2 model, errors of 22% and 8% are achieved for the first and second studies, respectively. Has it can be seen, the models give lower estimates than the study s results, for all variables. Table 7: PHEV study results vs. models results Study Energy CO 2 CO NO x PM Obs. (P. C. Baptista et al., 2012) PHEV Models GREET (19%) (22%) (68%) (32%) (55%) (P. Baptista et al., 2010) GREET PHEV Models (13%) (8%) A comparative analysis of results from the literature review was inconclusive due to the different methodologies used. 11

12 Conclusions and future developments The automotive sector is nowadays directly regulated in terms of energy consumption and pollutant emissions only in the vehicle use phase. The other phases of the vehicle lifecycle, as the vehicle materials lifecycle, remain unregulated. Moreover, it was concluded that the environmental impacts in the form of energy consumption and pollutants emissions, for the vehicle materials lifecycle, are increasingly significant as the technological evolution of the vehicles takes place. The components that contribute most to this are the propulsion batteries, whose impacts are directly proportional to their mass, and the fuel cell included in FCHEV. Thus, the main objective of this study was to develop simplified models for the estimation of these impacts relating to the vehicle materials lifecycle, which facilitate the comparison between different propulsion technologies and also contribute to the development of a vehicle environmental ranking. Finally, were obtained 45 models that estimate energy consumption and emissions (GHG, NO x, VOC, CO, SO x, PM 10 and PM 2.5), associated to the vehicle materials lifecycle of different vehicle technologies (, HEV, PHEV, BEV, FCHEV), which besides avoiding the use of large databases, are simple to use and have as Inputs variables of public access. If used within the certain conditions in terms of the ranges of the Input variables, of the relative composition of the vehicle materials and of component maintenance periods, these models turn out to be reliable, with acceptable estimation errors: mean absolute errors are lower than 12% for models of VOC, NO x and SO x and below 34% for the other categories. With regard to the possible improvements sought is considered interesting to expand the capabilities of the models in addition to the estimation of energy and emissions of conventional cars, by including for example light-weight vehicles that have less dense materials (e.g. carbon fiber) or biomaterials. Additionally, would be relevant to introduce input variables as the recycled content of Lithium, since this metal is the major constituent of propulsion batteries whose environmental impacts are significant. References Baptista, P. C., Silva, C. M., Farias, T. L., & Heywood, J. B. (2012). Energy and environmental impacts of alternative pathways for the Portuguese road transportation sector. Energy Policy, doi: /j.enpol Baptista, P., Tomás, M., & Silva, C. (2010). Plug-in hybrid fuel cell vehicles market penetration scenarios. International Journal of Hydrogen Energy, 35(18), doi: /j.ijhydene Baudoin, J.-M. (2007). Avaliação energética e ambiental da produção e reciclagem de materiais / componentes de veículos convencionais versus veículos híbridos. Dissertação para obtenção do Grau de Mestre em Engenharia Mecânica. Boureima, F., Messagie, M., Matheys, J., Wynen, V., Mierlo, J. V., Vos, M. D., & Caevel, B. D. (2009). Comparative LCA of electric, hybrid, LPG and gasoline cars in Belgian context. Fuel Cell, 1 8. Boureima, F., Vincent, W., Nele, S., Heijke, R., & Messagie, M. (2009). Clean Vehicles Research : LCA and Policy Measures LCA report. Renewable Energy. Delucchi, M. A. (2001). A lifecycle emissions analysis : urban air pollutants and greenhouse- gases from petroleum, natural gas, LPG, and other fuels for highway vehicles, forklifts, and household heating in the U. S. World Resources Review, 13, Granovskii, M., Dincer, I., & Rosen, M. a. (2006). Economic and environmental comparison of conventional, hybrid, electric and hydrogen fuel cell vehicles. Journal of Power Sources, 159(2), doi: /j.jpowsour Hu, Z., Pu, G., Fang, F., & Wang, C. (2004). Economics, environment, and energy life cycle assessment of automobiles fueled by bio-ethanol blends in China. Renewable Energy, 29(14), doi: /j.renene Leduc, G., Mongelli, I., Uihlein, A., & Nemry, F. (2010). How can our cars become less polluting? An assessment of the environmental improvement potential of cars. Transport Policy, 17, doi: /j.tranpol Liu, J., Wang, R., & Yang, J. (2007). A scenario analysis of Beijing s private traffic patterns. Journal of Cleaner Production, 15(6), doi: /j.jclepro Wu C, Ma X, Chen Y, Zhao Z, L. H. (2006). Vehicle life cycle assessment covering its manufacturing, use and recycling. Automotive Engineering, 28, Zachariadis, T. (2006). On the baseline evolution of automobile fuel economy in Europe. Energy Policy, 34(14), doi: /j.enpol

13 Appendix Table 8: models coefficients Variable Coefficient Energy GEE COV CO NO x PM 10 PM 2.5 SO x AP VehicleWeight c E E E E E E E E E-04 Lifetime c E E E E E E E E E-06 Oil c E E E E E E E E E-02 NaturalGas c E E E E E E E E E-02 Coal c E E E E E E E E E-02 Nuclear c E E E E E E E E E-02 Biomass c E E E E E E E-02 Renewables c E E E E E E E E E-02 RecycledSteel c E E E E E E E E-03 - Recycled W. Aluminum c E E Recycled C. Aluminum c E E E Recycled Lead c E Recycled Nickel c E Table 9: HEV models coefficients Variable Coefficient Energy GEE COV CO NO x PM 10 PM 2.5 SO x AP VehicleWeight c E E E E E E E E E-04 Lifetime c E E E E E E E E E-06 Oil c E E E E E E E E E-02 NaturalGas c E E E E E E E E E-02 Coal c E E E E E E E E E-02 Nuclear c E E E E E E E E E-02 Biomass c E E E E E E E E E-02 Renewables c E E E E E E E E E-02 RecycledSteel c E E E E E E E E E-04 Recycled W. Aluminum c E E E E E E E-04 - Recycled C. Aluminum c E E E E E E Recycled Lead c E Recycled Nickel c E BatteryWeight c E E E E E E E E E-04 Table 10: PHEV models coefficients Variable Coefficient Energy GEE COV CO NO x PM 10 PM 2.5 SO x AP VehicleWeight c E E E E E E E E E-04 Lifetime c E E E E E E E E E-06 Oil c E E E E E E E E E-02 NaturalGas c E E E E E E E E E-02 Coal c E E E E E E E E E-02 Nuclear c E E E E E E E E E-02 Biomass c E E E E E E E E E-02 Renewables c E E E E E E E E E-02 RecycledSteel c E E E E E E E E E-04 Recycled W. Aluminum c E E E E E E-04 - Recycled C. Aluminum c E E E E E Recycled Lead c E Recycled Nickel c E BatteryWeight c E E E E E E E E-04 13

14 Table 11: BEV models coefficients Variable Coefficient Energy GEE COV CO NO x PM 10 PM 2.5 SO x AP VehicleWeight c E E E E E E E E E-04 Lifetime c E E E E E E E E E-06 Oil c E E E E E E E E E-02 NaturalGas c E E E E E E E E E-02 Coal c E E E E E E E E E-02 Nuclear c E E E E E E E E E-02 Biomass c E E E E E E E E E-02 Renewables c E E E E E E E E E-02 RecycledSteel c E E E E E E E E E-04 Recycled W. Aluminum c E E E E E E E-04 Recycled C. Aluminum c E E E E Recycled Lead c E Recycled Nickel c E BatteryWeight c E E E E E E E E E-04 Table 12: FCHEV models coefficients Variable Coefficient Energy GEE COV CO NO x PM 10 PM 2.5 SO x AP VehicleWeight c E E E E E E E E E-04 Lifetime c E E E E E E E E E-06 Oil c E E E E E E E E E-02 NaturalGas c E E E E E E E E E-02 Coal c E E E E E E E E E-02 Nuclear c E E E E E E E E E-02 Biomass c E E E E E E E E E-02 Renewables c E E E E E E E E E-02 RecycledSteel c E E E E E E E E E-04 Recycled W. Aluminum c E E E E E E E E E-04 Recycled C. Aluminum c E E E E E E E-04 - Recycled Lead c E E-04 - Recycled Nickel c E E-04 - BatteryWeight c Fuel Cell Power c E E E E E E E E-03-14