Robert JOUMARD Eric SÉRIÉ MODELLING OF COLD START EMISSIONS FOR PASSENGER CARS

Transcription

1 Robert JOUMARD Eric SÉRIÉ MODELLING OF COLD START EMISSIONS FOR PASSENGER CARS INRETS report LTE 9931 December 1999

2

3 MEET Project - Contract N ST-96-SC.204 Methodologies for estimating air pollutant emissions from transport Task 1.3 : Cold start influence Deliverable N 8 - Public dissemination - Project funded by the European Commission under the transport RTD programme of the 4th framework programme. COST 319 Action Estimation of pollutant emissions from transport Subgroup A3b : Cold start emissions Robert JOUMARD Eric SÉRIÉ MODELLING OF COLD START EMISSIONS FOR PASSENGER CARS INRETS report LTE nd version December 1999

4 Authors : Robert JOUMARD, Senior Researcher, a specialist in air pollution research, LTE Eric SÉRIÉ, scientist in atmospheric physics, LEN Research units: LTE : LEN : Laboratoire Transports et Environnement, INRETS, case 24, Bron cedex, France. Tel.: +33 (0) Fax: +33 (0) joumard@inrets.fr ; Laboratoire Énergie Nuisances, INRETS. Acknowledgements The authors would like to thank all the laboratories which provided them with data. and more particularly people who sent data and comments : Mr Hammarström from VTI (Sweden), Mr Hassel and Mr Weber from TÜV Rheinland (Germany), Mr Laurikko from VTT (Finland), Mr Cernushi from Politecnico di Milano (Italy), Mr Rijkeboer from TNO (The Netherlands), Mr Jammernegg, Mr Hausberger and Mr Sturm from TU-Graz (Austria), Mr Hickman from TRL (England), Mr Keller from Infras (Switzerland), Mr Laguna from INTA (Spain), Mr Samaras and Mr Ntziachristos from Aristotle University Thessaloniki (Greece) and Mr Ramella from the Polytechnic School of Turin (Italy). Their remarks were very useful to improve this document. 2 INRETS report LTE 9931

5 Publication data form 1 UR (1st author) 4 Title LTE Modelling of cold start emissions for passenger cars 2 Project n 3 INRETS report n LTE Subtitle 6 Language E 7 Author(s) JOUMARD Robert SÉRIÉ Eric 8 Affiliation 9 Sponsor, co-editor, name and address 10 Contract, conv. n ST-96-SC Notes 11 Publication date December Summary After a survey among 39 European laboratories, data were obtained concerning 661 vehicles and measurements (1 measurement corresponds to 1 vehicle, 1 cycle and 1 pollutant). Data were measured during standardised and representative cycles. The data received were analysed in order to model cold-start-related excess emissions (defined as the difference between cold and hot emissions, for a same cycle) for 4 pollutants (CO, CO 2, HC, NOx) and fuel consumption as a function of various parameters: technology (diesel, catalyst and conventional cars), average speed, ambient temperature and travelled distance. In a second step the model is extended from a single trip to a whole traffic characterised by a number of parameters such as vehicle flow, average trip speed and environmental conditions, using measured distributions of average speed during cold start conditions versus overall average speed, engine start-up temperature versus ambient temperature, and cold start trip number versus trip length. A simple model is built, giving unit cold excess emissions in g/km, as a function of the average speed, the ambient temperature and the season. The model is provided for gasoline and diesel cars, present and near future till EURO 4 standard vehicles. The proposed final model is part of an emission inventory model. 14 Key Words Transport, emission, pollutant, cold start, driving cycle, catalyst, diesel, passenger car, ambient temperature, engine temperature, model, speed, travelled distance. 15 Distribution statement limited free X 16 Nb of pages 86 pages 17 Price F 18 Declassification date 19 Bibliography yes INRETS report LTE

6 Fiche bibliographique 1 UR (1er auteur) LTE 4 Titre 2 Projet n 3 Rapport INRETS n LTE 9931 Modélisation des émissions lors du départ à froid des voitures particulières 5 Sous-titre 6 Langue E 7 Auteur(s) JOUMARD Robert SÉRIÉ Éric 8 Rattachement ext. 9 Nom adresse financeur, co-éditeur 10 N contrat, conv. ST-96-SC Remarques 11 Date de publication décembre Résumé Après enquête auprès de 39 laboratoires européens, nous avons obtenu des données concernant 661 véhicules et mesures (1 mesure correspond à un véhicule, un cycle et un polluant). Les données ont été obtenues lors de mesures effectuées durant des cycles normalisés et des cycles représentatifs. Nous avons analysé ces données afin de modéliser la surémission des voitures particulières lors d'un départ à froid (définie comme la différence entre l émission à froid et l émission à chaud, pour un même cycle) pour 4 polluants (CO, CO 2, HC, NOx) et la consommation de carburant en fonction de divers paramètres : technologie (voitures diesel, essence avec et sans catalyseur), vitesse moyenne, température ambiante et distance parcourue. Dans un deuxième temps, le modèle est étendu d'un seul trajet à un trafic de véhicules, caractérisé par les paramètres flux de trafic, vitesse moyenne, et conditions environnementales. On utilise pour cela des distributions statistiques, issues de mesures, qui concernent la vitesse moyenne à froid en fonction de la vitesse moyenne générale, la température moteur de départ en fonction de la température ambiante, et le nombre de départs à froid selon la longueur des trajets. On obtient un modèle simple exprimant une surémission unitaire à froid en g/km, fonction de la vitesse moyenne, de la température ambiante et de la saison. Ce modèle est fourni pour les voitures essence et diesel actuelles et futures jusqu'à la norme EURO 4. Le modèle final proposé fait partie d un modèle d inventaire des émissions. 14 Mots clés Transport, émission, polluant, départ à froid, cycle de conduite, catalyseur, diesel, voiture particulière, température ambiante, température du moteur, modèle, vitesse, distance parcourue. 16 Nombre de pages 86 pages 17 Prix F 15 Diffusion restreinte libre X 18 Confidentiel jusqu'au 19 Bibliographie oui 4 INRETS report LTE 9931

7 Modelling of cold start emissions for passenger cars Content INTRODUCTION DATA Initial data Data correction General method Results INFLUENCE OF VARIOUS PARAMETERS Excess emission as a function of the cycle speed Excess emission as a function of ambient temperature Excess emission as a function of the travelled distance CALCULATION METHOD General formula of cold-start-related excess emissions of a trip Gasoline cars (with or without catalyst), Diesel cars without catalyst Diesel cars with catalyst Light duty vehicles (LDV) Heavy duty vehicles (HDV) Excess emissions for trip starting under engine mean temperature conditions Parking time approach Simplified approach Inventory of cold-start-related excess emissions General approach Calculation mode Final model Near future vehicles Discussion of the results...12 CONCLUSION...12 Annex 1: Laboratory acronyms, addresses and persons to contact...12 Annex 2: Vehicle distribution and type of cycles...12 Annex 3: Excess emissions (g) versus ambient temperature ( C)...12 Annex 3.1: Gasoline cars without catalyst...12 Annex 3.2: Diesel cars without catalyst...12 Annex 3.3: Gasoline cars with 3-way catalyst...12 Annex 4: Cold distance (km) as a function of the vehicle speed (km/h)...12 Annex 4.1: Gasoline cars without catalyst...12 Annex 4.2: Diesel cars without catalyst...12 Annex 5: Dimensionless excess emission versus dimensionless distance...12 Annex 5.1: Gasoline cars without catalyst...12 Annex 5.2: Diesel cars without catalyst...12 Annex 6: Dimensionless excess emission versus dimensionless distance. Case of NOx pollutant for gasoline cars without catalyst...12 Annex 7: Effect of air humidity for NOx pollutant...12 Annex 8: Parking time correction coefficient...12 Annex 9: Mileage percentage of the trips started at cold or intermediate engine temperature cm(s,vi)...12 INRETS report LTE

8 Modelling of cold start emissions for passenger cars Annex 10: Distribution pj of the trips as regards the cold average speed Vj and the overall average speed (%)...12 Annex 11: Percentage pk of the trips travelled with a start-up engine temperature Tk (%), according to the ambient temperature ( C)...12 Annex 12: Percentage pm of trips started with a cold engine and distance dm, for speed Vj with a cold engine (%)...12 Annex 13: Cold excess unit emission according to average speed and ambient temperature for conventional gasoline cars according to the season...12 Annex 14: Cold excess unit emission according to average speed and ambient temperature for conventional diesel cars according to the season...12 Annex 15: Cold excess unit emission according to average speed and ambient temperature for gasoline cars with catalyst according to the season...12 Annex 16: Cold excess unit emission according to average speed and ambient temperature for gasoline cars with catalyst EURO 2 according to the season...12 Annex 17: Cold excess unit emission according to average speed and ambient temperature for gasoline cars with catalyst EURO 3 according to the season...12 Annex 18: Cold excess unit emission according to average speed and ambient temperature for gasoline cars with catalyst EURO 4 according to the season...12 List of figures and tables...12 REFERENCES INRETS report LTE 9931

9 Modelling of cold start emissions for passenger cars Introduction This report is the result of a study carried out as part of COST 319 action and MEET programme. These two projects result from a working group named CORINAIR, working on emissions factors for calculating emissions from road traffic (Eggleston et al., 1993). In this group, the methodology developed for calculating the cold start emission was to introduce relative cold emission (cold/hot emission ratio) as a function of ambient temperature and trip length. For cold start emissions, indicators of the emission estimate were mainly defined from small set of measured data or from the available literature. It has been thus decided to update continuously these parameters and to make a thorough emission estimate. The COST 319 action (European Co-operation in the Field of Scientific and Technical Research) is an Europe-wide programme for the co-ordination of national research studies. Due to the fact that traffic is the major contribution to air pollution, considerable research effort addresses the problem of transport and the environment. This activity can be split into two major headings : - studies aimed at estimating the current situation and forecasting the future - studies aimed at developing and evaluating solutions. Today, the action involves over 80 active scientists from 24 countries. The wide field covered by the action imposed the formation of four main working groups : - emission factors and functions : engine maps, instantaneous vehicle emissions, average vehicle emissions, future vehicle and life cycle emissions - traffic characteristics : traffic management, driving behaviour, traffic composition, mobility. - inventorying tools : bottom-up and top-down approaches - non-road transport : inland, maritime and air transport. The COST programme includes three successive phases. First, a survey was carried out among a number of partners in order to collect data. Data was then analysed to build an emission model and eventually discussions were held with all the signatories. The MEET project (Methodologies for Estimating Air Pollution Emissions from Transport) is developing a comprehensive methodology and has three main objectives : - to provide a set of data and models to calculate pollutant emissions and fuel consumption of various transport modes - to develop a comprehensive calculation method from the available set of data and models - to make sure that this comprehensive method corresponds to the requirements of the potential users in terms of accuracy, simplicity and input data availability. Three phases are necessary to meet the general objectives of the project : - users' requirements : they were already identified in COST analysis phase : to collect and analyse emission data, elaborate emission sub-models specific to given circumstances (cold start, life cycle...) and to collect the external parameters which affect emissions - synthesis phase : to homogenise the different sub-models. The operational links between COST and MEET will be ensured by the bodies participating in both projects. Most of the outputs of this COST319/MEET project are available at including final reports (Joumard, 1999; Hickman et al., 1999). INRETS report LTE

10 Introduction This report is the result of a collaboration between many research organisations and has been reread by various researchers being part of COST and MEET actions. The aim of this study consists in modelling the cold start impact on road vehicle emissions as functions of the pollutant and the vehicle type, using all the collected data from COST and to propose a model to be discussed in COST and MEET. This model has been developed empirically considering the available data : excess emissions indeed, but also ambient temperature, start temperature, travelled distance and average speed. Measurements were made with passenger cars. In addition we propose an extension to duty vehicles and diesel passenger cars equipped with oxidation catalysts. In the following analysis, the terms "cold effect" and "cold emissions" will be considered. There are five different ways of presenting cold effect results. These are : - average cold emission factors (g/km) of the first (cold) cycle, - absolute emissions (g) per cold cycle, - the difference of average emission factors (g/km) between cold and hot cycles, - the ratio of cold and hot emissions (relative cold emission), - absolute excess cold start emissions (g) defined as the additional emission value obtained under cold conditions compared to the emission value that could have been recorded for the same period (cycle) under hot conditions. In this report, cold-engine-related excess emissions are addressed. 8 INRETS report LTE 9931

11 Modelling of cold start emissions for passenger cars 1. Data 1.1. Initial data In January 1994 an inquiry was sent to various laboratories studying vehicles emissions under cold start conditions as part of the COST 319 action, subgroup A3B (cold start emission). We obtained data from TNO, INRETS, TU-Graz, TÜV, Politecnico di Milano, TRL, EMPA, INTA, VTT, VTI, KTI, LAT (see Annex 1 for laboratory acronyms). We asked for information about cold start tests : number of test vehicles, vehicle type and characteristics (fuel type, model year, engine capacity), driving cycle type, ambient temperature, start condition, emission measurements. Only a few number of laboratories were able to provide data relative to cold start, i. e. measuring cold and hot emissions over a same driving cycle. The list of available data is given in Annex 2. A data represents one measurement for one vehicle during a cold cycle and a same hot cycle, independently of the pollutant recorded. Passenger cars were divided into four categories : - gasoline cars without catalyst - gasoline cars with catalyst - diesel cars without catalyst - diesel cars with catalyst For each vehicle, 3 types of cold and hot cycles were possibly followed (a summary description of these cycles is shown in Table 1.): - standardised cycles : ECE-15, ECE-15-1 (TRL laboratory), FTP short cycles (Inrets laboratory) : short slow urban, short free-flow urban and short road ; each of these cycles was repeated 15 times. These cycles were drawn up from travelled kilometres previously recorded all over France by 35 private cars (EUREV study, André, 1989) - cycles developed at TRL : congested traffic, TRL 1 and TRL 2. In addition TRL developed an other cycle : the urban road test. But for the hot and cold cycles, temperature conditions were differing and we did not take this cycle into account. Moreover, data concerning stabilised phases (in terms of temperature), such as FTP72-2 and the European EUDC, are not considered in the analysis below. It should be noted that the ECE-15-1 cycle corresponds to the first cycle of the ECE-15 cycle (i.e. a quarter). Concerning excess emission data as a function of the cycle, the total number of obtained data was 460 (gasoline cars without catalyst), 1784 (gasoline cars with catalyst), 315 (diesel cars without catalyst) and 9 (diesel cars with catalyst). All samples were selected by a number of laboratories so that the distribution was representative, to some extent, of the fleet corresponding to each country. The number of vehicles tested by each laboratory and the corresponding cycles are shown in Table 2 for 3-way catalyst cars. It should be noted that, for the main part of the Inrets data, the vehicles are the same for each cycle (short cycles or FTP cycle). It is thus possible to compare the obtained results. INRETS report LTE

12 Data Type Name Duration (s) Distance (m) Average speed (km/h) FTP Standard ECE ECE short slow urban Inrets short short free-flow urban short road congested traffic TRL cycles TRL TRL Table 1: Details of the various driving cycles. Measurements made with the same vehicles Lab CO 2 CO HC NOx FC Cycle Inrets Short slow urban Inrets Short free-flow urban Inrets Short road Inrets FTP72-1 TNO FTP72-1 TÜV FTP72-1 Inrets ECE-15 TNO ECE-15 TNO ECE-15 TÜV ECE-15 LAT ECE-15 TRL ECE-15 Table 2: Vehicle distribution versus average speed to calculate excess emission for gasoline cars with catalyst. For ambient-temperature-related excess emission data, we used data measured under various ambient temperature conditions (-9 C to 26 C). Table 3 shows a number of analysed data and the temperature range. Laboratory CO 2 CO HC NOx FC Cycle Temp. range ( C) Inrets FTP to 20 TNO FTP TÜV FTP to 20 TRL ECE to 26 Table 3: Number of data analysed in order to obtain excess emission as a function of ambient temperature for gasoline cars with catalyst. It should be noted that : - for diesel cars with oxidation catalyst, the data are rare and do not allow a full data processing, as for other vehicle categories. - for NOx pollutant, some laboratories made humidity correction but not all of them. We did not take into account such a correction factor and we think it would be better to have data without 10 INRETS report LTE 9931

13 Modelling of cold start emissions for passenger cars humidity corrections, i.e. actual emissions. Concerning these latter, Annex 7 gives the computation of the humidity correction factor Data correction General method Once all the data were collected, we had to take into account a number of parameters influencing the general method : - Great variety of data : - as can be seen in section 1.1, the number of vehicles analysed with standardised cycles is very significant ; but such cycles are not representative since they do not reflect the reality. Comparing, for a same speed, the standard deviation of acceleration between Inrets and standard cycles [Joumard et al. (1995a)] yielded significantly differing results, acceleration standard deviation being lower than for standardised cycles. - the representative cycles (real cycles) enabled a fine description of the emission evolution, but there was a limited number of analysed vehicles. - when the vehicles are tested using a standard cycle, the engine temperature is not always hot at the end of the cycle, according to Joumard et al. (1995b). Therefore, we had to introduce a light adjustment for each pollutant. Thus we obtained excess emissions over the entire cold period for different cycles, whether standard or not. Such an adjustment is needed especially for ECE-15-1 cycle since it is very short. - Adjustment of cold excess emissions measured during real cycles using deviation between the FTP cycles and these measures. This allowed us to take into account the large number of measurement data available from FTP cycles (with non representative conditions), and the small number of representative measurements. We were thus able to deduce the influence of average speed. - Ambient temperature : it must be taken into account, if possible, whatever the mean speed may be. Therefore, we have to look for a relation independent of the average speed, considering that initially the starts are made with a cold engine (engine temperature corresponds to ambient temperature at start). - we make the following hypothesis : excess emission depends on the engine start temperature only (as temperature parameter), this one being equal to the ambient temperature (real cold start) or greater than the ambient temperature (semi-cold or semi-hot engine). This hypothesis is necessary due to the lack of data. - taking into account the travelled distance after a cold start in order to decrease excess emission levels if the travelled distance is lower than the cold distance (i. e. the distance needed to stabilise the emission). It should be noted that some measurements correspond to the same cycle. But, they correspond to measurements made by various laboratories using various measuring devices, in various conditions, with various car models, different vehicle ages, etc.. So it results in differences for the same cycles themselves. INRETS report LTE

14 Data Results A recent study (Joumard et al., 1995b) showed that FTP72-1 and ECE-15 cycles do not cover entirely the cold period due to the cold start. So, we introduced a correction coefficient for these cycles to transform the measured excess emission during standard cycles into a full cold excess emission. This coefficient is deduced from measurement data recorded using Inrets short cycles, which cover the whole cold period. The distance necessary to warm up the engine is defined as the distance needed to stabilise the emission level (vehicles with constant emission). Using this cold distance (see Table 4) and the Inrets short cycles data, we calculate the correction coefficient to be applied to adjust the standardised cycles to the representative cycles. For example (see Figure 1), the ECE-15 cycle corresponds to an average speed of 18.7 km/h and a distance of 4052 m. For CO 2 pollutant, the cold distance (distance necessary to stabilise the emission level) is equal to 5.9 km for the representative cycle with the nearest average speed, i.e km/h (from Table 4). Regarding excess emission (normalised by the total excess emission) as a function of the distance, the ECE-15 cycle corresponds to 91 % of the total excess emission of the short free-flow urban cycle. Then the factor is equal to 1.10 (=1/0.91). We applied this method to all the pollutants and fuel consumption levels and for ECE-15, FTP and TRL cycles (see table 5). From Table 4, it can be noted that there are different cold distances between CO 2 and fuel consumption (FC). Small differences are observed for catalyst cars, but for diesel and non catalyst cars, there is a greater difference due to the different behaviour of CO 2, CO, HC and FC. Theoretically, owing to the carbon balance (see section 2.1.), this does not seem possible, but it is due to the inaccuracy in distance evaluation, when we consider the asymptotic shape of the curve (see an example Figure 1). Nevertheless we took into account these values since they correspond to real measurements. dimensionless excess emission 1 0,9 0,8 0,7 0,6 0,5 0,4 0,3 0,2 0, short Inrets cycle distance (km) distance of the ECE-15 cycle: km cold distance for CO2 pollutant: 5.9 km Figure 1: Cumulative dimensionless excess emission (ratio of absolute excess cold start emission to total absolute excess cold start emission) as a function of the distance (km) for short free-flow urban cycle. Correction calculation example of ECE-15 cycle for CO 2 pollutant and gasoline cars with catalyst. and gasoline cars with catalyst. 12 INRETS report LTE 9931

15 Modelling of cold start emissions for passenger cars Technology Speed (km/h) CO 2 CO HC NOx FC Gasoline cars without catalyst Diesel cars without catalyst Gasoline cars with catalyst Table 4: Distance (km) necessary to warm up the engine according to the pollutant and the cycle mean speed (km/h). (Joumard et al., 1995b) ; FC : Fuel consumption. Technology Cycle CO 2 CO HC NOx FC FTP ECE Gasoline cars ECE without Congested catalyst TRL TRL FTP ECE Diesel cars ECE without Congested catalyst TRL TRL FTP ECE Gasoline cars ECE with catalyst Congested TRL TRL Table 5: Correction factor of cold excess emission for various cycles, to take into account the too short distance of the cycles. Concerning ECE-15-1 cycle, this cycle corresponds to the first mini-cycle of ECE-15 cycle (including 4 mini-cycles). Indeed, correction coefficients for this cycle are much higher than for standard cycles. For some pollutants, we observed correction factors lower than 1. This is due to the fact that, in these cases, there is an excess emission level followed by a lowest emission level (amounting globally to an excess emission). INRETS report LTE

16

17 Modelling of cold start emissions for passenger cars 2. Influence of various parameters 2.1. Excess emission as a function of the cycle speed Only data concerning gasoline cars (with and without catalyst) and diesel cars without catalyst. are currently available For diesel cars with catalysts, measurements were made for only one cycle (FTP72-1) and therefore did not allow to take into account the speed for this technology type. In Figure 2, an example of excess emission as a function of the average speed for gasoline cars with catalyst is shown. The successive corrections are calculated and applied as follows : - data correction for standard cycles, as explained in section in order to homogenise the recorded data, the all laboratories data/inrets data ratio for FTP cycle excess emissions was calculated (see Table 6). For NOx emissions from diesel cars, this ratio is supposed equal to 1, since excess emission levels were recorded by the other laboratories and reduced emission levels were recorded by Inrets. - correction of Inrets short cycle emission levels according to these ratios. Then, using modified short Inrets data, we applied a linear regression in order to obtain the excess emission level [g] as a function of the average speed V [km/h] (see Table 7). When correlation coefficient of the regression was low (especially for CO 2 and FC in the case of diesel cars), the excess emission was supposed equal to a constant amounting to the mean value of all the recorded data. For CO and HC emissions from diesel cars, the calculated equation has to be positive. Therefore, when average speed is greater than 74 km/h for CO and 81.5 km/h for HC (concerning diesel cars), this equation has to be set equal to 0. It should be noticed that the regression is calculated using only three measurement points, each point corresponding to an average of ten measurements. So, the correlation coefficient has to be relativized, especially for a coefficient equal to 1. Using the calculated equations, we determined the correction coefficients (see Table 7) corresponding to the functions made dimensionless by dividing them by their values recorded at 20 km/h. For vehicle speeds greater than 50 km/h, these equations are valid using a linear extrapolation, as far as the boundary speed is not exceeded. For all the data, fuel consumption was calculated using the carbon balance method. We used the formula (1) where the ratio hydrogen/carbon r H/C is equal to 1.8 for gasoline (leaded or unleaded) and 2.0 for diesel. The HC mass has to be expressed in CH 4 equivalent. Fuel mass r H/C = CO mass 2 CO mass HC mass (Joumard et al., 1995b) + Particle mass This equation indicates that the fuel consumption is not the only factor of the CO 2 emission. Joumard et al. (1990) showed, for gasoline vehicles without catalyst, that the fuel conversion rates respectively into CO 2, CO and HC equals on average 78 %, 19 % and 4 %, with little variations depending on average speed. (1) INRETS report LTE

18 Influence of various parameters Absolute excess emission (g) Inrets lab - FTP cycle calculated ratio : 0.62 applied ratio : 0.62 All labs - ECE-15 cycle All labs FTP cycle short Inrets cycles, Inrets data : initial standard cycles, all labs : initial standard cycles, all labs : after first correction FTP cycle, Inrets data : initial linear regression, final curve linear regression, Inrets initial Speed (km/h) Figure 2: Method of calculation of CO absolute excess emission (g) as a function of the average speed (km/h) for gasoline cars with catalyst. Ratio CO 2 CO HC NOx FC Gasoline cars without catalyst Diesel cars without catalyst Gasoline cars with catalyst Table 6: All laboratories data Vs Inrets lab data for FTP cycle. Techno. Pollutant Correlation coeff. Excess emission equation ω V Correction coefficient f(v) Boundary Speed Gasoline CO 2 CO V V V V V<119 km/h - cars HC V V without NOx V V V>11 km/h catalyst FC (calc.) V V Diesel CO 2 CO V V V<74 km/h cars HC V V V<81 km/h without NOx V V V<64 km/h catalyst FC (calc.) Gasoline CO 2 CO V V V V cars HC V V with NOx V V V>5 km/h catalyst FC (calc.) V V Table 7: Equation describing the influence of mean speed (km/h) on excess emission V (g) and the associated dimensionless correction coefficients f(v) and boundaries. This equation results in a linear regression (best fitted line). The boundaries correspond to f(v) INRETS report LTE 9931

19 Modelling of cold start emissions for passenger cars 2.2. Excess emission as a function of ambient temperature Once the corrections over standard cycles (section ) were applied, in first approximation, we can determine the relation between cold start excess emission and ambient temperature. Ambient and engine temperatures were considered similar under cold start conditions. All the processed data concern gasoline cars (with and without catalyst) and diesel cars without catalyst. Some reports show a dependence on ambient temperature, especially for CO and HC (Keller et al., 1995 ; Rijkeboer and Havenith, 1996). It should be noted that the collected data are very scattered. Example of CO pollutant of catalyst car is shown in Figure 3 (other pollutants and technologies are given in Annex 3). We established a regression between excess emission (g) and ambient temperature T ( C). Results are given in Table 8. The determination coefficient represents the reliability of the regression curve and is equal to : ( Y i Y ˆ i ) 2 i where n corresponds to the total number of data, Y 2 i to the ith data and Y ˆ i to Y i 2 Y i i n i the estimated ith data CO CO Excess emission (g) Temperature ( C) Figure 3: Excess emission (g) as a function of ambient temperature for CO pollutant for gasoline cars with catalyst. A point represents a data and the line the data-related regression curve. Concerning CO 2 and NOx pollutants for gasoline cars, we didn t find a good correlation between these variables and the temperature. Lenner (1994) from Swedish Road and Transport Research Institute found a very low correlation coefficient for NOx pollutant too and Joumard et al. (1990) consider that CO 2 excess emission remains constant whatever the temperature may be. Cold start emissions of CO 2 and NOx for catalyst cars were thus approximated to be constant and equal to the average, i.e g and 1.17 g respectively. Therefore temperature correction INRETS report LTE

20 Influence of various parameters should not be applied for these two pollutants. The latter value for NOx is comparable to that found by Lenner (1994), i. e g. Concerning CO, HC, and FC, a reasonable correlation is found (see Table 8). The above equations enabled to calculate the correction coefficient (see Table 9) using a reference temperature of 20 C. For ambient temperatures higher than the reference temperature, the correction coefficient is to be equal to 0, whatever the pollutant may be. It can be seen that, in most cases, the increase in excess emission corresponds to a temperature decrease. For catalyst cars, Shayler et al. (1996) from the University of Nottingham found a fuel penalty increase factor of 2.1 when temperature decreases from 20 C to -20 C. Using the equation above, we found a fuel penalty increase factor of 3.2 for the same temperature decrease. According to Keller et al. (1995), the emission increase of CO and HC is produced by a delayed catalyst heat up, causing a steady decreasing in the air fuel ratio under cold start conditions. Technology Pollutant Correlation Determination Equation of ω T (mean chosen in the coefficient coefficient case of low correl. coeff.) Gasoline cars CO 2 CO T (148.0) T without HC T catalyst NOx T-0.51 (-0.15) FC (calc.) T Diesel cars CO 2 CO T T+3.66 without HC T+1.21 catalyst NOx T+1.56 FC (calc.) T+82.1 Gasoline cars CO 2 CO T (123.5) T with catalyst HC T NOx T+0.78 (1.17) FC (calc.) T Table 8 : Formula describing the excess emission T (g) as a function of ambient temperature T ( C) and the associated correlation and determination coefficients. In the case of low coefficient correlation (case of CO 2 and NOx pollutants for gasoline cars), the mean is preferred to the regression curve Excess emission as a function of the travelled distance Using all the short cycles Inrets data (for gasoline cars with and without catalyst and diesel cars without catalyst), excess emission was calculated as a function of the cycle distance. But if the travelled distance is lower than this cold distance (distance necessary to stabilise the emission), excess emission is lower. In order to calculate the excess emission versus distance equation, we first computed the cold distance as a function of the vehicle speed, technology, temperature and the studied pollutant. Then the travelled distance was made dimensionless by dividing it by the calculated cold distance and the equation describing excess emission was determined as a function of the travelled distance. To obtain excess emission as a function of travelled distance, this distance was considered dimensionless. In addition, cold distance d c in km (Joumard et al., 1995b) was calculated as a Technology Pollutant Correction coefficient g(t) Boundary T CO INRETS report LTE 9931

21 Modelling of cold start emissions for passenger cars Gasoline cars CO T T<30 C without catalyst HC T T<27 C NOx 1 - FC T T<43 C CO T T<41 C Diesel cars CO T T<36 C without catalyst HC T T<30 C NOx T T<31 C FC T T<42 C CO Gasoline cars CO T T<23 C with catalyst HC T T<27 C NOx 1 - FC T T<38 C Table 9: Correction coefficient to apply to the general formula for ambient temperatures (T in C) and their associated requirements. The boundaries correspond to g(t) 0. Technology Pollutant cold distance d c Correlation coefficient Boundary distance CO d c =0.15 V Gasoline cars CO d c =0.04 V without catalyst HC d c =0.09 V NOx d c =0.02 V FC d c =0.28 V CO d c =0.24 V Diesel cars CO d c =0.08 V without catalyst HC d c =0.08 V NOx d c =-0.07 V d c 0 FC d c =0.13 V CO d c =0.29 V d c 0 Gasoline cars CO d c =0.24 V d c 0 with catalyst HC d c =0.06 V NOx d c =0.19 V FC d c =0.24 V Table 10: Equation describing the cold distance d c (km) as a function of the average speed V (km/h). The boundary corresponds to d c 0. function of vehicle speed V in km/h (see Table 10). Due to the fact that measurements of cold distance were not performed as a function of the temperature, d c was assumed independent of the temperature. From the only three measurement points available, we obtained the functions by a linear regression : see the curves in Figure 4 for catalyst cars and Annex 4 for diesel and non catalyst INRETS report LTE

22 Influence of various parameters cars (owing to the low data number, the correlation coefficient has to be relativized). Thus, if the travelled distance is higher than d c then excess emission is equal to the calculated one in sections 2.1. and 2.2. Otherwise, we have to calculate excess emission as a function of the travelled distance d made dimensionless by dividing it by cold distance d c. Using excess emission as a function of the distance (see Figure 5 for catalyst cars and Annex 5 for other technologies), we calculate the equation describing this dependence. Equation is written in the form: 1 e overemission = overemission( a d d c V,T) (2) 1- e a where a is deduced from Figure 4 (for non catalyst and diesel cars, see Annex 4). We proposed one equation per pollutant. We obtained the equations given in Table 11 with their associated correlation coefficients. The correlation coefficient is calculated using a variable change, due to the nonlinearity of the curve. Thus, we obtained a decreasing linear curve and a negative correlation coefficient. The confidence interval reflects variation in the error, as well as variation in the parameter estimates and is calculated for the mean population. It should be noted that the chosen exponential function is well representative of the majority of the data. But in some cases, especially for NOx, the shape is much more complex, and often depends on the cycle speed (as shown in Annex 6 for NOx pollutant of gasoline cars without catalyst). As we prefer to model only the influence of the distance, we propose to use the exponential function in all the cases Cold distance (km) CO2 CO HC NOx FC Linear regression (CO2) Linear regression (CO) Linear regression (HC) Linear regression (NOx) Linear regression (FC) Speed (km/h) Figure 4: Cold distance (km) as a function of the vehicle speed (km/h) for gasoline cars with catalyst. 20 INRETS report LTE 9931

23 Modelling of cold start emissions for passenger cars 1 Dimensionless excess emission 0,9 0,8 0,7 0,6 0,5 0,4 0,3 0,2 0,1 0 CO FC HC CO2 NOx 0 0,1 0,2 0,3 0,4 0,5 0,6 0,7 0,8 0,9 1 Dimensionless distance CO2 CO HC NOx FC CO2 regression CO regression HC regression NOx regression FC regression Figure 5: Dimensionless excess emission as a function of the dimensionless distance (travelled distance made dimensionless by dividing it by the cold distance) and the associated regression. Case of gasoline cars with catalyst. Technology Pollutant Function h (δ) Correlation coefficient Mean and 95 % confidence interval CO 1-exp(-2.85 δ) ±0.15 Gasoline cars CO 1-exp(-6.70 δ) ±0.54 without HC 1-exp( δ) ±0.23 catalyst NOx 1-exp(-2.54 δ) ±1.3 FC 1-exp(-7.97 δ) ±0.15 CO 1-exp(-3.95 δ) ±0.17 Diesel cars CO 1-exp(-3.43 δ) ±0.14 without HC 1-exp(-2.48 δ) ±0.14 catalyst NOx 1-exp(-0.89 δ) ±0.16 FC 1-exp( δ) ±0.14 CO 1-exp(-3.01 δ) ±0.15 Gasoline cars CO 1-exp( δ) ±0.16 with catalyst HC 1-exp(-7.02 δ) ±0.2 NOx 1-exp(-2.30 δ) ±0.23 FC 1-exp(-7.55 δ) ±0.35 This great value is due to the low number of values (3 obtained). Table 11: Equation describing the dimensionless excess emission as a function of the dimensionless distance ( =d/d c ). Note that the function h( ) defined in section is equal to h ( )/h (1). INRETS report LTE

24

25 Modelling of cold start emissions for passenger cars 3. Calculation method We assumed that the general model is a function of temperature, average speed and travelled distance. Measurements were made using different cycles, these being characterised by their mean speed. The cycles can be characterised by other parameters such as mean product speed times acceleration (dynamics measurement), standard deviation of speed or standard deviation of the product speed times acceleration for example (Hassel and Weber, 1996). But, those parameters can not be used in the general model due to the fact that there is no available statistics General formula of cold-start-related excess emissions of a trip Gasoline cars (with or without catalyst), Diesel cars without catalyst In the previous pages, excess emissions (in g) are expressed versus average speed on the one hand and separately as a function of ambient temperature on the other, and eventually considering the distance travelled. Average speed and ambient temperature have an impact on overall excess emissions, i.e. over the whole distance covered with a cold engine. The general method must integrate these three functions ω V, ω T and h'. As cross impact values for speed, temperature and distance are not available, a number of extrapolations and assumptions are required. It is then first proposed to consider distance using a multiplying correction coefficient h(δ), δ being the ratio of the travelled distance to the distance covered under cold engine conditions. This function must be standardised for δ = 1. Thus we get h(δ) = h'(δ) / h'(1), where h'(δ) corresponds to the distance impact. Then speed (ω V ) and temperature (ω T ) impacts must be cumulated considering only common measurement conditions, i.e. those close to 20 km/h and +20 C. It is thus proposed to calculate excess emission under above speed and temperature conditions, and then to apply a correction coefficient as a function of speed and temperature. ω V (20) differing from ω T (20), basic excess emission ω is defined as the average value of these two values: ω = (ω V (20 km/h) + ω T (20 C)) / 2 Dimensionless values for ω V and ω T must be obtained to describe speed and temperature impacts; we thus get the following f(v) and g(t) functions: f(v) =ω V (V) / ω V (20) g(t) = ω T (T) / ω T (20) Combining f and g can be reached using various procedures: extrapolation can be multiplicative or additive. The main drawback of multiplicative extrapolation is error multiplication. Additive extrapolation is thus preferred, adding only impact values and therefore possible errors in order to minimise the total error when extrapolations are used. This hypothesis has to be made as no INRETS report LTE

26 Calculation method cross-distribution value for these two variables has been made available. Excess emission is then expressed: excess emission= [ ( ) + g( T) 1] h( ) f V where : excess emission for a trip is expressed in g V : mean speed in km/h during the cold period T : temperature in C (ambient temperature for cold start, engine start temperature for starts at an intermediate temperature - see section 3.2) = d d : travelled distance d : undimensionned distance with c d c : cold distance ω : reference excess emission (at 20 C and 20 km/h) and the functions f, g and h can be found respectively in Table 7, Table 9 and Tables 10 and 11. The ω coefficient can be found in Table 12. The functions f, g and f+g-1 must be positive or null. Technology Pollutant ω (g) CO Gasoline cars CO without HC 8.23 catalyst NOx FC CO Diesel cars CO 2.18 without HC 0.82 catalyst NOx 0.06 FC CO Gasoline cars CO with catalyst HC 4.62 NOx 1.77 FC CO Diesel cars CO 0.74 with HC 0.65 catalyst NOx 0.03 FC 55.4 Table 12: Coefficient corresponding to excess emission at 20 C and 20 km/h (in g). 24 INRETS report LTE 9931

27 Modelling of cold start emissions for passenger cars For example, for CO pollutant and catalyst cars : - using Table 7, f(v)= V using Table 9, g(t)= T using Table 11, h(δ)=(1-e δ )/(1-e ) and δ is deduced from Table 10 : δ=d/(0.24 V-0.14) - using Table 12, ω=28.71 and therefore : excess emission= ( V ) ( T ) d 1 e ( 0.24 V 0.14 ) f ( V ) gt ( ) 1 e h( d, V ) A comparison can be made for the temperature influence and the speed influence on excess emission. An example of influence of temperature decrease by 30 C (20 C to -10 C) and average speed increase by 40 km/h (10 km/h to 50 km/h) on the f and g functions is shown in Table 13. The joint influence of speed and temperature is ranging from 0.63 to 8.61, with an average of Thus the influence of temperature and speed is very significant, temperature influence being higher than speed influence. Technology Pollu -tant f f f(50 km/h) f(10 km/h) g g g(-10 C) g(20 C) f(50)+ g(-10)-1 f(10)+g(20) -1 (10 km/h) (50 km/h) (-10 C) (20 C) CO Gasoline CO cars HC without NOx catalyst FC CO Diesel CO cars HC without NOx catalyst FC CO Gasoline CO cars HC with NOx catalyst FC Table 13: Influence of temperature and speed on the speed and temperature functions f(v) and g(t). INRETS report LTE

28 Calculation method Diesel cars with catalyst Concerning excess emissions calculation for diesel cars with oxidation catalyst, we obtained only data for one cycle and one temperature. Then, we assumed that excess emission was proportional to excess emission of diesel cars without catalyst : excess emission for diesel cars with catalyst excess emission for diesel cars without catalyst = This coefficient has been calculated for FTP72-1 cycle, at 20 C considering the 9 tests for diesel cars with catalyst tested and the 66 tests for diesel cars without catalyst. Then, ω(diesel with catalyst)=α ω(diesel without catalyst), the functions f, g and h remaining the same. The corresponding ω coefficients are presented in table Light duty vehicles (LDV) For light duty vehicles, we think that excess emissions should be the same as for passenger cars (PC). This hypothesis has to be made due to the lack of data Heavy duty vehicles (HDV) In the same way, for HDV the excess emission could be calculated using: HDV excess emission PC excess emission (same fuel, no cat.) = HDV hot emission (at 20 km/h) PC hot emission (at 20 km/h) The second term of this equation has to be determined using other MEET results. PC excess emissions are given above Excess emissions for trip starting under engine mean temperature conditions From statistical data relating to the Drive-modem and Hyzem driving patterns (André et al., 1999), 41.9 % of the trips are started with a fully warmed-up engine (water temperature exceeding 70 C) for passenger cars, while about 19 % only of the trips are actually started with a cold engine, i.e. with an engine temperature equalling the ambient temperature. Similar results (20 %) were recorded during a measurement campaign conducted in Austria at a small scale (GVF, 1992). About two fifths of the trips are started with an intermediate engine temperature lying between ambient temperature and fully-warmed up engine temperature (70 C). For the studied cases two approaches are currently available to model excess emissions: - from parking time data, - from start-up temperature statistical data 26 INRETS report LTE 9931

29 Modelling of cold start emissions for passenger cars Parking time approach The dependence of the parking time upon start emissions has been measured over 5 cars equipped with catalysts, 3 gasoline non-catalyst cars and 4 diesel cars. The cars have been tested after a 0.5 h, 1 h, 2 h, 4 h, 8 h, 10 h and 16 h parking period at 20 C. This yielded the following correction factor for parking time, written in the form: excess emissions afteraparking time of x hours excess emissions afteraparking time of16 hours The results obtained are given in Figure 6 and the recorded values in Annex 8. For HC emissions from non-catalyst gasoline and diesel vehicles corresponding to a parking time of 10, 11 and 12 hours, the results obtained with the correction coefficient are not in good agreement with the measurement results: for these 3 time parameters, values were calculated assuming a linear progression for a parking time ranging from 4 to 9 hours. 1,2 1 Parking-time-correction-coefficient 0,8 0,6 0,4 0,2 0-0,2-0,4-0, FC (all cars) CO-catalyst CO-gasoline CO-diesel HC-catalyst HC-gasoline HC-diesel NOx-catalyst NOx-gasoline NOx-diesel PM-diesel -0,8 Parking time (h) Figure 6: Parking-time-correction-factor (GVF, 1992), in the form (excess emissions after a parking time of X hours)/(excess emissions after a parking time of 16 hours). This approach is very interesting but relies on a single parameter of ambient temperature and average speed. Statistical data relative to parking time are needed for implementing such an approach. An example from the Austrian study is given below (tab. 14) and wider statistical data should be defined from the modem/hyzem database. % of starts over a given period Vs parking-time before starting (h) Period 0-1 h 1-2 h 2-3 h 3-4 h 4-5 h 5-6 h 6-7 h 7-8 h > 8 h Total 0 to 4 h to 8 h to 24 h Average Table 14: Start Distribution versus parking-time and period of the day (GVF, 1992). INRETS report LTE