Electric Vehicle Simulator for Energy Consumption Studies in Electric. Mobility ... (EVSIM09 Project). .... To simulate energy consumption of EVs, SUMO's vehicle.
2011 IEEE Forum on Integrated and Sustainable Transportation Systems Vienna, Austria, June 29 - July 1, 2011
Electric Vehicle Simulator for Energy Consumption Studies in Electric Mobility Systems Ricardo Maia, Marco Silva, Rui Ara´ujo, and Urbano Nunes into account energy issues and traffic analysis, such as route planning, street connectivity and directions, and placement of recharging stations. Modeling and simulation methods are essential elements in design and operation of transportation systems. Several reasons justify the simulation task. Construction costs can be minimized with prior simulation; analysis may be done with minimum risk; dynamic analysis can be made without need of prototype construction; simulation analysis can be made in the design phase of the system at a fraction of the cost of construction. Transportation systems are the backbone connecting the vital parts of a city / region and thereby the indepth understanding of the transportation system components is essential for the planning, design and operational analysis of the city / region. The EV energy consumption can be reduced by many ways, namely by choosing best routes. Alves et al. [2] uses the ant colony optimization algorithm to improve route choice. In [3], the routing problem is characterized using multi-agent systems. The agent’s actions may use internal state information about the vehicle itself such as vehicle size, top speed or torque. A real-time carpooling system using Djisktra’s algorithm with an objective function combining waiting time and traveling time is proposed in [4]. In [5], Sghaeier et al. report an architecture for data collection and analysis performance from EV. Related works to route improvement apply some kind of computational intelligence, like genetic algorithm and fuzzy logic [6], [7], [8], [9] but none of them address issues related with EV, like performance, range and route optimization, aiming minimum energy consumption. EV employ regenerative breaking technology, which allows the conversion of kinetic energy into electrical energy when the vehicle is slowing down or is driving downhill. With regenerative breaking, the electric drive motor also functions as a generator, supplying energy back to the batteries. To simulate this characteristic, traffic simulators’ environment need to be three-dimensional, e.g., altitude has to be known and must be represented in the environment model. However, known traffic simulators are two-dimensional, i.e., the maps lie in Cartesian x-y projection, and therefore are not appropriate to simulate regenerative breaking. This paper describes components that extend the 2D traffic simulator package SUMO (Simulation of Urban MObility) [10] in a 3D simulation environment for electrical vehicles. The EV model used here follows closely the formulations described in [11]. It was developed to provide a means for conducting studies of electric mobility in urban areas. The model was inserted into the car-following model proposed
Abstract— One of the most important environmental problems in large cities is the vehicular emission. Electric Vehicles (EVs) are a growing alternative for internal combustion engine (ICE) vehicles. Since this kind of vehicle has low autonomy yet, it is important to optimize energy consumption, for instance by planning a suitable infrastructure of battery recharge and/or battery-switch stations. This paper presents an architecture for EV simulation, important to analyze traffic flow, its dynamics and the performance when there are obstructions or intense traffic. There are several tools for traffic simulation, SUMO (Simulation of Urban MObility) is one of them. But none of the existing traffic simulators integrates models of EV that allow, for example, perform simulation studies regarding energy consumption. SUMO is a portable open source simulator with multi-modal traffic feature capabilities that permit the simulation of various types of vehicles. This work is an extension of the SUMO, two-dimensional (2D) vehicular simulation package. To allow the simulation of energy consumption of EV, two extensions were incorporated in SUMO: EV models and modeling of altitude, transforming SUMO into a threedimensional (3D) simulator. The energy model effectiveness and correctness with 3D capabilities has been validated using two driving schedules (Urban Dynamometer Driving Schedule and Highway Fuel Economy Driving Schedule). This new tool will also support the study of better routes choice in 3D environment with EV aiming minimum energy consumption.
I. I NTRODUCTION The increasing number of motor vehicles on the streets and roads is demanding an active transport policy. This growth also results in traffic jams and appearance of a wide diversity of drivers with different driving behaviors, increasing the probability of accidents. These are only two reasons why developed countries have a great interest in Intelligent Transportation Systems (ITS). Nowadays, petrol’s high cost and serious environmental pollution problems deriving from fossil burning fuel led automotive industry to heavily investing in plug-in electrical vehicles (PHEV) as well as fully electric vehicles (EV). In [1], Spongenberg states that the number of electric cars on European roads is going to boost next years due to oil prices, climate change concerns and tough EU environment regulations. The use of EV raises several problems taking This work was partially supported by the Portuguese Foundation for Science and Technology (FCT) under grant PTDC/SEN-TRA099413/2008 (EVSIM09 Project). Rui Ara´ujo and Urbano Nunes are with the DEEC - Department of Electrical and Computer Engineering, University of Coimbra, Portugal. All authors are with the ISR-Institute of Systems and Robotics, University of Coimbra, Portugal. Ricardo Maia and Marco Silva also acknowledge PhD fellowships granted by FCT, respectively SFRH/BD/44644/2008 and SFRH/BD/38998/2007. Marco Silva is also with IPC-ISEC - Polytechnic Institute of Coimbra, Coimbra, Portugal. email: {rmmaia, msilva, rui, urbano}@isr.uc.pt
978-1-4577-0991-3/11/$26.00 ©2011 IEEE
228
appearing in (2) to (6) are physical constants, or related to vehicle’s physical characteristics, and their meaning are stated in Table I. It should be noted that EV have regenerative braking feature. This means that, if Fte is negative, the tractive force will not be applied from the electrical motor to the wheels, but from the wheels to the motor; and the current will flow into the battery, charging it. Frr and Fad are friction forces and they must be as low as possible to minimize energy consumption, which is achieved with a good design reducing µrr , A and Cd . Both Frr and Fad are non-negative; so, only Fhc and Fla , together, are able to make Fte negative. Fhc will be negative when the vehicle is going downhill (α < 0), and Fla will be negative when the vehicle is slowing down (a < 0). The mechanical energy required to move the vehicle is Z Ete = Fte × v dt (7)
TABLE I: Physical constants. constants g ρ µrr A Cd α m v a I G r ηg
meaning acceleration due to gravity air density coefficient of rolling resistance vehicle frontal area drag coefficient angle of slope or hill vehicle mass vehicle velocity vehicle linear acceleration moment of inertia of rotor of the motor gear ratio of the system radius of the tyre gear system efficiency
by Krauß [12], [13], but can be easily adapted for other carfollowing models. The paper is organized as follows. Section II establishes the EV model with focus on the energy consumption component. Section III presents the simulation software and its modifications. Section IV describes the simulation scenario and simulation relevant results. Finally, in Section V, some concluding remarks are drawn.
so that the energy taken from battery to be supplied to the traction motor to provide Ete (7), is E te , driven case, (8) ηm × ηg Emot in = Ete × ηm × ηg , regenerative case. (9)
where ηm and ηg are the motor and the gear system efficiencies, respectively. When the vehicle is being driven, it holds (8); but if the motor is being used to slow the vehicle down, the efficiency works in the opposite sense, supplying energy to the battery, and (9) takes place. Finally, it must be considered all other vehicle’s electrical systems (lights, heating, cooling, indicators, radio, etc.), Eac , which shall be added to the motor energy. Thus, the total energy required from the battery is
II. M ODEL D ESCRIPTION An EV is a complex system including several subsystems, such as: mechanical, electrical, control, magnetic, pneumatic, electrochemical and hydraulic, etc. In this work, most subsystems are abstracted, and only those needed to provide mechanical and electrical traction to vehicle will be characterized. A. Mechanical Traction The force needed to provide mechanical traction to propel the vehicle forward is the tractive effort. This force has to overcome the rolling resistance, aerodynamic drag, hill climbing force, the force to accelerate the vehicle and the force to provide angular acceleration to the traction motor. Thus, the tractive effort can be expressed as [11]: Fte
= Frr + Fad + Fhc + Fla + Fwa
(1)
= µrr mg (rolling resistence force); 1 = ρACd v 2 (aerodynamic drag); 2 = g sin(α) (vehicle’s weight component);
(2)
Ebat
Fad Fhc Fla Fwa
Emot in + Eac
(10)
B. Electrical Traction When in driven case, the power required from the motor to make the vehicle run at a certain speed is supplied from the battery. On the other hand, in regenerative case, the current flows into the battery. The current that flows from/into the battery is expressed by:
where: Frr
=
p 2 − 4R P Voc − Voc in bat , 2R in I= p 2 + 4R P −Voc + Voc in bat , 2R in
(3) (4)
= ma (force required to give linear acceleration); (5) G2 a (force required to give rotational = I ηg r2 acceleration to the traction motor); (6)
driven case,
(11)
regenerative case.
(12)
where Voc is the open circuit voltage from the battery, Rin is its internal resistance, and Pbat is the power produced by the current. As the motor drains current from battery, what is really needed to be known is how battery discharges while
Since frequently the motor’s moment of inertia is not known, in these cases it is reasonable to increase the vehicle’s mass by 5% in (5) and ignore Fwa . The physical quantities 229
Fig. 2: Interaction between SUMO’s traffic module and energy module.
A. Modifications To simulate energy consumption of EVs, SUMO’s vehicle class has been modified to receive attributes and methods to implement the functionalities explained in Section II. Moreover, the energy can be regenerated when the vehicle slows down or goes downhill. Thus, the bi-dimensional road network was changed to receive the z coordinate, related to the elevation of the network nodes. Fig. 1 shows a schematic of the main processes and data files involved in the simulation process using SUMO [14]. A new file, XML altitudes, was added in order to allow the specification of road elevation. This way, the NETCONVERT tool, gets the extra z attributes from XML altitudes file and outputs the road network (XML network) which incorporates the elevation information. EV physical parameters and routes information are provided to the SUMO simulator through XML consumption and XML route files, respectively.
Fig. 1: Simulation process with SUMO.
the EV is moving. The depth of discharge (DoD) is given by DoD
CR Cp
=
(13)
where Cp denotes the Peukert Capacity and CR is the charge removed. If in regenerative case Z CR = I dt (14) otherwise, in driven case CR
=
Z
k
I dt
Algorithm 1 Energy Model Pseudo-code. 1: Receive from CFM values of next velocity vi+1 , last value of velocity vi and acceleration a; 2: Calculate torque T needed to apply those acceleration and velocity; 3: if (T > Tmax ) then 4: Calculate new vi+1 and a values to reach Tmax ; 5: end if 6: Calculate Pmot ; 7: Pbat = Pmot + Pac ; 8: Calculate battery current I to provide Pbat , using (11) or (12); 9: Calculate DoDi+1 , the DoD in the next time step; 10: if (DoDi+1 > DoDlimit ) then 11: Re-calculate I and Pbat values; 12: Re-calculate vi+1 and a values; 13: Re-calculate DoDi+1 ; 14: end if 15: Return vi+1 ;
(15)
where k denotes the Peukert Coefficient. III. S IMULATION PACKAGE The consumption model described in the previous section was implemented as a module of SUMO. Here, we describe a few features of SUMO and the modifications made to incorporate new features. The main features of SUMO traffic simulator are: • • • •
support different vehicle types, capable of handling large road networks, handle bi-dimensional networks, computationally fast.
Its vehicles flow’s model (car-following model [12]) is based on microscopic routes, where each vehicle is treated individually with its own route. The model is also continuous in space and discrete in time. The package is open source, licensed under the GPL (General Public License) and highly portable.
The interaction between SUMO’s car-following module (CFM) and the new EV energy module is illustrated in 230
TABLE II: Battery Pack Parameters.
TABLE III: Range with NiMH Battery.
Manufacturer: Ovonic Energy Products Type: Nyckel Metal Hydride Number of Modules: 26 Weight of Module: 18.3 kg Weight of Pack(s): 481 kg Nominal Module Voltage: 13.2 V Nominal System Voltage: 343 V Nominal Capacity: 77 AH Stored Energy: 26.4 kWh
Drive Schedule UDDS HWFET
CEPA report 230.087km 244.568km
SUMO
variation
230.406km 253.588km
0.14% 3.68%
Depth of Discharge (%) 1 0.9 0.8 0.7 0.6
Fig. 2. In each time step of the SUMO simulation process, the desired vehicle velocity is determined depending on the velocity limit of the road, and the traffic demand represented by the distance to the vehicle ahead and its velocity. For the car-following module to provide the desired velocity, the energy module verifies each time step if there is enough energy in battery to produce the required power. For this propose, the torque is calculated by
0.5 0.4 0.3 0.2 0.1 0 0
10000
15000 time (s)
20000
25000
30000
25000
30000
Battery Voltage (V)
Fte × r (16) G where Fte is given by (1), r is the radius of the tyre, and G is the gear ratio. If (16) is higher than the maximum torque, Tmax , which the EV motor can provide then new velocity and acceleration values are calculated such that Fte produces Tmax . Next, Ebat is calculated by (10) and it is verified if there is enough energy in the battery to supply to the motor. If not, new velocity and acceleration values are calculated based on the information of the existing residual energy in the battery. Algorithm 1 summarizes the Energy Model’s processing and computations. T
5000
=
370 360 350 340 330 320 310 300 290 280 0
IV. S IMULATION S CENARIO
5000
10000
15000 time (s)
20000
Fig. 3: UDDS: DoD evolution (19 cycles) and battery pack voltage.
A. The vehicle The simulation experiments the vehicle model uses the parameters of the EV1 electric vehicle produced by General Motors from 1996 to 1998. This model was chosen since it has been widely used in previous studies. The EV1 parameters values are: A = 1.89 m2 , Cv = 0.19, µrr = 0.005, length = 4.31 m, maximum acceleration = 3.08 m/s2 , maximum deceleration = 1.0 m/s2 and maximum speed of 129 km/h. The considered density of the air was ρ = 1.25 kg/m3 . The motor and the gear system efficiencies, ηm and ηg respectively, were taken from [15]. The internal resistance Rin and open circuit voltage Voc of the battery pack were modeled according to [16], [17] and [18]. The battery pack main parameters ate be shown in Table II.
under 96 km/h. Table III shows the ranges of EV1 reported by CEPA [18] and ranges obtained by the simulations performed in SUMO extended with the EV and elevations models here proposed. These results show a close matching between the developed models implemented in SUMO and the CEPA data. The DoD (13) and the voltage drop during the 19 complete UDDS discharge cycles is plotted in Fig. 3. In Fig. 4, depicts the power supplied by the battery pack for the first UDDS cycle. The maximum drained power and the maximum regenerated power value is ≈ 34 kW and ≈ 21 kW, respectively. These values are within the motor/inverter capabilities. The discharged cycles were applied to 95% DoD to not exceed the cut-off voltage. The 3D circuit shown in Fig. 5 was used in the constant speed tests. The circuit connects three straight-line road segments. The first road segment ac has a positive slope of 3.24%. The second road segment connects points c and b with 0% slope. Finally, a road segment of negative slope of 6.68% links points b and a. After the a → c → b → a
B. The Road Network Two types of tests were performed to verify the range of the simulated vehicle: driving cycles tests [11] and constant speed tests. Two driving cycles specified by the U.S. Environmental Protection Agency [16] were used: the Urban Dynamometer Driving Schedule (UDDS), representing city driving conditions; and the Highway Fuel Economy Driving Schedule (HWFET) representing highway driving conditions 231
10000
Delivered / Regenerated Power (kW) 40
6000
30
b
10 0
.4
8%
%) 3.24 m( 4 9 299
3 99 14
2000
20
m
(6
c 16054 m (0%)
)
a 5000
10000
15000 20000 (a)
-10 -20
25000
30000
State of Charge 1
-30 0
200
400
600 800 time (s)
1000
1200
1400
a
0.9
b
0.7
Fig. 4: First UDDS cycle power request.
a
c
0.8
a
0.6
b
0.5
c a
0.4
TABLE IV: Constant speed range scenario. Speed (slope / no slope) 60km/h 60km/h 80km/h 80km/h
(slope) (no slope) (slope) (no slope)
c
0.3
b a
0.2
Range
b c
0.1
260km 430km 196km 302km
0 0
50000
100000
150000
200000
250000
time (s)
(b) Battery current (A) 60
trajectory the vehicle is driven back to point a but following the opposite way a → b → c → a. This sequence was repeated until the empty charge condition was attained (DoDlimit = 0.95). Figures 5b-5c represent the evolution of the State of Charge (SoC = 1−DoD) in a constant speed profile through the 3D road test network. Though the EV has a constant speed, the current is not constant, as can be seen in Fig. 5c. In fact, as the cycles are repeated, the SoC decreases due to the energy drained from the battery pack. The battery pack voltage also decreases (Fig. 3), which explains the fact that, to achieve the same required power to maintain a constant speed value, an increase is required in the consumed current. This happens even in zero slope road bc segment. In the first pass of the vehicle in the direction a → b the current measured at b point is 57A, and in the corresponding point at the last pass the current is 59A as highlighted in Fig. 5c. Table IV summarizes results for two situations considering the route defined in Fig. 5c: (a) with slopes as specified in the figure, and (b) with all three segments being horizontal. In both cases it was considered that the vehicle carried a passenger with 75kg weight. In zero slope and low speed tests, the two major influence influencing range are the rolling resistance and the aerodynamic drag. The EV1 range at constant speed of 72 km/h reported in [16] is 355 km which is in the range of obtained values in our zero-slope tests (430 km for 60 km/h and 302 km for 89 km/h). The regenerative breaking effect is clearly observed in the Fig. 5b in the road segments with negative slope. Finally, the model was applied on a simulates subnetwork of the Coimbra city. The target area, a circuit of approximately 8 km length, is illustrated in Fig. 6. The
a
b
50
40
30
20 b
c
10
(c)
Fig. 5: (a) 3D road test network, (b) evolution of the SoC along the complete test, and (c) zoomed and overlaid currents at the first (red) and last (blue) passes of the vehicle over the a → b → c segment.
starting/ending point of the circuit is represented by a red dot on the map. The maximum difference of altitude in the covered circuit is 54.4 m. As shown in Fig. 7, the circuit is diversified in terms of altitude allowing exploration of the regenerative breaking characteristic of the vehicle. For this study, one EV was injected at the starting location, and run along the route specified by the arrows. There was no other traffic along the circuit. To prove the effect of altitudes over energy consumption, two different tests were performed: 1) Using original map, with the true altitudes; 2) Using a planified map, with no altitude differences. Both simulation tests consisted of one complete turn along the circuit. The corresponding energy results are shown in 232
way the battery discharges over time (with charging periods occurring in negative slope segments or vehicle deceleration periods). The EV consumption model was validated with two types of well-known driving cycles and in constant speed mode. This model has been also applied on a sub-network area of Coimbra city. With the enhancements reported in the paper, SUMO framework was endowed with suitable tools that allow large scale simulation of electric mobility systems. R EFERENCES [1] H. Spongenberg, “Euobserver / eu states plug in to electric cars,” http: ////euobserver.com/882/26594, August 2008, retrieved 2010-03-08. [2] D. Alves, J. van Ast, Z. Cong, B. D. Schutter, and R. Babuˇska, “Ant colony optimization for traffic dispersion routing,” 13th International IEEE Conference on Intelligent Transportation Systems (ITSC), pp. 683–688, Sep 2010. [3] S. Boskovich, K. Boriboonsomsin, and M. Barth, “A developmental framework towards dynamic incident rerouting using vehicle-tovehicle communication and multi-agent systems,” 13th International IEEE Conference on Intelligent Transportation Systems (ITSC), pp. 789–794, Sep 2010. [4] V. Suresh, G. Hill, P. T. Blythe, and M. Bell, “Smart infrastructure for carbon foot print analysis of electric vehicles,” 13th International IEEE Conference on Intelligent Transportation Systems (ITSC), pp. 949–954, Sep 2010. [5] M. Sghaier, H. Zgaya, S. Hammadi, and C. Tahon, “A distributed dijkstra’s algorithm for the implementation of a real time carpooling service with an optimized aspect on siblings,” 13th International IEEE Conference on Intelligent Transportation Systems (ITSC), pp. 795–800, Sep 2010. [6] Y. Chen, M. G. H. Bell, and K. Bogenberger, “Reliable pretrip multipath planning and dynamic adaptation for a centralized road navigation system,” IEEE Transactions on Intelligent Transportation Systems, vol. 8, Issue: 1, pp. 14–20, 2007. [7] H. T. Masaya Yoshikawa, “Car navigation system based on hybrid genetic algorithm,” World Congress on Computer Science and Information Engineering, 2009. [8] B. Chakraborty and R. C. Chen, “Fuzzy-genetic approach for incorporation of driver’s requirement for route selection in a car navigation system,” IEEE International Conference on Fuzzy Systems, pp. 1645– 1649, 2009. [9] A. J. S. Kumar, J. Arunadevi, and V. Mohan, “Intelligent transport route planning using genetic algorithms in path computation algorithms,” European Journal of Scientific Research, vol. 25 (3), pp. 463– 468, 2009. [10] SUMO, “Simulation of urban mobility,” http://sourceforge.net/ apps/mediawiki/sumo/, January 2010, retrieved 2010-03-18. (www.dlr.de/ts). [11] J. Larminie and J. Lowry, Electric Vehicle Technology Explained. John Wiley & Sons, 2003. [12] S. Krauß, “Microscopic modeling of traffic flow: Investigation of collision free vehicle dynamics,” Ph.D. dissertation, Mathematisches Institut, Universit¨at zu K¨oln, 1998. [13] S. Krauß, P. Wagner, and C. Gawron, “Metastable states in a microscopic model of traffic flow,” Phys. Rev. E, vol. 55, no. 5, pp. 5597–5602, May 1997. [14] SUMO, “Sumo user documentation,” http://sourceforge.net/apps/ Jan mediawiki/sumo/index.php?title=SUMO User Documentation, 2010, retrieved 2010-03-08. [15] A. Campbell, A. Rengan, and J. Steffey, “The simulation of 42-volt hybrid electric vehicles,” http://www.math.msu.edu/Academic Programs/ graduate/msim/MSIMProjectReports/MCP2.May.2001.report.doc, retrieved 2011-05-31. [16] U.S. Departament of Energy, “1999 general motors EV1 w/nimh,” http://www1.eere.energy.gov/vehiclesandfuels/avta/pdfs/ fsev/eva results/ev1 eva.pdf, retrieved 2011-05-31. [17] S. Golbuff, “Optimization of a plug-in hybrid electric vehicle,” Master’s thesis, Georgia Institute of Technology, the Netherlands, 2006. [18] A. R. Board, “2000 zero emission vehicle program - staff report,” http: //www.arb.ca.gov/msprog/zevprog/2000review/staffreportfinal.pdf, retrieved 2011-05-31.
Fig. 6: Road circuit simulated (source by Google). Altitude profile 80
70
circuit end
Altitude (m)
60 circuit beginning 50
40
30
20
0
1000
2000
3000
4000 5000 Position (m)
6000
7000
8000
Fig. 7: Altitude profile along the circuit.
TABLE V: Range in urban scenario, one cycle one passenger (75kg) top speed 60km/h (3D) 60km/h (2D) 80km/h (3D) 80km/h (2D)
DoD 3.27% 2.15% 4.21% 3.12%
Table V. As can be seen, the altitude significantly influences consumption. The energy is drained differently depending on the altitude profile of the circuit: with altitudes the consumption is higher then with the flat map. V. C ONCLUSION A simulation framework for electric vehicles in terms of energy consumption has been presented. SUMO traffic simulator was enhanced with components that allow 3D simulation of EV energy consumption. EV performance depends on the terrain slope, with a direct impact on the 233