Modelling andsimulation of anelectrichybrid Bus in City Traffic

Transcription

1 Modelling andsimulation of anelectrichybrid Bus in City Traffic Tekn.Lic. J. Andersson, Tekn.Dr. B. Jacobson, Ing. R. Axelsson, Ing. Lars Lundmark (posthumous) Machine & Vehicle Design, Chalmers University of Technology, Göteborg, Sweden, Summary Many new vehicle propulsion concepts aim for lowering fuel consumption and emissions. Simulation is an efficient tool for design of such systems. An electric series hybrid city bus, driven on a specific route, is used as a case study for involving simulation in the design process. The first objective of this work is to verify models for transport task and driver. The second objective is to optimize the energy storage size with respect to fuel consumption, emissions and performance. The third objective is to evaluate the optimized hybrid bus. Comparisons are made with a conventional bus. 2. Introduction Alternative vehicle propulsion systems occur in many different concepts, from using unconventional fuels in rather conventional internal combustion engines to completely new concepts like energy storing or fuel cell driven ones. In common for most of the new solutions is that they aim for lowering the emissions and/or the energy consumption. Two main methods can be used for evaluation of new propulsion systems: Testing prototypes or computer simulations. In the early development of a new propulsion system computer simulation are often preferable due to the large cost for building prototypes. Many research and development teams, involved in alternative propulsion systems, have adopted modeling and simulation as tool, e.g. /7/ and /8/. Reference /7/ treats city buses and points out the importance of questioning the traditional concept of driving cycles, i.e., prescribing speed vs time. This paper focuses on energy storing systems, often referred to as hybrid propulsion systems, but has a methodology approach applicable to most alternative propulsion systems for transient traffic. The system used for a case study is described in Chapter 4, e.g. Figure 7. The overall aims for the hybrid project, from which this paper originates can be discussed in the context of Figure. This figure describes one view of the design process for a vehicle propulsion system.

2 model transport task and driver propose driveline (components & configuration of the system) model driveline propose control strategy implement control strategy in model simulation optimum strategy? no yes optimum system? no yes tests verification OK? no yes change driveline changecontrolstrategy changemodel Figure. One view of the design process for a vehicle propulsion system. The task for us, when starting the process from the left in the left of Figure, is: Design a propulsion system for a certain transport task and driver. The result when leaving the process in the right is a proposed system design, including its control strategy. In this work the process sketch serves a frame for presenting the three Work Packages. These are presented in the following and they all refer to a case where the vehicle is a hybrid bus for scheduled city traffic. Work Package A: Model transport task and driver. This work package is connected to model transport task and driver and the loop involving verification in Figure. A conventional bus was modelled along with a transport task and a driver. Simulation results are compared with measurements. More on this in Chapter 3. Work Package B: Optimize energy buffer size for hybrid vehicle. This work package is connected to model driveline and the loop involving optimum system in Figure. A hybrid bus was modelled along with a transport task (a real city bus route) and a driver. The same concept for modeling transport task and driver was used as in Work Package A. Hereby, the accuracy of these models are judged in Work Package A. The hybrid driveline model is not claimed to be verified, since Work Package A did use a conventional bus. A series of simulations was made with different sizes of the energy storage. More on this in Chapter 4. Work Package C: Evaluate hybrid vehicle. This work package is also connected to the loop involving optimum system in Figure, but in a more evaluating than optimizing sense. The best hybrid bus from Work Package B was compared to a conventional bus performing the same transport task with the same driver. More on this in Chapter 5.

3 3. Work Package A: Model transport task and driver The model is hierarchical according to Figure 2, where the two highest hierarchical levels are shown. Model top level Driver Environment Environment Vehicle Attitude Vehicle Operate litre diesel engine automatic gearbox incl. converter with lock-up and final gear chassis Figure 2. Overview of model as visible on computer screen. The model is developed from the model library VehProp, /5/, which is developed in the software Dymola, //. The continuous state variables in the model are: In diesel engine: crankshaft speed, turbo shaft speed, captured air mass in intercooler In chassis: position, speed In driver: pedal position The discrete state variables are typically the speed limit in the road model or the mode of the driver. The environment model is a developed version of the one described in /2/. It models the transport task in two ways: The Road part and the Stop part, see Figure 3. The used transport task on the test ground is also shown in Figure 3. During the tests, the driver was informed to target the speed limits but otherwise drive as usual. City traffic and conceptually new propulsion systems stress the usage of a transport task model instead of a driving cycles, where speed is prescribed versus time.

4 Environment Position Road Stop =Stop with stop time Road gradient Speed limit for present road segment 5 km/h SpeedLimit Speed limit for next road segment 35 km/h Distance to next road segment 25 km/h Distance to next stop Time to start from next stop s s s s Stand time at next stop m 3m 6m 9m Distance Figure 3. Left: Principal variable flow of the environment model. Right: Environment model for the route at the test ground, Hällered. Road gradient is also included, but in this case it is almost identical to zero and therefore not plotted. The transport task model described requires a special driver model, which can determine accelerator and brake pedal signals from knowing the output from the environment model and vehicle position, speed and acceleration. The driver model is divided into a attitude or strategic part and an operate part. For more description and discussion on the environment and driver models, see references /2/ and /6/. The diesel engine model is described in /3/. Additionally, a steady state torque converter model and an ideal automatic transmission model is used. The gearbox has 4 gears plus lock-up of the converter. The chassis is modelled simply as a translating mass with air, rolling and road gradient resistance. Measured and simulated result are shown in Figure 4 and Figure 5. Speed [m/s] 5 speed limit measured speed simulated speed Accelerator pedal Position [m] Figure 4. Speed and accelerator position for a measurement (dotted lines) along with results from one simulation (solid lines), concerning the test route.

5 4 Torque [Nm] Speed [rad/s] 4 Torque [Nm] Speed [rad/s] Figure 5. Measurements (right plot) of the torque and rotational speed for the propeller shaft along with one simulated result, concerning the test route. A strict numerical comparison of measurements and simulations is difficult to do. However, it is obvious that measurements and simulations have some resemblance, since they have the same characteristics. The mismatch between measured and simulated pedal position is due to the calibration of measured signal for values between zero and one but the values are not of major importance. The fact that the pattern is captured is more important. It can be seen that the simulation shows higher acceleration and lower deceleration, but this can easily be tuned by adjusting the two corresponding driver parameters. It turns out that two measurements often spread more than a single measurement differ from the corresponding simulation. Therefore, it is probably no idea to look for a very strict way to compare measurement with simulations. Before one can claim that the models are accurate a similar comparison has to be made concerning driving in real traffic with a hybrid vehicle. However, it is already claimed that the models probably are better than just using conventional driving cycles, i.e., speed versus time. With driving cycles arising from measurement of conventional vehicles, a hybrid vehicle would be driven with the same speed vs time trajectory as the conventional one, which is not very realistic. See also Figure 3 and reference /7/. 4. Work Package B: Optimize energy buffer size for hybrid vehicle The transport task is modelled in the same spirit as in Work Package A. Since it now corresponds to real bus traffic and not a route on the test ground, some stops are real bus stops why they also need departure times and number of passengers, which are new items for the driver model to reflect on. The transport task model used is presented in Figure 6. The letters

6 A-F refers to critical places along the route. They show how the transport model have a high degree of correspondence with the actual route, represented by the map. The values of speed limit is found by looking at measurements of buses in traffic at the route. Additionally, the road gradient is described as function of position. The transport task used involves about 6 stops distributed over a distance of approximately 5 km. The route is mostly downhill in its first half and then mostly uphill. The road gradient has been slightly reduced in order to make also the simulations with the smaller energy buffers to handle the transport task, see also Figure. B A Bus stop Curve Pedestrian crossing Curve Street crossing Bus stop C F A B C D E F D E speed / [m/s] solid line : speed limit (defined in road model) dotted lines: speed trajectories (output from simulations Figure 6. The route in city traffic, part of bus route No 85 in Göteborg. The vehicle model is shown in Figure 7. The propulsion system is similar to the one described in /4/ and two prototypes are built and used in scheduled traffic in Göteborg, but no reliable measured data is yet available. It is a series electric hybrid primarily driven by a gas turbine, see Figure 8. The energy is stored in conventional lead-acid batteries, see Figure 9. The difference, expressed by the continuous state variables in the model, compared to the model used in Work Package A is: In gas turbine (replaces the diesel engine): shaft speed In energy storage: state of charge (SOC), temperature In control system: 2 integrators in control of traction force and SOC

7 controlsystem gas turbine high speed generator dc/dc converter energy storage motor final gear chassis Figure 7. Vehicle model for the hybrid bus model as visible on computer screen. torque / [Nm] throttle / [-] 5 5 Gas turbine torque at fully open throttle speed / [rad/s] Gas turbine efficiency (%, 2%, 28%) % 2% 28% speed / [rad/s] Torque / [Nm] Voltage=4 V High Speed Generator Voltage=65 V Speed / [rad/s] Figure 8. Gas turbine and high speed generator characteristics. The dynamics of the gas turbine is captured in a simple flywheel model: MassInertia*d(Speed)/d(time)=ProducedTorque-ShaftTorque Produced torque, efficiency and emissions are read from empirically derived maps. For each cell in the battery: Uc=EMK-Ri*Ic, where EMK and Ri are dependent on SOC (State of Charge) and temperature. Energy storage interface variables, U and I, are found through: U=Ucell*Ncell*SBatt and I=Icell*PBatt. The time derivative of SOC is modelled as proportional to I. The time derivative of temperature is modelled as a function of energy loss and cooling.

8 - one cell: voltage=ucell EMK + Ri current=icell one battery: Ncell cells connected as: - voltage=ub 2 3 Ncell... + current=ib one energy storage: PBatt*SBatt batteries connected as: 2 - voltage=u SBatt current=i... PBatt Figure 9. Energy storage model. The reference case is Ncell=6, SBatt=27 and PBatt=2. Pbatt is then used as a parameter for optimization of energy storage size. Simulations with different size of battery is performed and presented in Figure and Table. From Table it is found that storage size 2 is the best. It corresponds to 4 MJ, but the utilization,.52<soc<.55, covers only 3 percent of it, i.e. 4.2 MJ. The storage is used as a booster that allows usage of a low emission prime mover (the gas turbine) instead of a diesel engine. speed / [m/s] and road gradient / [scaled for nice plot] State of Charge / [dimensionless] Figure. Simulation result for different energy storage sizes. Size is varied with the number of batteries in parallel. The number tested are 3, 2, and.5. Upper diagram: Speed trajectory for the size. Lower diagram: Variation in SOC, where a slower variation refers to a larger storage.

9 fuel per distance / [mass/distance] Storage size: NOx per distance / [mass/distance] Figure. Simulation result for battery size optimization. Fuel consumption and emissions (NOx). Curves present the continuous quantity accumulated mass divided by travelled distance. energy storage size Table. Results of simulation with different energy storage size. fuel ranking ( is best) emissions (NOx) performance (keeping the time table) sum of ranking (low is best) The actual conclusion that this storage is best should not be read as an absolute truth, primarily since the properties (fuel, emissions and performance) can be weighted differently but also since the control strategy is not optimized. However, the study calls for attention to a important

10 fact, which was implicitly stated already in Figure : A fair comparison can not be made with less than that the control strategies are separately optimized. In order to make the comparison fairly relevant, the energy storage size is used as a parameter in the control system. E.g., the PI-controller for keeping SOC limited has coefficients depending on the size of the energy storage. Also, the current in the electric motor and in the DC/DC converter are limited by a model based non-linear controller, where the size occurs as a parameter. 5. Work Package C: Evaluate hybrid vehicle Figure 2 shows a simulation with a conventional bus, used Work Package A. The same transport task and driver models as in Work Package B are used. Figure 3 shows a close-up view where the conventional bus can be compared to the hybrid bus. It shows, as claimed earlier, that different vehicles perform their transport task differently, although they are driven by the same driver performing the same transport task. curve above zero: speed / [m/s] curve below zero: gear (-,-2,-2.5,-3 means st,2nd,2nd with lock-up,3rd gear) Figure 2. Simulation with conventional bus driven on bus route No 85 in Göteborg. Compared to the optimized hybrid bus from Work Package B, it is found that the conventional bus uses less fuel but produces more emissions per distance. Depending on weight factors on properties, either one of the buses can be found as a winner. The exact values of the properties are intentionally not given here, since we presently are more concerned of showing the possibilities in using modelling and simulation than presenting a basis for judging the hybrid propulsion technology.

11 7 6 conventional bus speed / [m/s] hybrid bus Figure 3. Simulation with a conventional bus and a hybrid bus with two different energy storage sizes. 6. CONCLUSIONS AND FUTURE WORK The overall conclusion with this work is that modelling and simulation is a very useful tool in the design of alternative propulsion systems. We suggest, however, that the concept of driving cycles, i.e. prescribing speed vs time, steps aside in favour of more realistic transport task and driver models, especially for design of city buses. It is important, but not easy, to verify such models. Some appropriate tasks suggested, in order to really evaluate the propulsion system concept are: Verify the transport task and driver models also through measurements on the real bus route, and preferably using both conventional and hybrid buses. Develop/optimize the control strategy, in order to have a more fair comparisons. A longer route, typically some tenths of km, is probably better, since it would reduce the influence of, e.g., different initial and final SOC in the energy storage. Extend the simulations to cover the influence of traffic interaction. Include accessories in driveline, such as pneumatic compressors and air conditioning.

12 7. ACKNOWLEDGEMENT This work is a part of research project carried out by Volvo Bus Cooperation and Machine & Vehicle Design at Chalmers University of Technology. The project is sponsored by Swedish Automotive Research Programme. Valuable contributions came from Benny Carlsson at Volvo Bus. Also Anders Malmquist at Royal Institute of Technology and the people at Volvo Technical Development Cooperation deserves many thanks. One of us, Lars Lundmark, modelled most of the hybrid driveline components. Even if Lars is not longer with us, after a tragic accident, we would like to express our sincere thanks to him for his excellent work and especially for being a great colleague and friend. 8. References // Elmquist, H., Dymola User s Guide. Dynasim AB, Lund, Sweden, 997. (See also: information on the modeling and simulation software Dymola). /2/ Eriksson, Anders. Simulation Based Methods and Tools for Comparison of Powertrain Concepts, Thesis for degree of Licentiate of Engineering, Machine and Vehicle Design, Chalmers University of Technology, S GÖTEBORG, Sweden, /3/ Berglund, S. and Karlsson, J., A Modular Diesel Engine Toolbox for Studies of Charging and Control System Influence on Emissions and Performance, 3th International Symposium on Automotive Technology & Automation, Firenze, Italy, June 6-9, 997. /4/ ECB Project Group, Built for the next century - the Volvo ECB. Volvo Technology Report No. 2, pp , 996 /5/ Jacobson, Bengt. Final report from the project: Modular Simulation Tool for Vehicle Propulsion concerning Energy Consumption and Emissions, Machine and Vehicle Design, Chalmers University of Technology, S GÖTEBORG, Sweden, 997. Web version: /6/ Andersson, J., Axelsson, R. and Jacobson, B.. Control Strategies for an Electric Hybrid Bus Adapted for a Specific Route, 4th International Symposium on Advanced Vehicle Control 998 (AVEC'98), Nagoya, Japan, September 4-8, 998. /7/ Berta, G. L., Durelli, E., Nymann, I. Simulation models for hybrid buses, Journal of Automotive Engineering, 998 Vol 22, No D, pp /8/ Powell, B. K., Bailey, K. E., Cikanek, S. R. Dynamic Modeling and Control of Hybrid Electric Vehicle Powertrain Systems, IEEE Control Systems, October 998, pp 7-33