Simultaneous optimisation of fuel consumption and ...

2 downloads 0 Views 548KB Size Report
in Simulink and incorporated into the vehicle simulation software Advisor. .... transmission and synchronously rotating with the input shaft, the stator being fixed.
204

Int. J. Vehicle Design, Vol. 46, No. 2, 2008

Simultaneous optimisation of fuel consumption and emissions for a parallel hybrid electric SUV using fuzzy logic control Jian Chen and Yan Li* School of Manufacturing Science and Engineering, Sichuan University, Chengdu, 610065 PR China E-mail: [email protected] E-mail: [email protected] *Corresponding author

Jian Wang School of Mechanical and Aerospace Engineering, Queen’s University Belfast, Stranmillis Road, Belfast BT9 5AH, Northern Ireland, UK E-mail: [email protected] Abstract: A Fuzzy Logic Control Strategy (FLCS) for a parallel hybrid electric SUV is described. In this control strategy, an optimal torque provided by the Internal Combustion Engine (ICE) is first evaluated based on the ICE characteristics. Then, with the consideration of control-state constraints and the characteristics of other power components, the acquired optimal torque is modified by a fuzzy logic controller. Consequently, the actual required output torques of the ICE are figured out. The proposed control strategy was modelled in Simulink and incorporated into the vehicle simulation software Advisor. The simulation results demonstrate that the proposed control strategy is more effective overall than the conventional. Keywords: hybrid electric vehicle; HEV; fuzzy logic control; mathematical model; simulation. Reference to this paper should be made as follows: Chen, J., Li, Y. and Wang, J. (2008) ‘Simultaneous optimisation of fuel consumption and emissions for a parallel hybrid electric SUV using fuzzy logic control’, Int. J. Vehicle Design, Vol. 46, No. 2, pp.204–218. Biographical notes: Jian Chen received his BS from the School of Manufacturing Science and Engineering, Sichuan University, Chengdu, China in 2003. Presently, he is a Postgraduate student in the same school. His research interests include hybrid electric vehicles and product creative design. Yan Li received his PhD in Mechanical Engineering from the Liverpool John Moores University, Liverpool, UK, in 1996. From 1997 to 1998, he was a Postdoctoral Research Associate at Intelligent Systems Research Laboratory, School of Engineering, University of Wales, Cardiff, UK. From 1998 to 1999, he worked as a researcher at the Engineering Design Centre, Engineering Department, Cambridge University, UK. Since 1999, he has been working as a Professor in the School of Manufacturing Science and Engineering Copyright © 2008 Inderscience Enterprises Ltd.

Simultaneous optimisation of fuel consumption and emissions

205

of Sichuan University, China. His research interests include hybrid electric vehicles, intelligent manufacturing and product creative design. Jian Wang is a Lecturer in School of Mechanical and Aerospace Engineering at Queen’s University, Belfast. He is an active researcher in the Research Cluster on Integrated Aircraft Technologies. His research and teaching address dynamic response analysis for complex structures, aircraft design and manufacturing. He is one of the key investigators for several industry-funded projects. He received his BEng from the National University of Defence Technology (China) in Solid Mechanics; his MSc from the University of Science and Technology of China, and his PhD from the University of Wales, Cardiff, UK. He is a member of the American Institute of Aeronautics and Astronautics.

1

Introduction

The automobile industry will be confronted with tremendous challenges in the 21st century owing to environmental issues and declining petroleum resources. According to International Energy Agency statistics (http://www.iea.org), road transport guzzles about 80% of the transportation oil used and directly contributes to over a fifth of greenhouse gas emissions. Although the purely Electric Vehicle (EV) is a promising technology for the long-term goal of energy efficiency and for reducing atmospheric pollution, its limited driving distance, high consumption and lack of supporting infrastructure may hinder its public acceptance (Butler et al., 1999). A Hybrid Electric Vehicle (HEV), combining a conventional powertrain system with an EV powertrain system, can also satisfy the requirements of energy efficiency and low emissions without sacrificing the conventional vehicle’s merits of long driving range and high drivability. Thus, for the short-to-mid-term, HEV has the most potential as an alternative to conventional vehicles. Regardless of topology of the vehicle, the essence of the energy control problem for HEV is how to distribute the driver’s demand to each power source and manage the power flows, while achieving the overall design objectives of a vehicle (Pisu et al., 2003). Many control strategies for HEV have been presented in the literature. These include: conventional Electric Assist Control Strategy (EACS) (Johnson et al., 2000), optimal control strategy (Delprat et al., 2004; Sciarretta et al., 2004) and FLCS etc., among which the application of fuzzy logic control in HEV energy management is now becoming a popular subject of HEV research worldwide. Baumann et al. (2000) and Rajagopalan et al. (2003) presented a rule-based fuzzy logic control method which optimises the energy efficiency through the control of the power flow of a parallel HEV by commanding the engine to operate within its efficient operating region. Kheir et al. (2004) and Schouten et al. (2002) developed a power controller that optimises the operations of major HEV components. Won (2003) presented the vehicle incorporating mode-based fuzzy torque distribution control for a parallel HEV. According to the Yema HEV’s design requirements set by the Sichuan Automobile Industry Group, this paper proposes a FLCS which can get a trade-off between fuel economy and emissions objectives by dynamically choosing a set of weights on fuel consumption and emissions. Using the State of Charge (SOC), the driver’s requested

206

J. Chen, Y. Li and J. Wang

driving torque and the Electric Motor (EM) speed as fuzzy logic controller inputs, a set of 108 rules was developed to effectively distribute the driver’s demand both on the ICE and on the EM. After the proposed FLCS was modelled in Simulink and incorporated into the vehicle simulation software Advisor, the simulation results finally demonstrated that this strategy is more effective overall than the conventional EACS. Therefore, an environmentally friendly and efficient FLCS is presented.

2

System configuration

Typically, HEVs fall into three basic categories (Schouten et al., 2002): Series Hybrid Electric Vehicle (SHEV), Parallel Hybrid Electric Vehicle (PHEV), and Series-Parallel Hybrid Electric Vehicle (PSHEV). The Yema HEV adopts the parallel mode, and the baseline of this PHEV is the Yema SUV, developed by Sichuan Automobile Industry Group in 2004. Figure 1 shows the configuration of the Yema HEV system, which consists of the following components: •

ICE. MPI gasoline engine, 2.0L, 4-stroke, 4-cylinder, 66 kW/5000 rpm, 152 N.m/3500 rpm.



EM. Brushless permanent magnet motor, including two main parts: ring-shaped stator and ring-shaped rotor, the rotor being installed on the input shaft of the transmission and synchronously rotating with the input shaft, the stator being fixed on the input transverse plane of the transmission case, with 30 kW as the rotated power for continuous operation



Transmission. Automated Manual Transmission (AMT), 5 speed.



Battery pack. Nominal capacity and voltage for single cells are 8 Ah and 1.2 V respectively. The battery pack consists of 12 models, each having ten cells connected in series. These 120 cells are all connected in series for a total voltage of 114 V.



Total vehicle curb mass. 1800 kg.

There are five different operating modes for the Yema HEV depending on the flow of power. •

the purely EV mode: all thrust is provided by the EM and all power comes from the battery pack



the ICE only mode: when the driving power at the vehicle wheels reaches a level where operation of the ICE is of high efficiency, the EM turns off



the hybrid mode: both the ICE and EM operate simultaneously



the ICE drive and charge mode: the ICE provides thrust for the wheels along with electricity for charging the battery pack



the regeneration charge mode: as the vehicle is in the braking state, the EM operates as a generator so that the regenerative kinetic energy flows back to recharge the battery pack.

Simultaneous optimisation of fuel consumption and emissions

207

Modes 1–3 can be classified as the general driving modes, which occupy the most operational time of HEV. Figure 1

3

System configuration

Energy control strategy

The energy control strategy for HEV is more complicated than for conventional ICE-only vehicles (Lin et al., 2003), especially for the driving power management/distribution. When the PHEV operates, the engine output power (torque) and motor output power (torque) need to be selected to achieve optimal fuel economy, emissions reduction and drivability. If the power distribution is figured out, the driving mode of the HEV operating at has been decided too. Besides, there are two control-state constraints that must be monitored during the entire operation period: x

The driver’s request for driving power has to be satisfied consistently and the driver should not feel different from driving a conventional ICE-only vehicle.

x

The battery SOC has to be maintained balance and not drop below a certain level to avoid damage of the battery pack. Consequently, the battery pack should probably be charged or discharged continually during entire HEV operating period.

3.1 Electric Assist Control Strategy (EACS) Toyota Prius, the most successful commercially available HEV in the world, has adopted EACS. The EACS uses the ICE as the primary power source, including maintenance of SOC and the EM for additional power whenever needed by the vehicle. Figure 2(a) (Johnson et al., 2000) shows that when the SOC is higher than its low limit, the EM can be used for all driving torque below a certain minimum vehicle speed or a certain minimum required torque. The ICE is turned off and the vehicle is in purely EV mode. Figure 2(b) (Johnson et al., 2000) shows that when the SOC is below its limit, additional torque is required from the engine to charge the battery. It happens in two circumstances:

208

J. Chen, Y. Li and J. Wang



if the driver’s request driving torque is higher than the off torque envelope (case 1), the actual output torque of the ICE is equal to the sum of the driver’s request driving torque and charge torque



if the driver’s request driving torque is lower than off torque envelope (case 2), the ICE runs on the off torque envelope with the excess torque used to charge the battery.

Figure 2

Electric Assist Control Strategy: (a) for high SOC and (b) for low SOC

(a)

(b)

The energy control system for HEV is a non-liner, Multiple-Input Multiple-Output (MIMO) and sophisticated system. The Electric Assist Control belongs to the type of fixed value control, so the ability of this control strategy to optimise vehicle performance is limited. Furthermore, the EACS only attempts to minimise engine energy usage and there is no direct way to effect reductions in emissions (Johnson et al., 2000).

Simultaneous optimisation of fuel consumption and emissions

209

3.2 Fuzzy Logic Control Strategy (FLCS) Fuzzy logic control has been applied as an effective control method in various fields. The advantages of this strategy are its inherent robustness and ability to handle both non-linearities and linguistic knowledge. It also has immunity to imprecise measurements and to component variability. The robust property of FLCS enables the HEV to be operated with the improved battery charge balance, regardless of various disturbances. Therefore, FLCS is a suitable method for PHEV energy control characterised by its non-linearities and uncertainties. The FLCS proposed for the Yema HEV is an instantaneous control strategy. It is primarily used to control the ICE torque output, while the EM is used to perform load-levelling. Thus, in this strategy, the optimal ICE torque is firstly calculated based on the ICE characteristics. Then, with the consideration of control-state constraints and other power components’ characteristics, the optimal torque is modified by a fuzzy logic controller so that the actual output torque of the ICE is obtained (see Figure 3). It finally reaches the target of simultaneously optimisation of both fuel consumption and emissions. Additionally, when the ICE is off at a certain control period, if the AMT’s input shaft speed is bigger than the ICE’s minimum speed, the ICE current speed here should be substituted by the current AMT’s input shaft speed; else the ICE current speed equals the ICE’s minimum speed. Figure 3

Block diagram of the FLCS

3.2.1 The ICE optimal torque calculation The foundations of the FLCS are mathematical models of the ICE. These models consist of the fuel consumption model, the CO emission model, the HC emission model and the NOx emission model. Figure 4 shows one of the models, the fuel consumption 3D map. When any sample speed is specified and its representative vertical plane cuts the fuel consumption 3D map, an intersection line is obtained. There is a point with optimum fuel economy on this line and this point corresponds to a certain torque value. For a series of sample speed planes, there exists a series of optimal points. When these optimal points are connected and displayed in a speed-torque plane, they create a curve called “optimal fuel economy line”. In the same way, the three optimal lines for the other three emission models can be obtained. Figure 5 shows these four lines displayed in the same speed-torque plane with a line representing maximum torque.

210

J. Chen, Y. Li and J. Wang

Figure 4

Fuel consumption 3D map

Figure 5

Optimal ICE lines for fuel consumption and emissions individually

As presented in Figure 5, the locus of optimum fuel economy does not coincide with the loci of optimum emissions at the same speed. In order to solve the problem of mutual contradiction between lower fuel usage and lower emission, a method of dynamically assigning weights for the optimising targets is introduced here; that is to say, in these 3D maps, for a specific current engine speed, continuously choosing a series of points in the torque direction, such that each point corresponds to a set of definite values of fuel consumption and emissions. However, these values cannot be directly compared

Simultaneous optimisation of fuel consumption and emissions

211

with one another because of their different dimensions. In order to weigh the interrelationship of the four optimising objectives with a uniform standard, it is necessary to normalise these values here by utilising the optimal values of fuel consumption and emissions at that current speed. After multiplying the normalised values by their individual weightings (w1, w2, w3, w4), a syntheses criterion value F for that point’s fuel consumption and emissions is acquired (see equation (1)). Yet, for a series of points, the torque with minimal syntheses criterion value is just the optimal ICE torque at that specific speed. F = w1

NO x fuel CO HC + w2 + w3 + w4 (fuel)opt (CO)opt (NO x )opt (HC)opt

(1)

where ‘fuel’ is the actual value of fuel consumption for the chosen operation point at the specific speed plane; ‘(fuel)opt’ is the optimal value of fuel consumption in the specific speed plane. ‘CO’, ‘NOx’ and ‘HC’ have analogous meanings with ‘fuel’, and ‘(CO) opt’, ‘(NOx) opt’ and ‘(HC)opt’ are defined in the same manner as ‘(fuel)opt’ does; w1, w2, w3 and w4 are the weights and they can be varied during PHEV running. According to the characteristics of the chosen ICE, the sensitive levels of each emission index, induced by the fuel consumption fluctuation, have been roughly evaluated. Based on these results, it concludes that when the engine is at the hot-stabilised state, the weights corresponding to the fuel consumption and three emissions are set to 2, 0.15, 0.2 and 0.15 respectively. However, when the engine starts in a cold state, the tailpipe emissions become very high because the three-way catalysts do not reach their ‘light-off’ temperature to work efficiently. In a typical drive cycle, there are 4–5 times increases in HC and CO emissions compared to the stable operation state. Even the production ratio of NOx is relatively low in the cold cylinder, but the tailpipe emission for NOx is likewise serious, an increase of about 2–3 times, because of the catalysts’ poor efficiency in the cold state. In this case, higher weights should be assigned to emissions, and the weight of fuel consumption should be decreased relatively. Besides, during the course of Yema HEV operation, there may be a short-term stop for the ICE. As the engine restarts, it belongs to a ‘part warm’ state which is between cold-start and warm-start, and the catalytic converters can perform at partial function only. Consequently, for this FLCS, the engine coolant temperature is introduced here to judge the states and dynamically tune the weights (see Table 1). Table 1

Assignment table of weights Fuel

CO

NOx

HC

80

2

0.15

0.2

0.15

Engine coolant temperature (°C)

3.2.2 Fuzzy correction on the optimal torque The optimal ICE torque is calculated based on the ICE characteristics but without consideration of the control-state constraints and some other powertrain components. In order to distribute the driver’s demands into each power source properly, the optimal

212

J. Chen, Y. Li and J. Wang

ICE torque should be modified by a fuzzy logic controller. The fuzzy inputs are battery SOC, the driver’s request driving torque and EM speed. Finally, the actual output of the ICE is acquired. The essence of the fuzzy modification/correction is: •

When the driver’s request driving torque is smaller than the optimal ICE torque at that speed a If the battery SOC is higher than a certain level where it is not allowed to charge the battery pack any more, the actual torque output of the ICE will deviate away from the optimal torque point and primarily satisfy the driver’s demand. b If the battery SOC is lower than the level referred in (a), the EM can generate some electricity and keep the SOC balanced. The generation torque is determined by the SOC and EM speed characteristics. In this case, the ICE is able to operate near or even on optimal torque point.



When the driver’s request for driving torque is bigger than the optimal ICE torque at that speed (such as during an acceleration) c If the battery SOC is lower than a certain level where the EM is not allowed to supply tractive torque for preventing battery damage, the ICE’s actual output torque will inevitably deviate from the optimal torque point. d If the battery SOC is higher than the level referred in (c), the EM can work as an auxiliary tractive source to partly or even wholly make up the torque gap between the driver’s demand and optimal ICE output. The value of the auxiliary torque depends on the SOC and EM speed characteristics.

The fuzzy logic controller used in this research is the Sugeno-Takagi type (Takagi and Sugeno, 1985; Le, 1990). The inputs are membership functions and the output is a regular value. The driver’s request driving torque is first transformed as a torque demand to the ICE, and the variation range of the demand is scaled from 1 to 9. According to the characteristics of the selected ICE, the fuel consumption is more sensitive to the torque change in the lower torque area. The calculated optimal torque, with trade-offs between fuel economy and emissions, is relatively close to the maximum ICE torque. Therefore, in the scaling process, the 1, 6 and 9 stands for the zero torque, optimal ICE torque and maximum ICE torque, respectively. Figure 6(a) shows the membership functions of the driver’s request driving torque, ‘down-L’, ‘down-Mid’, ‘down-H’ and ‘down-v-H’ symbolising down low, down middle, down high and down very high respectively. The SOC fuzzy set uses four trapezoidal membership functions for too-low, low, normal and high (see Figure 6(b)). The EM speed fuzzy set consists of three trapezoidal membership functions for low, normal, and high (see Figure 6(c)). The output of the controller is scaled from 1 to 9, where 1, 6 and 9 denote the same meanings as aforementioned. And the scaled output matches the actual output torque of the ICE.

Simultaneous optimisation of fuel consumption and emissions Figure 6

213

Fuzzy inputs membership functions: (a) driver’s request for driving torque; (b) battery SOC and (c) EM speed (rad/s)

(a)

(b)

(c)

In the fuzzy logic controller, the output is related to the inputs by a list of rules. Based on the components characteristics and design requirements of the Yema HEV, a group of 108 rules was developed here. A subset of rules is given by the following list and the Figure 7 shows the 3D fuzzy rule maps, where Ts.f represents the driver’s request for driving torque and opt_Ts.f is the scaled value corresponding to the actual ICE’s output torque. •

If (Ts.f is down-L) and (SOC is too-low) and (EM is optimal) then (opt_Ts.f is 6)



If (Ts.f is down-v-H) and (SOC is low) and (EM is optimal) then (opt_Ts.f is 7)



If (Ts.f is up-High) and (SOC is normal) and (EM is low) then (opt_Ts.f is 7)



If (Ts.f is optimal) and (SOC is low) and (EM is optimal) then (opt_Ts. f is 8)



If (Ts.f is down-H) and (SOC is high) and (EM is optimal) then (opt_Ts.f is 1).

214

J. Chen, Y. Li and J. Wang

Figure 7

4

3D fuzzy rule maps

Simulation of the Fuzzy Logic Control Strategy (FLCS)

Computer modelling and simulation can reduce the expense and length of the HEV design cycle by testing configuration and energy management strategies before prototype construction begins (Butler et al., 1999). The simulation of the FLCS is implemented in the vehicle simulation software Advisor2002, which is developed in the Simulink environment. The Advisor was first developed by the National Renewable Energy Laboratory (NREL) in 1994 (Wipke et al., 1999; Markel et al., 2002). It was designed as an analysis tool to assist the US Department of Energy (DOE) in developing the new generation of vehicles. Both the backward-facing approach and forward-facing approach are integrated into this software. The former has the merit of fast execution while the latter is able to obtain accurate results and a good prediction of vehicle performance limits. The Advisor consists of several blocks, such as ‘drive cycle’, ‘fuel converter’ and ‘gear box’ etc., and the lines with arrows indicates the flow directions of the simulation data among these blocks. Because this blocking configuration has a capability of extension and modification, it is simple for the user to set up their own HEV simulation models. Based on the foregoing control theory of the Yema HEV, the FLCS simulation block is set up in Simulink software (see Figure 8). This block is mainly made up of four subsystems: optimal torque calculation subsystem, driver’s request driving torque scaling

Simultaneous optimisation of fuel consumption and emissions

215

subsystem, fuzzy logic controller and actual ICE output torque scaling subsystem. In the optimal torque calculation subsystem, the multiport switch can realise the function of weights assignment. The leftmost inputs of the FLCS block get simulation data from the previous clutch block and the rightmost output of the FLCS block flows to the next fuel converter block. After the FLCS block of the Yema HEV is embedded in Advisor2002 and displaces the default energy control block, the simulation can be conducted for the whole vehicle’s performance. Figure 8

5

Simulation block of the FLCS

Simulation results and discussion

Based on the characteristics of the Yema HEV components, the parameters belong to the ‘fuel converter’ block and ‘motor’ block etc., are modified first, then the simulation experiment are carried out for EACS and FLCS separately, with the same vehicle components, configuration and the choice of drive cycle UDDS (see Figure 9). The drive cycle UDDS lasts 1369 s, and covers a distance of 11.99 km. Other simulation parameters of the cycle are listed as follows: •

maximum speed = 91.25 km/h



average speed = 31.51 km/h



maximum acceleration = 1.48 m/s2



maximum deceleration = –1.48 m/s2



average acceleration = 0.5 m/s2



average deceleration = –0.58 m/s2



idle time = 259 s.

Figure 10(a) and (b) illustrate the SOC variation trends for the FLCS and EACS respectively. It can be seen that the SOC drops at the end of the cycle for the FLCS are smaller than the drop for the EACS and the depth of discharge by using the FLSC is

216

J. Chen, Y. Li and J. Wang

controlled to be