Fuel Economy Comparisons of Series, Parallel and HMT Hydraulic Hybrid Architectures

Transcription

1 2013 American Control Conference (ACC) Washington, DC, USA, June 17-19, 2013 Fuel Economy Comparisons of Series, Parallel and HMT Hydraulic Hybrid Architectures Zhekang Du, Kai Loon Cheong, Perry Y. Li and Thomas R. Chase Abstract Since hydraulic hybrid vehicles are power dense and do not require costly batteries, they have the potential to achieve high fuel economy and performance and at low cost. The three main classes of hydraulic hybrid architectures are series, parallel and hydro-mechanical (HMT) or powersplit. These architectures have intrinsic differences in transmission efficiencies and effectiveness in engine management. This paper compares the fuel economies and performance of these architectures and validates these features. Using a Toyota Prius like engine and chassis as common factors, fuel economies are compared for the optimal design for each architecture which considers both the physical designs and the engine/energy management. Physical design variables include pump/motor sizes and gear ratios. The effect of pump/motors efficiencies, extra gears and different engine efficiency maps are also studied. To improve computational efficiency in evaluating fuel economy, engine operation is restricted to several operating modes and the accumulator pressure is assumed to be constant. These simplifications enable the Lagrange multiplier method to be employed so as to quickly determine the optimal engine management control and the resulting fuel economy for each design iteration. Full optimal control computations without the simplifying assumptions for the optimized design for each architecture (using dynamic programming) verify that, despite these simplification, the estimated fuel economies are close. It is shown that hybrid HMT offers the best fuel economy for various hydraulic efficiencies, followed by parallel and series architectures. However, the difference between HMT and parallel architectures diminishes if the engine has a wide efficient speed range of operation. It is also shown that an extra mechanical gear ratio can significantly increase fuel economy for all three architectures. I. INTRODUCTION Hybrid power-trains can play an important role to improve fuel economies of vehicles by allowing better engine management and re-using braking energy. In contrast to a hybrid electric power-train which uses an electric battery to store energy, and electric motor/generators for energy conversion, a hydraulic hybrid power-train stores energy in hydraulic accumulators and uses hydraulic pump/motors for energy conversion. Hydraulic accumulators and pump/motors have an order of magnitude higher power densities than the electric counterparts, can be manufactured cost-effectively and do not pose significant concerns when recycling. While hydraulic accumulators have low energy densities, it can be shown that with effective control strategies, the capacity of energy storage for hydraulic hybrids do not need to be very The authors are with the Center of Compact and Efficient Fluid Power, Department of Mechanical Engineering, University of Minnesota, Minneapolis, MN duxxx139@umn.edu, cheo0013@umn.edu, perry-li@umn.edu, trchase@me.umn.edu Fig. 1. a) Series; b) Parallel; c) HMT (power-split) hybrid architectures high. Thus, hydraulic hybrids have the potential for high fuel economies, high performance, compactness and low cost. The three basic hydraulic hybrid architectures are series, parallel and power-splits (Fig. 1). The series architecture eliminates the mechanical linkage between the engine and vehicle speed. It allows the engine to operate at any operating point and offers freedom to operate the engine more efficiently. However, all engine-generated power must be transmitted through the hydraulic pump/motors which are less efficient than mechanical transmission. The parallel architecture preserves the efficient mechanical gearbox and uses a hydraulic pump/motor to augment engine operating torque and to store and reuse braking energy. Engine speed is however coupled with the vehicle speed so that engine operation restricted. Hydraulic power-split architecture, a.k.a /$ AACC 5974

2 between the fuel economies of the three main hydraulic hybrid architectures. To achieve a fair comparison of the potential of each architecture, the physical design (such as pump/motor sizes, gear ratios) as well as the energy management control are optimized. Given the limited effciencies of hydraulic components, the sensitivity of the overall powertrain efficiency with respect to the variation of hydraulic pump/motors efficiency is also studied. The rest of the paper is organized as follows: Section II describes the different operating modes of the three major hybrid architectures. Section III explains the optimization process of the power-train in order to compare the different hybrid architectures fairly. Section IV present comparison results and discussions. Section V contains concluding remarks of this study. Fig. 2. Overall power-train efficiencies as a function of hydraulic efficiency for the 3 architectures with η engine 0.39 for series and HMT; η engine 0.35 for parallel; % hydraulics 0.6 for HMT and parallel; η mechanical hydro-mechanical transmission (HMT) uses a pair of hydraulic pump/motors and a planetary gear set. It allows a portion of power to be transmitted mechanically and the engine to operate at arbitrary operating points. There are several types of HMTs. The input-coupled HMT shown in Fig. 1 is used in paper. Other HMTs are output coupled and compound HMT [8], [2]. Because of these features, the power-train efficiency for a series, parallel and HMT architecture will follow roughly this formula 1 : Energy output η overall F uelenergy η engine [ %hydraulic η hydraulics + (1 % hydraulic) η mechanical ] 1 where η engine, η hydraulic and η mechanical are the efficiencies of the engine, hydraulic transmission and mechanical transmission; % hydraulic is the proportion of the energy transmitted hydraulically. For series architecture, % hydraulic 1 and for parallel and HMT, % hydraulics depends on control strategy. Since parallel architecture does not allow arbitrary engine operation, η engine is expected to be lower than for series or HMT. With currently available components, η hydraulics << η mechanical. Fig. 2 shows an example comparisons of the overall drive-train efficiencies as a function of the efficiency of the hydraulic components η hydraulic for the 3 architectures. Analysis, design and control of specific hydraulic hybrid architectures, especially power-splits are reported in many articles, including [4], [1], [8], [10], [7]. Comparisons between different hydraulic and electric power-split hybrids were presented in [9]. What is lacking in the literature, and what this paper aims to do, is a direct comparison 1 braking energy recuperation is neglected for simplicity (1) II. HYBRID VEHICLE OPERATING MODES The operating modes of each architecture are presented in this section. Besides the basic operating modes (series, parallel and HMT), additional operating modes are introduced to further improve fuel-economy. These involve by de-clutching and shutting off the engine or locking up / free-spinning individual hydraulic units. In evaluating fuel-economy in the design iteration/optimization process, the engine operating point in each mode is pre-defined based upon engine efficiency alone. This is done in order to improve computational efficiency at the expense of the operation not being truly optimal, since component efficiencies are often not taken into account by this choice. When evaluating the fuel economy of the final optimized design, however, the engine operating point is not restricted a-priori. A. Series Architecture The operating modes for the series architecture are given in Table I. The normal series operating mode is that the engine operates at its more efficient point, and PM2 is controlled to propel the vehicle as required by the drive cycle. In PM2-only mode, PM1 is locked up and the engine is shut off. Energy is provided or captured by the accumulator only. Modes Series TABLE I SERIES ARCHITECTURE OPERATING MODES PM2-only B. Parallel Architecture Operation of modes Series operation, Engine operating at most efficient point Engine disengaged, PM1 unit locked up The operating modes for the series architecture are given in Table II. In the normal parallel mode, the engine and PM1 speeds are governed by the vehicle speed. For a given engine speed, we restrict the engine to operate at its most efficient torque at that speed. Any excess torque is absorbed by PM1. In the second - PM1 only mode, the engine is shut-off and de-clutched and the vehicle is propelled or braked hydraulically only. 5975

3 Modes Parallel PM1-only Fig. 3. Transmission iterative design loop TABLE II PARALLEL ARCHITECTURE OPERATING MODES C. HMT Architecture Operation of modes Parallel operation, Engine operating along maximum torque curve Engine disengaged The operating modes for the HMT architecture are given in Table III. The normal HMT mode involves operating the engine, PM1 and PM2 simultaneously. For a given desired vehicle speed/torque, the choice of engine operating point uniquely determines how PM1 and PM2 should operate. The consequences of the choice are a) the amount of power taken or stored in the accumulator; and b) the instantaneous losses. In [1], this is resolved with the mid-level static control map which determines the optimal engine operating point for a given accumulator power and desired vehicle speed/torque. This map however is expensive to compute. To reduce computational burden, the engine, when on, is restricted to operate at its most efficient point only. Additional modes are obtained by locking up PM2 when engine is on ( Parallel mode), and by locking up PM1 and PM2 respectively when the engine is dis-engaged (PM2-only, PM1-only modes). In [8], more additional modes can be added but these are not used here. Modes HMT PM2-only PM1-only Parallel TABLE III HMT ARCHITECTURE OPERATING MODES Operation of modes Power-split operation, Engine operating at most efficient point Engine disengaged, PM1 unit locked up Engine disengaged, PM2 unit locked up PM2 unit locked up, Engine operating along maximum torque curve When multiple gears are allowed, the total modes are duplicated for each gear ratio. III. OPTIMIZATION OF VEHICLE DESIGN AND ENERGY MANAGEMENT CONTROL To compare the fuel economies of the architectures, the physical design and control for each are optimized. The same vehicle weight, wind and road drag, drive cycle and the engine are used in all the architectures. However, the pump/motor sizes and gear ratios are allowed to change. Table IV shows the vehicle parameters used in this study. The design optimization procedure is similar to that proposed in [8] and is briefly described next. Pump/Motor Engine Vehicle total mass Drag coefficient 0.26 Frontal area 2.33m 2 Rolling resistance coefficient f o scaled from 28cc bent axis pump 1.5L Prius Engine(ADVISOR[12]) 1500kg Rolling resistance coefficient f s Drive cycle Combined UDDS and highway cycle TABLE IV VEHICLE PARAMETERS APPLIED IN THE SIMULATION A. Components sizing Gear ratios: The gear ratios define kinematic relations between speeds and torques of different components: Series: Parallel: HMT ( TP M1 ) ( ) ( ) 1/R 0 Teng T P M2 0 1/G i T veh ( ) ( ) ( ) ωp M1 R 0 ωeng ω P M2 0 G i ω veh ( TP M1 T P M T veh /(G i R) T eng /R, (2) ω P M ω eng R ω veh RG i (3) ) ( ) ( ) R1 2R 1 /G i Teng, T P M2 0 1/R 2 T veh ( ) ( ) ( ) ωp M1 1/R1 0 ωeng ω P M2 2R 2 /G i R 2 ω veh To investigate the benefits of having multiple gear ratios in hydraulic hybrid architectures, more than one option for the gear ratio G i is allowed: i.e. G 1 and G 2 for 2-speeds and G 1 only for 1-speed. The ratios G i as well as the other ratios R, R 1, R 2 are to be optimized. Varying hydraulic pump/motors sizes: The maximum displacements of the hydraulic pump/motors are to be optimized. To do this, a 28cc bent axis pump/motor is used as a reference pump/motor and the torque and flow loss maps as functions of speed, pressure and displacements are available. To extrapolate these maps to pump/motors of other sizes, they are scaled relative to the maximum displacement of the pump/motors. B. Scaling the losses for hydraulic components To study how the overall efficiency of each architecture depends on the efficiency of the hydraulic pump/motors (see Eq.(1)), a loss factor is introduced to scale the mechanical loss and volumetric loss linearly. The loss factor of 1 means the pump/motor is not scaled, while the loss factor of 0 means the pump/motor is ideal. (4) 5976

4 C. Optimization of hybrid operation For a given vehicle design and drive cycle, the fuel economy depends on the hybrid operation or energy management strategy. For a fair comparison, energy management for each design is optimized. Maximizing the fuel economy over a prescribed drive cycle is equivalent to minimizing the total losses from the engine and the hydraulic components. The mode of operation in Section II specifies how much power loss is incurred at each instant and how much accumulator flow is used. Thus, the optimal energy management problem for a given drive cycle is to minimize: min tf mode( ) t 0 tf subject to Loss (t, P (t), mode(t)) dt t 0 P (t) Q acc (t, mode(t))dt 0 (5) where Loss(t, P (t), mode(t)) is the total loss of the powertrain at time t of the drive-cycle if the accumulator pressure is P (t) and the operating mode if mode. The constraint is to ensure that net accumulator energy use is zero. Generally, the accumulator pressure dynamics and size constraint will also imposed. Solving Eq.(5) is typically computationally expensive. However, by assuming that system pressure is constant and accumulator size is unconstrained, the Lagrange Multiplier method can be used to speed up the evaluation of the energy management control and the evaluation of the fuel economy for a given design [1], [8]. The optimal control becomes: tf max min [Loss(t, P, modes(t))+ λ t 0 mode(t) λ P Q acc (t, mode(t))] dt (6) which is computationally very fast. D. Validation of Optimal Design without Simplifying Assumptions In order to verify that the simplifying assumptions made have minimal effect on the performance of the optimized design, dynamic programming is used for the optimized design to find the optimal power-train operation and fuel economy. Here the accumulator pressure dynamics are included and finite accumulator volume constraint is imposed. In this dynamic programming, volume of the accumulator is discretized and is used as the control variable. The power-train optimization procedure is illustrated in Fig. 3), A. Component Size IV. RESULTS AND DISCUSSIONS The optimized gear ratios, pump/motor sizes for the three hydraulic hybrid architectures are shown in table V. Generally, total displacements for the series architecture are larger than either HMT or parallel architectures. Hydraulic component efficiencies do not have a large effect on pump/motor sizes. However, when 2-speed operation is allowed the total Loss factor HMT PM1 [cc/rev]: PM2 [cc/rev]: G: R1: R2: HMT PM1 [cc/rev]: speed PM2[cc/rev]: G1: G2: R1: R2: Parallel PM1[cc/rev]: G: R: Parallel PM1[cc/rev]: speed G1: G2: R: Series PM1[cc/rev]: PM2[cc/rev]: G: Series PM1[cc/rev]: speed PM2[cc/rev]: G1: G2: TABLE V OPTIMIZED PUMP/MOTORS SIZES (IN CC) COMPARISON OF THREE HYBRID ARCHITECTURES pump/motor sizes decrease (20% for HMT, 40% for parallel, little change for series). B. Fuel economy comparison Fig. 4 shows the fuel economy and overall power-train efficiency for each optimized (1-speed) architecture for various hydraulic efficiencies (i.e. via loss factor). The maximum fuel economy with this engine is 90MPG. For each hydraulic efficiency, the HMT architecture is more efficient than either the parallel and the series architecture. Only when the loss factor is close to 0 (i.e. when the pump/motors are nearly ideal), series architecture has better fuel economy than parallel architecture. This is consistent with the expectation in Fig. 2. Also, the hydraulic efficiencies of the pump/motors with the same loss factors are higher in HMT than parallel and series. This implies that the extra degree of freedom in HMT enables the components to operate in more efficient regimes. C. Engine Operation Figure 5 shows the engine operating points over the drive cycle for the HMT and parallel architectures. Series architecture always operates the engine at its peak efficiency of 39%. Both HMT and parallel operate close to the maximum torque line, which is also where the engine is most efficient (with efficiencies range between 37%-39%). Note that HMT engine operation does not cluster around the engine sweet spot as expected. This is because the Prius engine has a flat contour at highest efficiency around the maximum torque region. As a result, even as the engine speed deviates from the sweet spot, the decrease in efficiency is minimal and is made up for by allowing other components to operate 5977

5 Fig. 4. Fuel economy and overall power-train efficiency of three hybrid architectures (1-speed) versus mean pump/motor efficiencies. Fig. 6. Fuel of 1-speed or 2-speed gear shifts on fuel economy for the 3 architectures (with loss-factor 1). TABLE VI HMT FUEL ECONOMIES COMPUTED USING THE LAGRANGE MULTIPLIER AND DYNAMIC PROGRAMMING METHODS. Loss Factor Lagrange Method [MPG] DP [MPG] Error 1.3% 0.7 % 1.2 % 1.4% 0.7 % Fig. 5. spot) Engine operating points(red: Parallel Green: HMT Star: sweet more efficiently. This will discussed further in Section IV-F. Normally, series is expected to be better than the parallel architecture when the pump/motors are efficient enough. However, the Prius engine, which is an Atkinson cycle engine, has a flat high efficiency range through the entire speed range (Fig. 5). This promotes the parallel architecture and makes it not as important to always operate the engine at the sweet spot. (with simplifying pressure dynamics and accumulator size assumptions) and using dynamic programming (without these assumptions). The results from the two approaches are very close for different pump/motor efficiencies. This verifies the assumptions and the optimization process we did are valid and the fuel economy value we get from the Lagrange multiplier method is reasonable and also practical. In Tab. VI, a 10gallon (38L) accumulator is assumed. Fig. 7 shows the effect of accumulator size on fuel economy. After a certain value, increasing accumulator size does not increase fuel economy significantly. The chosen accumulator size is close to this value. Figure 8 shows the state of charge of the accumulator (in terms of pressure) over the drive cycle for the 1-speed HMT architecture, and with the unscaled pump/motors. The accumulator pressure is kept low most of the time. The control strategy tends to maintain low pressure such that pump/motors will operate at higher displacements D. Effect of Gearshifts Fig. 6 shows that allowing a second gear increases the fuel economy of all three architectures. The fuel economy of the parallel architecture becomes very close to that of the HMT configuration. E. Dynamic Programming Verification Table VI shows the fuel economies for the optimized HMT architecture estimated using the Lagrange Multiplier method 5978 Fig. 7. Effect of accumulator size with fixed transmission design

6 Fig. 8. Accumulator SOC through the drive cycle while achieving the same torque requirements, since the efficiency is generally higher. Most of the high pressure portions are the braking events. The fact that the global optimal solution given by DP prefers to operate the accumulator at low pressure values for most of the time explains the minor effect on fuel economy that the simplifying constant pressure assumption in the Lagrange multiplier optimization method has. F. Dependency on Engine As mentioned earlier, the engine efficiency characteristic may affect the choice of the optimization of the engine operation.to study the effect of the engine efficiency map on the hybrid operation, an engine with a narrower efficient operating region (Fig. 9) is selected to investigate the effect on HMT and parallel architectures. The efficiency map is constructed by scaling the Saturn engine data provided by ADVISOR [12] to match the Prius engine power. The two-speed parallel architecture s fuel economy is 55.2MPG and that of the 1-speed HMT is 57.6MPG. The series architecture s fuel economy is 48.5MPG. The engine operating points in Figure 9 shows that for this engine, HMT does keep the engine operating close to the most efficient point; series operates exactly at the sweet spot, and parallel operates the engine over a large speed range at lower engine efficiencies. V. CONCLUSION This paper presented a generalized process of evaluating and comparing series, parallel and HMT hydraulic hybrid architectures. Results are that the HMT architecture have better fuel economy than series and parallel architectures. The HMT architecture allows full engine management and utilizes the efficient mechanical path to avoid energy loss through the relatively inefficient hydraulic components. The HMT architecture also operates the pump/motor more efficiently than series and parallel architecture. Additional gear shifts allow useful improvement in fuel economy, and reduce the pump/motor sizes, especially for the parallel and HMT architectures. Fig. 9. Engine operating points. (Red: parallel, Green: HMT.) which is supported by the National Science Foundation under grant number EEC REFERENCES [1] Li, P. Y. and F. Mensing, Optimization and Control of a Hydro- Mechanical Transmission based Hybrid Hydraulic Passenger Vehicle, in 7th International Fluid Power Conference, Aachen [2] James J. Kress, Hydrostatic Power-Splitting Transmission for Wheeled Vehicles - Classification and Theory of Operation, in Society of Automotive Engineers, number , [3] A. Brahma, Y. Guezennec, and G. Rizzoni, Optimal energy management in series hybrid electric vehicles, Proceedings of the 2000 American Control Conference, vol. 1, no. 6, pp.60-64, [4] C. Lin, H. Peng, J. Grizzle, and J. Kang, Power management strategy for a parallel hybrid electric truck, IEEE Transactions on Control Systems Technology, vol. 11, no. 6, pp , [5] P. Matheson and J. Stecki, Development and Simulation of a Hydraulic Hybrid Powertrain for use in commercial heavy vehicles, in SAE paper, [6] B. Wu, et. al., Optimal Power Management for a Hydraulic Hybrid Delivery Truck, in Vehicle System Dynamics, vol. 42, pp , [7] T. P. Sim and P. Y. Li, Analysis and Control Design of a Hydro- Mechanical Hydraulic Hybrid Passenger Vehicle, Proceedings of the 2009 ASME Dynamic Systems and Control Conference/Bath Symposium PTMC #2763, Hollywood, [8] Kai Loon Cheong, Perry Y. Li, and Thomas R. Chase, Optimal Design of Power-Split Transmission for Hydraulic Hybrid Passenger Vehicles, in Proceedings of the American Control Conference, [9] K. L. Cheong, P. Y. Li, S. Sedler and T. R. Chase, Comparison between input coupled and output coupled power-split configurations in hybrid vehicles, 2011 International Fluid Power Exhibition (IFPE), Paper number:10.2, Las Vegas, NV, March, [10] C-T. Li and H. Peng, Optimal Configuration Design for Hydraulic Split Hybrid Vehicles, in American Control Conference, Baltimore, MD, [11] A. Bryson nad Y.C. Ho, Applied Optimal Control, published by Taylor and Francis Group, [12] ADvanced VehIcle SimulatOR (ADVISOR), software developed by NREL. VI. ACKNOWLEDGMENTS This material is based upon work performed within the Center for Compact and Efficient Fluid Power (CCEFP) 5979