Boost Composite Converter Design Based On Drive Cycle Weighted Losses in Electric Vehicle Powertrain Applications

Transcription

1 Boost Composite Converter Design Based On Drive Cycle Weighted Losses in Electric Vehicle Powertrain Applications Hyeokjin Kim, Hua Chen, Dragan Maksimović and Robert Erickson Department of Electrical, Computer and Energy Engineering University of Colorado Boulder Boulder, Colorado, 839 Abstract A weighted design optimization is introduced to minimize total loss of electric vehicle drivetrain power electronics over EPA standard drive cycles. It is shown that the net loss of the conventional boost converter can be reduced by a factor of.5 with this approach, while computational effort is reduced by three orders of magnitude. Even larger efficiency improvements are achieved by optimized boost composite converters: losses are reduced by factors of 4.5, 2.9, and 4.3 for US6, UDDS, and HWFET driving cycles, respectively. These design optimization results are experimentally verified with a 3 kw laboratory prototype boost composite converter, which demonstrates 98.4% average efficiency over the US6 driving cycle. I. Introduction Improvements in efficiency of the power conversion unit (PCU) in electric vehicle (EV) or hybrid electric vehicle (HEV) powertrains translate directly into increased MPGe (Mile-per-Gallon equivalent), as well as in size and cost benefits associated with thermal management. In the case of light duty vehicles, city or highway MPGe is measured over EPA standard test driving cycles: UDDS (urban dynamometer driving schedule) or HWFET (highway fuel economy test) [], while CAFE (corporate average fuel economy) MPGe is calculated by weighting the city MPGe by 55 % and the highway MPGe by 45 %. Figure shows one possible PCU architecture consisting of a motor inverter and a boost converter module and conventional architecture is explained in [2]. This architecture decouples battery and machine optimization and control. As a result, inverter, motor, and system efficiencies can be higher compared to a single-stage architecture where the battery is connected directly to the inverter [3], [4]. To further increase system efficiency, DC bus voltage control has been introduced [4], [5]. With the DC bus control, inverter and boost converter achieve loss reduction, compared to single-stage architecture. However, although DC bus voltage control reduces the converter average loss over driving cycles, magnetics design of boost converter significantly contribute to the average efficiency. Therefore, converter optimization takes into important role to achieve high average efficiency over driving cycles. In [6], multiple parameters such as switching device, switching frequency, and magnetic volume are taken account for optimization and loss model is developed. However, optimization is performed at one operating point which does not represents V battery Boost composite converter V bus Inverter Motor Fig.. Cascaded power conversion unit (PCU) employing boost composite converter driving cycles. In [7], multi-criteria optimization is proposed to take into account of motor losses and converter losses. Based on the analytical modeling of motor, three motor operating points, acceleration, rated power, and over speed region, are considered for optimization. However, considering only three operating points excludes major operating points over driving cycles. The most effective method for converter optimization over driving cycles is to perform point-by-point loss evaluation over a driving cycle. However, such a brute-force optimization requires a very large computational effort. To reduce the computational effort without loss of accuracy, a weighted loss optimization is proposed in this paper. The objectives are: system-level design optimization of the boost converter shown in Fig. based on the weighted loss method, verification of the performance of proposed weighted loss method, and demonstration of the results in a 3 kw prototype operating from 2-3 V input and capable of generating up to 8 V output. The electric vehicle simulation is developed in MATLAB/Simulink and the resulting simulation data and analysis are discussed in Section II. The weighted loss method is explained in Section III and the design optimization of composite boost converter based on the weighted loss method is presented in Section IV. Projected losses of the composite boost converter are reduced by factors of 4.5, 2.9, and 4.3 for US6, UDDS, and HWFET driving cycles, respectively, relative an equivalent conventional boost design. Section V presents experimental results for the 3 kw composite converter prototype, which demonstrates 98.6 % peak efficiency

2 Speed ref Driver Torque Current regulator V q V d PMAC Torque Wheel Force vehicle + _ Mv S Speed Speed id iq idq iabc ia ib ic Force resistance Air and rolling resistance Fig. 2. Block-diagram of a simplified electric vehicle powertrain simulation model and projects 98.4 % average efficiency over the US6 driving cycle. II. Electric vehicle powertrain simulation model TABLE I EV simulation parameters Vehicle Mv (Vehicle weight) [kg] Cd (Aerodynamic drag coefficient).28 Av (Veh. front cross-section area) [m 2 ] 2.2 Wheel radius [m].334 Maximum speed 95 mph Gear ratio 7.5 Environment Cr (Rolling Ω coefficient). ρ (air density) [kg/m 3 ].24 PMAC traction motor Pole 2 Ls [mh].76 φ f [Wb].8 A powertrain simulation model of an electric vehicle is developed in MATLAB/Simulink. A simplified top-level block diagram is shown in Fig. 2. The vehicle model follows a speed reference using a closed-loop controller represented as the driver in the diagram. The torque command output of the driver is fed into the current regulator which drives a traction motor. Parameters of the permanent magnet traction AC (PMAC) machine and the vehicle are listed in Table I. The vehicle parameters are imported from a Nissan LEAF [8] and motor parameters are estimated based on the Parker GVM PMAC motor. PMAC is driven by the conventional constant angle with field weakening control scheme. To reduce converter and inverter losses, a variable DC bus voltage control scheme is employed for inverter DC bus voltage control. With this complete vehicle model and control scheme, US6, UDDS, or HWFET driving cycles are simulated based on the following assumptions: Space vector control scheme is applied for motor control. Mechanical brake is disabled so that deceleration is achieved by regenerative braking. Parasitic circuit inductance and capacitances, oscillation during switching transitions, and dead-time of PWM control are neglected. Motor stator inductance and resistance are constant. Battery voltage is constant, 25V, and considered as an ideal voltage source. The speed schedule of US6 driving cycle and resulting data, required inverter bus voltage, and magnitude of motor power, are shown in Fig. 3. To minimize system losses, the required bus voltage which is the reference voltage for converter control can be represented as V bus,re f = α(vds 2 + V2 qs) () while V ds and V qs are the voltage of d and q stationary axis, and α is the number to achieve minimum converter and inverter losses while avoid field weakening control. In this case, α is taken to be 3.. The required bus voltage and motor power are distributed over a wide operating range. Operating points in the normalized power versus bus voltage plane are shown as density plots in Figs. 4, 5, and 6, for US6, UDDS, and HWFET driving cycles. Higher frequency counts of operating points are represented by darker shadings. One may note that the most frequently encountered operating points are approximately 73 V for V bus and 3% for motor power over the US6 driving cycle, 3 V for V bus and 6% for motor power for UDDS, and 65 V for V bus and % of motor power for HWFET. Consideration of only one operating point for powertrain module optimization, however, results in significant average efficiency degradation. To maximize average efficiency while reducing computational effort without loss of accuracy, a weighted loss method is proposed and its resulting improvement is discussed in section III. III. Design optimization based on a drive-cycle based weighted loss model To optimize the powertrain power conversion unit, an exhaustive search method is performed based on the comprehensive loss model described in [9]. The objective is to find the set of parameters that minimize the loss over a selected drive cycle. Brute-force, point-by-point loss evaluation over a drive cycle requires a prohibitively large computational effort. Instead a weighted loss method is applied as follows:

3 MPH 5 Vbus [V] Time [s] Power [kw] 3 Fig. 3. Speed schedule of US6 driving cycle and simulation results, required inverter bus voltage, and magnitude of motor power Fig. 5. UDDS driving cycle density plot of operating points Fig. 4. US6 driving cycle density plot of operating points Fig. 6. HWFET driving cycle density plot of operating points loss weighted = N loss op(n) probability op(n) (2) n= where N is the number of operating points considered for optimization, which is also the number of bins for the DC bus voltage (X-axis) in the plots shown in Fig. 4 through Fig. 8. The goal of this optimization is to find the optimum set of design parameters that minimize the weighted loss, Eq. 2. In this paper, the US6 driving cycle is considered for system optimization. The operating points to be considered are shown in Fig. 7. The Y-axis represents the most-used normalized power at the corresponding bus voltage. For example, 3% motor power is the most frequently used at V bus = 4V and the value can be found in Fig. 4. Also, V bus = 4V with 3% motor power has approximately 2% probability over the US6 driving cycle. Figure 8 shows the probability of each operating point represented in Fig. 7. For larger N, a more accurate optimization result is obtained, but the computation time increases proportionally. The normalized computation times and the results for average efficiency of US6 driving cycle based on the conventional boost converter as function of N are listed in Table II. Compared to the brute-force approach, the weighted loss model with N = 6 reduces the computation time by more than 3 orders of magnitude, with essentially no loss in accuracy. Also, in the case when only the most frequently used operating point is employed, V bus = 73V and 3% motor power, the lowest average efficiency design is obtained.

4 Most used normalized power at corresponding bus volt. [P / Prated] IV. Composite converter design based on the weighted loss optimization The proposed weighted loss method is further applied to the composite converter []. The composite converter architecture consists of boost, buck, and DCX (DC transformer) modules. The operating mode is represented in Fig. 9. The operating points to be considered for optimization is shown in Fig. and the back ground color represents operating region which corresponds the back ground color in Fig Vbus [V] Fig. 7. Most frequently encountered normalized power vs. bus voltage over the US6 driving cycle.3.25 Probability of most used normalized power Vbus [V] Vin Fig. 9. Operating region map of composite converter Vbus [V] Pbus [kw] Probability [%] Mboost Mbuck DCX (off) (off) Fig. 8. Probabilty of most frequent normalized power at corresponding bus voltage over the US6 driving cycle passthrough.58.5 (on) TABLE II Comparison of computation time and projected average efficiency as functions of N Number of Computation time US6 data [Normalized value] avg. efficiency N = Full data N = 28. N = 64.7 N = 6.3 N = 4.2 N = (most used op.) >. 96. % 94.9 % passthrough Fig.. Operating points of modules to be considered for converter optimization Based on the weighted loss method, the composite converter is optimized over the US6 driving cycle using an exhaustive search method. The flow chart of this optimization method for the boost module is shown in Fig.. The projected

5 efficiencies of the conventional boost converter and the composite converter over US6, UDDS, or HWFET driving cycles and CAFE average efficiency are listed in Table III. With the proposed weighted loss method, the composite converter achieves loss reductions by factors of 4.5, 2.9, and 4.3 over US6, UDDS, and HWFET driving cycles, compared to the conventional boost converter. Core material Library Core size US6 operating points data analysis Vin Prob. [%] Mboost Winding AWG Winding turns Winding strands fs Exhaustive search Search minimum Fig kw rating composite converter prototype with Silicon Super- Junction MOSFET Fig.. method Flow chart of converter optimization based on the weighted loss TABLE III Average efficiency over US6, UDDS, or HWFET driving cycle of conventional boost or composite boost converter Driving Conventional Composite cycle boost converter boost converter US % 98.4 % UDDS 97. % 99. % HWFET 9.8 % 98. % CAFE 94.7 % 98.6 % V. Experimental Results Based on the composite converter optimization summarized in Section IV, the 3 kw composite converter prototype shown in Fig. 2 has been fabricated. The module specifications are listed in Table IV. Fig. 3 shows the measured waveforms of DCX primary and secondary, buck, and boost switching node voltages and DCX primary, buck, and boost magnetic currents ripple at V in = 25V, V bus = 65V, and at 5 kw. Also, the losses of individual modules are predicted. The measured waveforms at V in = 25V, V bus = 65V, and rated power of 3 kw, are shown in Fig. 4. One may note that all modules achieve high efficiency at rated power. The comparison of measured efficiency, the loss model efficiency and conventional boost converter at the operating point of V in = 25V, V bus = 65V vs. power is shown in Fig. 5. The boost composite converter achieves 98.6 % peak efficiency at 5 kw power and maintains high efficiency over a remarkably wide operating range. The efficiency contour plot at 25V in in the bus voltage versus power plane is shown in Fig. 6. As red color, 99 % of efficiency is represented. Also, converter operating points over US6 driving cycle are overlayed as a blue solid line. Over the US6 driving cycle, the converter operates over high efficiency region. Based on the optimization results of the weighted loss method, the projected average efficiency of US6 driving cycle is 98.4%. TABLE IV Composite converter design summary Buck / Boost Number of MOSFET die MOSFET IPW65R4CFD2 Switching frequency 2 khz Inductance 6 μh for Boost, 48 μh for Buck Inductor core material Amorphous alloy DCX Number of MOSFET die 28 MOSFET IPW65R4CFD2 Switching frequency 33 khz Tank inductance 4.5 μh Transformer turns ratio 8:2 Transformer core material Ferrite VI. Conclusions This paper is focused on minimization of losses over a drive cycle in power conversion unit for electric vehicle powertrain applications. Over a drive cycle, the converter is operating over very wide ranges of voltage and power levels, which complicates design optimization. A weighted loss method is introduced to reduce the number of operating points to be considered, resulting in substantially reduced computing effort without loss of accuracy. The complete EV powertrain

6 DCX Pri. Sw. node voltage DCX Pri. Sw. node voltage (a) DCX Sec. Sw. node voltage Buck Sw. node voltage (a) DCX Sec. Sw. node voltage Buck Sw. node voltage Boost Sw. node voltage Boost Sw. node voltage DCX Pri. Tx. current DCX Pri. Tx. current (b) (b) Boost inductor current ripple Buck inductor current ripple Boost inductor current ripple Buck inductor current ripple Buck (D=), Pass-through 6W, =99.8% 25V DCX 43W, =98.4% 375V Buck (D=), Pass-through 44W, =99.7% 25V DCX 466W, =96.9% 32 V (c) 65 Vbus (c) 65 Vbus 25 Vin Boost (D=.9) 76W, =98.8% 275V 25 Vin Boost (D=.26) 283W, =98.2% 338 V Fig. 3. Measured waveforms at 25 V in, 65 V bus, 5 kw, (a) DCX, buck and boost switching node voltages, (b) DCX, buck, boost magnetic currents ripple, and (c) predicted efficiency and loss of modules Fig. 4. Measured waveforms at 25 V in, 65 V bus, 3 kw, (a) DCX, buck and boost switching node voltages, (b) DCX, buck, boost magnetic currents ripple, and (c) predicted efficiency and loss of modules simulation model is developed and resulting data is analyzed for operating points to be considered. Then, the weighted loss method is applied to conventional boost and boost composite converter. It is shown how the optimized conventional boost converter losses are reduced by a factor of.5 for the US6 driving cycle, compared to the losses of the conventional boost converter which is optimized for only most used operating point, while computational effort is reduced by more than 3 orders of magnitude. Even larger efficiency improvements compared to the conventional boost converter are obtained using the optimized boost composite converter: losses are reduced by factors of 4.5, 2.9, and 4.3 for the US6, UDDS and HWFET cycles, respectively. The design optimization results are experimentally verified on a 3 kw laboratory prototype of the boost composite converter, which demonstrates 98.6% peak efficiency and high overall efficiency over wide operating points. 98.4%, 99.%, and 98.% efficiencies over the three considered drive cycles are projected, respectively. Efficieny [%] Measured Comprehensive loss model Conventional DC-DC Power [kw] Fig. 5. Comparison of measured efficiency, comprehensive loss model efficiency, and conventional boost converter efficiency at V in = 25V and V bus = 65V, versus power

7 Fig. 6. Projected efficiency of composite converter at 25 V in and operating points of converter (blue solid line) over US6 driving cycle Acknowledgement The information, data, or work presented herein was funded in part by the Office of Energy Efficiency and Renewable Energy (EERE), U.S. Department of Energy, under Award Number DE-EE692. References [] J. Gonder and A. Simpson, Measuring and reporting fuel economy of plug-in hybrid electric vehicles, National Renewable Energy Laboratory, Conference Paper NREL/CP , 26. [2] K. Muta, M. Yamazaki, and J. Tokieda, Development of new-generation hybrid system THS II Drastic improvement of power performance and fuel economy, in SAE paper 24-, 24. [3] J.-S. Lai and D. J. Nelson, Energy management power converters in hybrid electric and fuel cell vehicles, Proceedings of the IEEE, vol. 95, no. 4, pp , 27. [4] F. Caricchi, F. Crescimbini, G. Noia, and D. Pirolo, Experimental study of a bidirectional dc-dc converter for the dc link voltage control and the regenerative braking in pm motor drives devoted to electrical vehicles, in Applied Power Electronics Conference and Exposition, 994. APEC 94. Ninth Annual IEEE, 994, pp [5] W. Qian, H. Cha, F. Z. Peng, and L. M. Tolbert, 55-kw variable 3x dc-dc converter for plug-in hybrid electric vehicles, Power Electronics, IEEE Transactions on, vol. 27, no. 4, pp , 22. [6] W. Martinez, M. Yamamoto, P. Grbovic, and C. A. Cortes, Efficiency optimization of a single-phase boost dc-dc converter for electric vehicle applications, in IECON 24-4th Annual Conference of the IEEE Industrial Electronics Society, 24, pp [7] Z. Wu, D. Depernet, and C. Espanet, Optimal design of electrical drive and power converter for hybrid electric powertrain, in 2 IEEE Vehicle Power and Propulsion Conference, 2, pp. 8. [8] J. G. Hayes and K. Davis, Simplified electric vehicle powertrain model for range and energy consumption based on epa coast-down parameters and test validation by argonne national lab data on the nissan leaf, in Transportation Electrification Conference and Expo (ITEC), 24 IEEE, 24, pp. 6. [9] H. Kim, H. Chen, D. Maksimovic, and R. Erickson, Design of a high efficiency 3 kw boost composite converter, in Energy Conversion Congress and Exposition (ECCE), 25 IEEE, Sept 25, pp [] H. Chen, K. Sabi, H. Kim, T. Harada, R. Erickson, and D. Maksimovic, A 98.7% efficient composite converter architecture with applicationtailored efficiency characteristic, Power Electronics, IEEE Transactions on, vol. 3, no., pp., 26.