Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
Abstract— This paper presents a method for efficiency esti- of high-performance power devices and enabled high-power
mation of boost-derived continuous conduction mode power density solutions through high switching frequency operation.
factor correction (CCM-PFC) converters for electric vehicle (EV) At high frequency, medium-to-high power applications the
onboard chargers. The proposed methodology incorporates con-
verter nonidealities, especially caused by magnetic components. major portion of power losses arises from magnetic cores (PFC
The value of magnetizing inductance in an inductor or trans- inductors) and the switching transients (i.e., turn-ON and
former core does not remain constant over variable current turn-OFF losses). A reasonably accurate estimate of power
levels, which causes nonuniform power losses at different current losses in individual components is essential in order to eval-
levels. The method proposed in this paper considers a time- uate the efficiency of the converter before building up the
variant inductance over various current levels and accordingly
establishes a dynamic model of loss estimation. As a proof-of- prototype.
concept verification, the approach is applied to three different To evaluate the switching and conduction losses of the
PFC topologies for EV applications and the estimated conversion MOSFETs, several studies have been conducted and mod-
efficiencies exhibit good agreement with experimentally obtained els are set up to describe the losses in a rigorous way.
efficiency values over a wide range of load power from 400 W Since the parasitic output capacitance varies with the drain–
to 4.6 kW. The deviation of the efficiency predicted from the
experimental data is considerably lower in comparison with the source voltage, it is required to estimate instantaneous value
existing estimation methods with fixed inductance assumption. of COSS as a function of VDS in order to accurately get
the parasitic loss. The turn-ON losses of a MOSFET are
Index Terms— CCM PFC, dynamic, electric vehicle, loss
estimation, on-board charger. strongly influenced by the stored amount of charge in its
output capacitance Coss . Thus, in order to accurately model
the turn-ON losses, the method in [2] and [3] determines the
I. I NTRODUCTION
total charge Q OSS at particular blocking voltages. In addition,
excitation winding and induced voltage across a sensing wind- contribution of this paper. This paper proposes an approach,
ing over one switching period with a nonsinusoidal ac input which considers an inductance as a current-varying function
excitation is proposed in [1] and [8]. In this approach the in a CCM-operated PFC converter. The method considers
magnetic flux density (B) is calculated piecewise depending a triangular variation of the current instead of assuming it
on different excitation voltages in multilevel full/half-bridge to be constant in a switching cycle. Accordingly, the max-
topologies. Furthermore, the research work in [9] and [10] imum intensity and minimum field intensity (Hmax , Hmin)
shows that zero applied voltage (constant flux) does not are calculated at each switching interval and thus, core-loss
necessarily imply zero core loss due to relaxation process. energy is computed. Cumulative addition of the core loss
Accordingly, a new core-loss modeling approach is taken energy over a line cycle provides the effective core loss.
and verified experimentally. However, these methods are not Furthermore, the proposed method helps estimating the con-
directly applicable in a PFC-based system, since core loss has duction and switching losses in every cycle by accurately
a high dependence on current ripple information. determining the rms value of switch current. Also, for the
Second, the work in [11] discusses a modified analytical purpose of verification, the proposed method is applied to
method for core losses calculation in magnetic laminations different existing PFC topologies for EV charging applications.
for wide range of frequencies and flux densities. This method This paper is organized as follows. Section II describes the
assumes a known value of magnetic flux density, which varies approximations and idealities in the aforementioned methods
with the change in current level in a PFC-based converter. with fixed inductance assumption for efficiency estimation.
An overall power loss estimate (including magnetic core loss) The proposed efficiency estimation method is discussed in
requires an accurate flux density estimation in an excited Section III. Section IV presents the validation and results
inductor core, which is not outlined in [11]. for verification of the proposed concept by applying it to
Third, based on Maxwell’s equation and flux density varia- three different efficiency data sets. Section V puts forward
tion patterns using 2-D finite-element methods, core-loss char- conclusions with relevant discussions.
acteristics of stator permanent-magnet machines are analyzed
and calculated in [12]. In addition, core-loss calculation of II. A PPROXIMATIONS AND I DEALITIES IN
different soft magnetic composite (SMC) materials used in C ONVENTIONAL E STIMATION M ETHOD
several electrical circuits is modeled using a 3-D finite-element A typical loss breakdown for a PFC converter contains
analysis and the loss model is validated with experimental inductor core loss, inductor copper loss, switching loss,
measurements of a SMC motor prototype [13]. In each of the conduction loss of MOSFETs, and equivalent series resis-
aforementioned methods, the fundamental assumption of the tance (ESR) loss on dc-link capacitors. Other loss sources
loss estimation is to have a correct set of values for magnetic such as PCB traces and parasitic resonance among components
flux density (B) and magnetizing field strength (H ); how- and traces are generally negligible if proper design practices
ever, accurately determining these parameters require highly are implemented. To estimate these losses, the high-frequency
time-complex computations, which makes the implementation current ripple of input current must be estimated first. How-
highly challenging. ever, either assuming an ideally fixed inductance value or fixed
There are several attempts to estimate the power losses for amplitude of high frequency current ripple will not be accurate
the entire PFC converter [14]–[16]. The power loss evaluation as it dismisses three factors which will impact the accuracy of
for a PFC unit was presented in [14] using the temperature the calculation:
measurements of the individual components and surrounding 1) nonlinearity of the inductor as a function of inductor
ambience. The temperature profile of an ac–dc converter unit current level and excitation frequency;
at different operating points is reported and accordingly power 2) high-frequency component in the input current;
loss estimates for each component are obtained using the 3) time-variant nature of duty cycle.
mutual correlation between temperature gradient and the tested Therefore, the impact of those factors on the loss estimation
power level. In addition, a simplified method to evaluate the should be investigated and clarified.
power losses for boost-derived PFC converters is proposed
in [16]. However, none of aforementioned approaches consider
A. Nonlinearity of Inductors and High-Frequency Component
nonlinearity of magnetic components into their modeling.
of Input Current as Loss Factors
In practice, most of the core loss calculation methods deal
with fixed inductance method (assuming effective permeability As shown in Fig. 1, the input current of a CCM-operated
same as its initial value) and fixed dc bias method [17], which PFC contains two different components: low-frequency sinu-
accounts for the effective permeability at the rms current soidal and high-frequency triangular components. As shown
level. Neither of these two existing practices considers the in Fig. 1, the high-frequency component in the current has a
instantaneous variation of permeability with current level, time-variant amplitude. This phenomenon is caused by mainly
which would lead to inaccuracies in core-loss estimation. The two reasons: the time-variant duty ratio and input line voltage,
variation of permeability will further impact the accuracy of as well as nonlinearity of the inductor. At any instant, the high
switching and conduction loss estimation, which will also be frequency inductor current ripple can be calculated by the
discussed in this paper. following:
Accounting for these nonidealities and establishing a deter- D(t)V in (t)
i L (t) = . (1)
ministic expression for overall power loss is the main Lfs
LU et al.: DYNAMIC STRATEGY FOR EFFICIENCY ESTIMATION 547
cycle (Hmax and Hmin ) into (6), respectively. Hmax and Hmin
are determined by the maximum and minimum current levels
in a switching cycle.
small perturbation to its previous value, which is equivalent Plugging (18) into (17) and equating it to (16), the sufficient
to simulating the feedback loop. This process is repeated and necessary condition for μrms = μ is obtained as the
iteratively until the check condition satisfies. It is noteworthy following:
to mention that this iterative loop to track the input current is
particularly valid for a CCM-operated converter, which is kept bi (t)2 + 2b I i (t)sin(2π f line t)+di (t)4
as the main focus of this paper. + 4dI 3 i (t)sin(2π f line t) + I i (t)3 sin(2π f line t)
Then, using (9), real-time core loss can be calculated for aI cI 3 dI4
each cycle, and be integrated over the entire line voltage cycle. + 6d I 2 i (t)2 sin(2π f line t) = √ + √ − . (19)
2 2 2 8
One of the fundamental variables in core-loss calculation
is the inductance value, considering instantaneous variation The coefficients a, b, c, and d are the inputs from the
of inductance due to permeability variation over the current core material properties, which in general, can be any value.
level. An empirical relation, established in [17], between Therefore, (19) cannot hold true at any given conditions, which
permeability and current level is given by means the deviation exists between the estimations of core
loss by assuming μrms and μ, respectively. Moreover, this
μ = 1 + a Iin + b Iin2 + cIin3 + d Iin4 (15)
difference is also verified through experiments and calculation
where the phase current “Iin ” is related to the magnetizing examples in the later sections of this paper.
field intensity as: Iin = le H /N.
In the existing practical methods [15]–[17], μ is assumed
constant at μrms as expressed in (16), which is the permeability B. Capacitor ESR Loss
calculated using (15) √ under the condition of the rms input In a PFC converter, the loss on dc-link capacitor is normally
current, i.e., Iin = (I / 2) assuming Iin (t) = I sin(2π f line t). considerable. ESR of capacitor can be calculated by the
Furthermore, the core loss is proportional to the change in dissipation factor (DF) obtained from datasheet
magnetic flux density, which is again proportional to the μrms
aI bI2 cI 3 dI4 DF
μrms = 1 + √ + + √ + . (16) ESR = (20)
2 4 2π f test C
2 2 2
On the other hand, in the proposed method, the magnetic where f test is the testing frequency.
core loss is a function of time-averaged permeability μ. In an n-phase boost-derived PFC (e.g., n = 2 if PFC is a
Assuming the phase current Iin (t) = I sin(2π f line t) + i (t), two-leg interleaved converter), the average capacitor current
which is the superposition of the line-frequency sinusoidal and over one switching cycle can be calculated by
the high frequency triangular, the time-averaged permeability
can be written into the form of i c (t) = n Iin (t) − Io (21)
μ = 1 + aIin + b Iin2 + c Iin3 + d Iin4 (17) where Iin (t) represent the input current per phase.
With the predicted real-time current waveform, capacitor
where
current containing high-frequency information can be accu-
Iin = I sin(2π f line t) + i (t) rately obtained and hence, the change of output voltage for
2 each switching cycle can be estimated as well. Then, ESR loss
I = I 2 sin2 (2π f line t) + i (t)2 +2I i (t)sin(2π f line t)
in3 can be determined by
Iin = I 3 sin3 (2π f line t) + i (t)3 + 3I 2 i (t)sin(2π f line t)
+ 3I i (t)2 sin(2π f line t) PCap = [4C f line (Vo.max − Vo.min )]2 ESR. (22)
4
Iin = I 4 sin4 (2π f line t) + i (t)4 + 4I 3 i (t)sin(2π f line t)
+ 4I i (t)3 sin(2π f line t) + 6I 2 i (t)2 sin(2π f line t). C. Conduction Loss and Switching Loss
Note that over any line cycle, both I sin(2π fline t) and i (t) With the predicted real-time current waveform, accurate
are odd functions, then the averages of different exponents of estimation of conduction loss and switching loss is straightfor-
the phase current can be simplified as shown in the following ward using (10). Based on all the modeling work, the flowchart
equation: of dynamic loss estimation strategy can be summarized as
⎧ shown in Fig. 3. In Fig. 3, D is the change step of the
⎪
⎪ Iin = 0 duty cycle to simulate the function of the feedback control
⎪
⎪
⎪
⎪ I2 loop in a CCM-PFC; T is the time step of the calculation,
⎪
⎪ Iin2 = + i (t)2 + 2I i (t)sin(2π f line t)
⎨ 2 which is set to be the pulse width modulation switching period.
Iin3 = 0 Moreover, there is a trade-off between calculation accuracy
⎪
⎪
⎪
⎪ 4 3I 4 and the computation time by tuning the value of ξ in Fig. 3.
⎪
⎪ Iin = + i (t)4 + 4I 3 i (t)sin(2π f line t)
⎪
⎪ 8 An unreasonably small ξ can result in an undesirably long
⎩
+ I i (t)3 sin(2π fline t) + 6I 2 i (t)2 sin(2π fline t). computation time but ξ must be relatively small to ensure a
(18) good accuracy.
550 IEEE TRANSACTIONS ON TRANSPORTATION ELECTRIFICATION, VOL. 3, NO. 3, SEPTEMBER 2017
f s 1.29
As an illustration to the differences among the core-loss PCore = 193 Aele (B) 2.01
. (30)
1000
estimation results from the proposed dynamic method and
the other two methods, the topology of interleaved bridgeless However, in proposed dynamic method, the time-variant
PFC in the well-established literature [21] is used as an exam- nature of the inductance leads to the time-variant ripple current
ple. As shown in Table I, the Kool-Mu powder 77071 core is amplitude. Therefore, the power loss density of the core at
selected due to its low loss feature. At any given Vin and Pin , different instants during the line cycle varies. To address this
the input current per phase can be calculated as nonideality of the inductor for higher calculation accuracy,
√ the cycle-by-cycle, real-time calculation is implemented. The
2Pin
Iin (t) = sin(2π fline t). (25) calculation results of this example are summarized in Table II.
2Vin The results of validation are presented in Figs. 5 and 6,
With the time-variant current flowing through the inductor, respectively. Fig. 5 shows the results on a two-leg interleaved
the permeability and the inductance will vary. The datasheet boost PFC converter at different input voltages and power
from the manufacture provides the relation between the perme- levels. Fig. 6 shows the results on an interleaved bridgeless
ability and the magnetic field intensity, shown in (26), as well PFC at different input voltages and load conditions. As can be
LU et al.: DYNAMIC STRATEGY FOR EFFICIENCY ESTIMATION 551
TABLE II
C ALCULATION R ESULTS
R EFERENCES
[1] A. Mallik, W. Ding, and A. Khaligh, “A comprehensive design approach
to an EMI filter for a 6-kW three-phase boost power factor correction
rectifier in avionics vehicular systems,” IEEE Trans. Veh. Technol.,
vol. 66, no. 4, pp. 2942–2951, Apr. 2017.
[2] Y. Xiong, S. Sun, H. Jia, P. Shea, and Z. J. Shen, “New physical insights
on power MOSFET switching losses,” IEEE Trans. Power Electron.,
vol. 24, no. 2, pp. 525–531, Feb. 2009.
[3] C. L. Shen and Y. S. Shen, “Conversion efficiency estimation for an
interleaved high-step-up converter,” in Proc. Int. Conf. Adv. Mater. Sci.
Fig. 8. Comparison of ripple current versus time between calculated and Eng. (ICAMSE), Tainan, Taiwan, 2016, pp. 661–664.
experimental results. [4] J. Galvez, X. Perpinya, X. Jorda, M. Vellvehui, and J. Millan, “Semi-
conductor power losses estimation method for matrix converter appli-
cations,” in Proc. IEEE Eur. Conf. Power Electron. Appl., Sep. 2009,
pp. 1–8.
[5] C. Zhao, S. D. Round, and J. W. Kolar, “An isolated three-port bidi-
rectional DC-DC converter with decoupled power flow management,”
IEEE Trans. Power Electron., vol. 23, no. 5, pp. 2443–2453, Sep. 2008.
[6] M. S. Ortmann, S. A. Mussa, and M. L. Heldwein, “Evaluation of
semiconductor losses and efficiency in single-phase multilevel multistate
switching cells based PFC rectifiers,” in Proc. IEEE Southern Power
Electron. Conf., Fortaleza, Brazil, Nov. 2015, pp. 1–5.
[7] A. Shahin, A. Payman, J. P. Martin, S. Pierfederici, and
F. Meibody-Tabar, “Approximate novel loss formulae estimation
for optimization of power controller of DC/DC converter,” in Proc.
36th Annu. Conf. IEEE Ind. Electron. Soc., Glendale, AZ, USA,
Nov. 2010, pp. 373–378.
[8] V. J. Thottuvelil, T. G. Wilson, and H. A. Owen, Jr., “High-frequency
measurement techniques for magnetic cores,” IEEE Trans. Power
Fig. 9. Comparison of efficiency between calculated and experimental results. Electron., vol. 5, no. 1, pp. 41–53, Jan. 1990.
[9] J. Muhlethaler, J. Biela, J. W. Kolar, and A. Ecklebe, “Improved core-
loss calculation for magnetic components employed in power electronic
systems,” IEEE Trans. Power Electron., vol. 27, no. 2, pp. 964–973,
Feb. 2012.
[10] S. Barg, K. Ammous, M. Hanen, and A. Ammous, “An improved
empirical formulation for magnetic core losses estimation under non-
sinusoidal induction,” IEEE Trans. Power Electron., vol. 32, no. 3,
pp. 2146–2154, Mar. 2017.
[11] A. M. Takbash and P. Pillai, “A modified analytical method for core
losses calculation in magnetic laminations for a wide range of frequency
and flux density,” in Proc. IEEE Electr. Mach. Drives Conf. (IEMDC),
May 2015, pp. 1109–1114.
[12] S. Zhu, M. Cheng, J. Dong, and J. Du, “Core loss analysis and
calculation of stator permanent-magnet machine considering DC-biased
magnetic induction,” IEEE Trans. Ind. Electron., vol. 61, no. 10,
pp. 5203–5212, Oct. 2014.
[13] Y. Guo, J. Zhu, H. Lu, Z. Lin, and Y. Li, “Core loss calculation for soft
Fig. 10. Calculated inductance versus magnetic field intensity H . magnetic composite electrical machines,” IEEE Trans. Magn., vol. 48,
no. 11, pp. 3112–3115, Nov. 2012.
[14] K. Viswanathan and R. Oruganti, “Evaluation of power losses in a boost
magnetic field intensity for high-flux and Kool-Mu materials PFC unit by temperature measurements,” IEEE Trans. Ind. Appl., vol. 43,
are shown in Fig. 10. no. 5, pp. 1320–1328, Oct. 2007.
[15] A. Mallik and A. Khaligh, “Variable-switching-frequency state-feedback
control of a phase-shifted full-bridge DC/DC converter,” IEEE Trans.
V. C ONCLUSION Power Electron., vol. 32, no. 8, pp. 6523–6531, Aug. 2017.
[16] F. Musavi, D. S. Gautam, W. Eberle, and W. G. Dunford, “A simplified
In this paper, a novel efficiency estimation method for power loss calculation method for PFC boost topologies,” in Proc. IEEE
Transp. Electrific. Conf. Expo. (ITEC), Detroit, MI, USA, Jun. 2013,
CCM-PFC converters is introduced and analyzed. The anom- pp. 1–5.
alies between experimentally obtained efficiencies and their [17] (2013). Magnetics Powder Core Catalog. [Online]. Available:
theoretical calculation by the existing practices arise due to https://www.mag-inc.com/Design/Technical-Documents/Powder-Core-
Documents.aspx
ignoring several converter nonidealities including the inductor [18] A. Mallik and A. Khaligh, “Control of a three-phase boost PFC converter
core nonlinearity, which is addressed in this paper. The pro- using a single DC-link voltage sensor,” IEEE Trans. Power Electron.,
posed method exhibits a good agreement with the actual exper- vol. 32, no. 8, pp. 6481–6492, Aug. 2017.
[19] C. Shi, H. Wang, and A. Khaligh, “Interleaved SEPIC power factor
imental data and error magnitudes of estimated efficiencies are preregulator using coupled inductors in discontinuous conduction mode
below 1% for two different PFC topologies (i.e., interleaved with wide output voltage,” IEEE Trans. Ind. Appl., vol. 52, no. 4,
boost and interleaved bridgeless), whereas the estimations pp. 3461–3471, Aug. 2016.
[20] D. S. Gautam, F. Musavi, M. Edington, W. Eberle, and W. G. Dunford,
ignoring nonidealities have a higher deviation (2%–8%) from “An automotive onboard 3.3-kW battery charger for PHEV application,”
the actual values. This paper presents the accuracy of this IEEE Trans. Veh. Technol., vol. 61, no. 8, pp. 3466–3474, Oct. 2012.
LU et al.: DYNAMIC STRATEGY FOR EFFICIENCY ESTIMATION 553
[21] F. Musavi, W. Eberle, and W. G. Dunford, “A high-performance Alireza Khaligh (S’04–M’06–SM’09) is currently
single-phase bridgeless interleaved PFC converter for plug-in hybrid an Associate Professor with the Department of Elec-
electric vehicle battery chargers,” IEEE Trans. Ind. Appl., vol. 47, no. 4, trical and Computer Engineering (ECE), the Insti-
pp. 1833–1843, Jul./Aug. 2011. tute for Systems Research in the University of
Maryland (UMD), College Park, MD, USA. He has
authored or co-authored over 160 journal and con-
Jiangheng Lu (S’16) received the bachelor’s degree ference papers. His current research interests include
in electrical engineering from the Huazhong Uni- modeling, analysis, design, and control of power
versity of Science and Technology, Wuhan, China, electronic converters for transportation electrifica-
in 2014. He is currently pursuing the Ph.D. degree tion, renewable energies, energy harvesting, and
with the Electrical and Computer Engineering microrobotics.
Department, University of Maryland, College Park, Dr. Khaligh received various awards and recognitions including the
MD, USA. 2017 Outstanding Young Alumnus Award from Illinois Institute of Tech-
He is involved in hardware and control devel- nology, the 2016 E. Robert Kent Junior Faculty Teaching Award and the
opment, thermal management, and EMI filter 2016 Junior Faculty Outstanding Research Award from the Clark School of
design. His current research interests include mod- Engineering, UMD, the 2015 Junior Faculty Fellowship from the Institute for
eling, analysis, design and development of On-bard Systems Research, UMD, the 2013 George Corcoran Memorial Award from
Charger and High step-down auxiliary power module for electric vehicles. the ECE Department, UMD, three Best Vehicular Electronics Awards from
the IEEE Vehicular Electronics Society, and the 2010 the Ralph R. Teetor
Ayan Mallik (S’14) received the B.Tech and Educational Award from the Society of Automotive Engineers. He is the
M.Tech degrees (under 5 years Dual Degree Pro- Area Editor of “Vehicular Electronics and Systems Area” of the IEEE
gram) in electrical engineering from IIT Kharagpur, T RANSACTIONS ON V EHICULAR T ECHNOLOGY. He is an Associate Editor
Kharagpur, India, in 2014. He is currently pursuing of the IEEE T RANSACTIONS ON P OWER E LECTRONICS , the IEEE J OURNAL
the Ph.D. degree with the Electrical and Computer OF E MERGING AND S ELECTED T OPICS IN P OWER E LECTRONICS , and the
Engineering Department, University of Maryland, IEEE T RANSACTIONS ON T RANSPORTATION E LECTRIFICATION. He was the
College Park, MD, USA. General Chair of the 2016 IEEE Applied Power Electronic Conference and
His current research interests include design, mod- Expo, Long Beach, CA, USA, and the 2013 IEEE Transportation Electrifi-
elling, and control of power electronic converters, cation Conference and Expo, Dearborn, MI, USA, and the Program Chair of
highly efficient and high-power density solutions the 2011 IEEE Vehicle Power and Propulsion Conference, Chicago, IL, USA.
for power conversions in the applications of more- He is a Distinguished Lecturer of the IEEE Vehicular Technology Society and
electric-airplanes, electric vehicles, and data centers. the IEEE Industry Applications Society.