You are on page 1of 11

IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 62, NO.

5, MAY 2015 3305

Comparative Analysis of Bidirectional


Three-Level DCDC Converter for
Automotive Applications
Serkan Dusmez, Student Member, IEEE, Amin Hasanzadeh, Member, IEEE, and
Alireza Khaligh, Senior Member, IEEE

AbstractIn battery/ultracapacitor electric vehicles, a


bidirectional dc/dc converter is employed to process the
power according to the power references obtained from
the energy management controller. The selection of this
converter is of critical importance for the overall system
efciency and size. This study proposes using a three-
level dc/dc converter and provides a comprehensive com-
parison with the conventional two-level and interleaved
bidirectional buck/boost converters in terms of magnetic Fig. 1. Drivetrain of the battery/UC vehicle with a bidirectional
component size/weight and overall efciency. Unlike the converter.
comparative studies presented in the literature, where the
efciency comparison of converters is conducted based on
given xed input and output parameters, power references
obtained from a wavelet-transform-based energy manage-
ment strategy with varying energy source voltages and
traction power are considered in this paper. The results
of the analyses show that a three-level converter exhibits
higher overall efciency and has smaller size inductor. A
1-kW bidirectional three-level dc/dc converter is designed
as a proof of concept, which exhibits 93.2% peak efciency
at 200-kHz switching frequency. Fig. 2. Nonisolated bidirectional buck/boost converters. (a) Two-level
(CBC). (b) Interleaved (BIC). (c) Three-level (TLC).
Index TermsElectric vehicles (EVs), interleaved con-
verter, nonisolated dcdc converter, three-level converter,
connected to the dc link and serves as a power buffer. High-
ultracapacitor (UC).
power requirement can be met by the UC through controlling its
voltage within the inverters upper and lower operation voltage
I. I NTRODUCTION
limits, where the only controllable unit is the bidirectional dc/dc

U LTRACAPACITOR (UC)-assisted electric vehicles (EVs)


are attractive alternatives to battery-only EVs with high-
power batteries [1][8]. Different drivetrain architectures, in
converter. The selection of the bidirectional converter topology
is of critical importance since it greatly affects the overall
system efficiency.
which UCs are either actively or passively controlled, are pro- Many studies proposed and evaluated bidirectional dcdc
posed and compared in terms of efficiency, control complexity, converters for EVs, and a majority of them provided the effi-
and performance [9][11]. The most common type of connec- ciency comparison, which are only evaluated for some specific
tion is shown in Fig. 1. In this architecture, a UC is directly power, with fixed input and fixed output parameters [12][18].
However, the efficiency evaluation of dc/dc converters in EVs
Manuscript received March 9, 2014; revised April 23, 2014; accepted should consider the wide traction power and battery/UC oper-
May 31, 2014. Date of publication July 8, 2014; date of current version ation voltage ranges. This study proposes using a three-level
April 8, 2015. This work was supported in part by the National Science converter (TLC) [19], [20] as the bidirectional dc/dc converter
Foundation Electrical, Communications and Cyber Systems Division
under Grant 1307228 and in part by the Genovation Cars Inc. in EVs and compares it with candidate counterparts, i.e., two-
S. Dusmez was with the Power Electronics, Energy Harvesting and level and interleaved bidirectional buck/boost converters (CBC
Renewable Energies Laboratory (PEHREL), Electrical and Computer and BIC) [21], [22] given in Fig. 2, considering Urban Dy-
Engineering Department, University of Maryland, College Park, MD
20742 USA. He is now with the Electrical and Computer Engineering
namometer Driving Schedule (UDDS) as the drive cycle. The
Department, University of Texas at Dallas, Richardson, TX 75080 USA power distribution obtained by a three-level discrete wavelet
(e-mail: serkan.dusmez@utdallas.edu). transformation in [23], together with the corresponding UC
A. Hasanzadeh and A. Khaligh are with the Power Electronics, Energy and battery voltages, are used in the efficiency and magnetic
Harvesting and Renewable Energies Laboratory (PEHREL), Institute for
System Research and Electrical and Computer Engineering Depart- component size calculations. Thereby, the overall efficiency of
ment, University of Maryland, College Park, MD 20742 USA (e-mail: the converters over the full drive cycle is computed through
hasanzad@ece.umd.edu; khaligh@ece.umd.edu). developed efficiency models, and the converter with the highest
Color versions of one or more of the figures in this paper are available
online at http://ieeexplore.ieee.org. efficiency and smaller magnetic component size is identified.
Digital Object Identifier 10.1109/TIE.2014.2336605 The presented roadmap for the analysis can be further extended
0278-0046 2014 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
3306 IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 62, NO. 5, MAY 2015

where de is
de = (ds2 0.5) + (ds3 0.5). (2)
Here, dS2 and dS3 denote the duty cycles of switches S2 and
S3 . It is worth mentioning that the duty cycles of S2 and S3 are
the same; hence, the effective duty cycle can be rewritten as
de = 2d 1 (3)
where d = dS2 = dS3 . The output-voltage-to-input-voltage ra-
tio, when d < 0.5, can be found by equalizing mode II or III
and mode IV, i.e.,
VUC 2
= . (4)
VB 2 de
The effective duty cycle for this mode is the sum of the
Fig. 3. Switching scheme of the three-level converter for boost mode. individual duty cycles
(a) d > 0.5. (b) d < 0.5.
de = ds2 + ds3 = 2d. (5)
to other applications such as hybrid EVs, plug-in EVs, and fuel-
cell vehicles.
III. C OMPARISON OF T HREE -L EVEL C ONVERTER W ITH
II. B IDIRECTIONAL T HREE -L EVEL DCDC C ONVERTER THE S TATE - OF - THE -A RT C ONVERTERS

In a battery/UC-powered drivetrain, where the battery is con- TLC provides several advantages over the state-of-the-art
nected to the high-voltage dc link via a bidirectional converter, CBC and BIC converters. Switching losses highly depend on
the converter is power controlled through stepping-up the the voltage applied across the switch. Particularly, even without
battery voltage during propulsion and stepping-down the high- soft switching, the switching loss of the parasitic capacitance
voltage dc link to the battery voltage during regenerative can be significantly reduced in comparison to CBC as the
braking. A majority of the studies in the literature use the switches are subjected to half the output voltage. In fact, the
conventional two-level buck/boost converter (CBC) due to its parasitic capacitance losses are expected to be even lower as
simple structure and control [24][31]. However, in high-power low-voltage switches are used. On the passive component side,
applications, the boost inductor becomes the major component the diode reverse recovery losses are lower as reverse voltage is
that increases the volume, weight, and cost of the system. only half the output voltage, and low-voltage diodes typically
Moreover, high-voltage switches must be used, which, in turn, recover faster. This section compares bidirectional buck/boost
causes higher losses. By using TLC, significant improvements dcdc converter topologies, given in Fig. 2, in terms of mag-
can be achieved over the CBC. netic component size and efficiency over the full drive cycle
The switching scheme for the TLC converter is given in range, considering the dynamic variations of the battery and UC
Fig. 3. The switch turn-on times are phase-shifted, resulting in voltages as well as the power processed by the converter.
an effective inductor current ripple frequency that is equal to
twice the switching frequency. This effective period is shown
as Tse in Fig. 3. The boost and buck operation modes of the A. Drive Cycle Characteristic
TLC are shown in Figs. 4 and 5, respectively. Since operations The efficiency of the bidirectional converter varies to a great
of the buck and boost modes are similar, only boost mode is extent with respect to the load power and dynamic voltage
explained. The operation of the circuit can be divided into four variations of the energy sources. In the drivetrain configuration
modes. Based on the duty cycle value, the sequence of the in Fig. 1, the power of the converter is equal to the power
equivalent circuits is separated into two. The climax value of the requested from/to the battery. The sizing and power/energy
duty cycle of the switches, i.e., d, is 0.5. Basically, if d < 0.5,
management of energy storage systems are important for the
the sequence of equivalent circuits is IVIIIVIII and keeps
efficiency of the system, as it impacts the voltages of the energy
repeating. Since the control signals of each switch is displaced
sources and the battery power. In this study, the UDDS drive
by 180 , as the duty cycle gets higher than 0.5, the control
cycle, which simulates an urban route of 7.5 mi with frequent
signals begin to overlap. This overlap causes a new equivalent
stops, is chosen. The maximum and average speeds are 56.7
circuit to emerge. The new equivalent circuit sequence becomes
IIIIIII. and 19.6 mi/h. It is assumed that daily commute consists of
The output-voltage-to-input-voltage ratio for continuous- four UDDS drive cycles, corresponding to 30 mi. A 15.8-kWh
conduction mode (CCM) operation can be found through an- Li-ion battery pack with a nominal voltage of 350 V, which goes
alyzing the inductor current ripple within one cycle of the up to 407 V at a fully charged state, is considered. The battery
effective switching period. When d > 0.5, the voltage conver- voltage varies between 380 and 350 V in the linear operation
sion ratio can be expressed as region. The UC voltage is limited by the minimum operation
voltage of the inverter and set to 350 V as it is connected to
VUC 2 the dc link directly, whereas the upper voltage limit is set to
= (1)
VB 1 de 600 V. As energy stored in the UC is proportional to the square
DUSMEZ et al.: BIDIRECTIONAL THREE-LEVEL DCDC CONVERTER FOR AUTOMOTIVE APPLICATIONS 3307

Fig. 4. Boost-mode operation. (a) Mode I (S2 and S3 are on). (b) Mode II (S2 and D4 are on). (c) Mode III (D1 and S3 are on). (d) Mode IV
(D1 and D4 are on).

Fig. 5. Buck-mode operation. (a) Mode I (D2 and D3 are on). (b) Mode II (D2 and S4 are on). (c) Mode III (S1 and D3 are on). (d) Mode IV
(S1 and S4 are on).

size and efficiency analyses will be conducted based on these


data in the following sections.

B. Magnetic Component Size


The size of the magnetic component is related to the peak
flux density, which, in turn, is related to the peak current passing
through the core. In CBC, the current ripple of the battery can
be expressed as
Vo
ibat_boost (d) = d(1 d). (6)
Lfs
The battery current ripple for TLC is
Vo
ibat_TL (d) = de (1 de ). (7)
2Lfs
For the TLC converter, the relation between the current ripple
and duty cycle of a switch can be found by substituting the
effective duty cycles given in (3) and (5), i.e.,
Fig. 6. Dynamic variations of state variables [23]. (a) Converter power. 
Lf 
(b) UC voltage. (c) Battery voltage. Vo
d(1 2d) 
s d0.5 
ibat_TL (d) = (8)
Vo (1 d)(2d 1) .
Lfs
of its voltage, sufficient energy can be stored (C 118.75 kJ) d>0.5

with a properly chosen C for 600350-V voltage swing, where The input current of the BIC converter is the sum of two
C represents the UC capacitance. inductor currents. When the duty cycles of the switches are less
In this study, we discuss the battery and UC reference power than 50%, the input current ripple becomes
achieved using three-level wavelet decomposition [23]. The  
converter power and voltage variations of UC and battery for 2Vin Vo Vin
ibat_Interleaved (d) = 1 . (9)
four consecutive UDDS cycles are presented in Fig. 6. For Lfs Vo
this specific case, the battery is assumed to have 80% initial
When the duty cycle is greater than 50%, the input current
state of charge (SOC), and the UC has 91.6% SOC. The
ripple is expressed as
battery voltage varies between 381 and 357 V, the UC voltage  
varies between 575 and 384 V, whereas converter power varies Vo 2Vin Vin
ibat_Interleaved (d) = (10)
between 10.78 and 45.32 kW throughout the drive cycle. The Lfs Vo
3308 IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 62, NO. 5, MAY 2015

TABLE I
C ONVERTER C URRENTS D URING F OUR UDDS (FS = 20 kHz)

where the current ripple of each interleaving inductor is the


same as (6), i.e.,

Vo
iL_Interleaved (d) = d(1 d). (11)
Lfs

The maximum input ripple current for BIC and TLC occur
at 25% and 75% duty cycles, where for CBC, it occurs at 50%.
For the same maximum battery current ripple, CBC requires
four times larger inductance compared with TLC.
Fig. 7. Flowchart of the inductor design algorithm based on the core
To determine the magnetic core size, the required inductance geometry estimation approach.
and peak inductor current have to be quantified. In this regard,
the drive cycle given in Fig. 6 has been evaluated considering 2
where E is 0.5LIpeak . The electrical coefficient is a function of
a 20-kHz switching intermediate buck/boost converter, where the magnetic and electrical quantities as
the effective frequency of the input current ripple is 40 kHz
in BIC and TLC. The battery current ripple is limited to 20% Kg = 0.145 Po Bmax
2
104 . (13)
of the discharging current at the maximum discharge rate,
corresponding to 40% inductor current ripple for BIC. Based on The estimation approach is based on finding and evaluating
assumed ripple current, switching frequency, and the operating Kg . Some core manufacturers directly provide the Kg [cm5 ]
maximum voltage across the inductor with corresponding duty value for each core, whereas most of the manufacturers just
cycle, the inductances for CBC, BIC, and TLC are calculated provide area product Ac Wa [cm4 ], where Ac and Wa denote
as 400 H, 200 H (2), and 100 H, respectively. However, cross-sectional and winding areas, respectively. The relation
as will be discussed in the latter sections, 400 H results in a between the area product and Kg can be expressed as
very large magnetic core. Therefore, the inductance of CBC is A2c Wa
chosen as 200 H. Kg (14)
MLT
As seen from Table I, the maximum battery current reaches
to 138.7 A with a ripple of 32.9 A in CBC. The maximum where MLT is the mean length per turn. The current density can
inductor current of BIC is calculated as 77 A with an inductor be extracted from the maximum flux density and area product as
current ripple of 30.8 A. Although this ripple is not reflected on 2 E 104
the battery, still, it would cause high core losses, which will be J= . (15)
Bmax Ac Wa Ku
investigated in the latter sections.
To calculate the approximate size of the magnetic compo- The current density dictates the wire size, which is
nent, the core geometry approach is utilized. To estimate the
core size, several parameters such as peak current (Ipeak ), rms Irms
Aw = . (16)
current (Irms ), maximum flux density (Bmax ), regulation (), J
maximum output power (Po,max ), required inductance (L), Based on the given wire size and winding area, the number of
and window utilization factor (Ku ) should be determined. turns can be calculated as
is defined as the ratio of the voltage drop across the inductor
to the output voltage and is related with the copper losses. Wa Ku
N= . (17)
The energy handling capability, i.e., E, is a function of , Aw
core geometry coefficient Kg , and an electrical coefficient Ke , The flowchart of the inductor core selection is presented
which is determined by the magnetic and electrical conditions. in Fig. 7.
Thus Maximum flux density depends on the selected core material.
Magnetics or iron powder cores are distributed air-gap cores
E 2 = Kg Ke (12) whose saturation flux density is typically between 1 and 1.5 T
DUSMEZ et al.: BIDIRECTIONAL THREE-LEVEL DCDC CONVERTER FOR AUTOMOTIVE APPLICATIONS 3309

TABLE III
A NALYZED C ONVERTER S PECIFICATIONS

Fig. 8. Kg values of BIC and TLC cores at various switching


frequencies.
TABLE II
I NDUCTOR S IZE OF CBC, TLC, AND BIC (PO = 45.23 kW,
Bmax = 0.5 T, KU = 0.4, ALFA = 0.1%)

thermal properties in comparison with powdered iron cores, are


evaluated using the iterative algorithm given in Fig. 7. The area
product of this type of core is up to 791 cm4 . The algorithm
chooses the feasible core among the ones stored in the database.
The results related to core design are presented in Table II. The
weights of the cores are 21.3 kg, 1.3 kg, and 60.262 kg for
CBC, TLC, and BIC converters, respectively.
C. Semiconductor Selection Guidelines
Before choosing the power switches, the voltage and current
stresses on the switches should be analyzed. The switches in
CBC and BIC are exposed to output voltage, whereas in TLC,
[32]. These cores are primarily used in power inductor ap- the switches are exposed to half the output voltage. From the
plications, specifically in switched-mode power supply output analyses given in Section III-A, the maximum UC voltage is
filters, which are also known as dc inductors. According to 575 V. Therefore, CBC and BIC switches should be chosen
(12)(14), the area product of the core, which is an estimation with ratings higher than 575 V, whereas the voltage ratings
of the core size, is reduced when the maximum operating flux is of TLC switches should be higher than 287 V. In addition,
increased; however, core losses increase in turn. The maximum as the converters are hard-switched, the peak voltage across
flux density is typically chosen as Bmax /2, which, in this case, the switches can be well over this level due to the resonance
is assumed to be around 0.5 T. Fill factor Ku is typically set between the parasitic inductances of the circuit and the parasitic
to 0.4. In low-power applications, the regulation is typically output capacitances of the switches. Therefore, a safety margin
chosen as 1%; however, it is usually limited to a lower bound should be determined. Here, it is determined as 2, which
as power increases. As maximum power reaches 45.23 kW, dictates choosing 1200-V switches for CBC and BIC and 600-V
regulation is limited to 0.1%. The other parameters (Ipeak , Irms , switches for TLC.
I, L, and Po,max ) required to calculate the product area and The peak currents of the switches are equal to the peak
thereby estimate the core size are obtained from the dynamic current of the inductors, which has been given in Table I. For
simulation of four UDDS drive cycles. the given drive cycle, the peak switch currents are 138, 130,
Kg is a significant indicator of core size. Fig. 8 presents the and 77 A for CBC, TLC, and BIC, respectively. Based on these
calculated Kg values for TLC and BIC, at various switching values, IGBTs are potential semiconductor devices, whereas
frequencies. The Kg value of the CBC is significantly higher MOSFETs could be suitable switching devices for switching
and, therefore, is not given in Fig. 8. As shown in the figure, Kg frequencies in the range of 2040 kHz. However, it is difficult
is almost the same for TLC and BIC converters, whereas it is to find 1200 V and even 600 V discrete low-cost MOSFETs
slightly higher for TLC. It should be noted that the results given in the market with high current ratings. These limitations dic-
for BIC includes total Kg for two identical cores. Although tate two distinct choices of using either 1) high-power IGBT
Kg provides prior information on the size of the cores, it does modules or 2) parallel-connected MOSFETs. Considering the
not necessarily hold true in implementation as one cannot find switching frequency and cost of the switches, the parallel-
cores on the market with exactly the same winding and cross- connected MOSFET structure is found to be viable. To have
sectional areas as calculated. a fair comparison, 600- and 1200-V MOSFETs are chosen
To achieve approximate core sizes for the given specifica- from the same brand and category/family as different manufac-
tions, Kool Mu (sendust core) cores, which are well known turers define the semiconductor parameters in their own stan-
for their moderate cost, lower losses, and substantially better dards to advertise the products differently. Taking these criteria
3310 IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 62, NO. 5, MAY 2015

TABLE IV
C URRENT E XPRESSIONS FOR CBC C ONVERTER

into account, the switch IXFB30N120P is selected for two-level by charging the gate capacitance by gate voltage VG and then
converters CBC and BIC, and IXFX80N60P3 is selected for dumping the charge to the ground in every switching cycle;
TLC, where IXFB30N120P and IXFX80N60P3 are 1200-V/ hence, it is categorized under the switching losses. In addition,
30-A and 600-V/80-A power MOSFETs, respectively. In total, the charge stored in the parasitic output capacitor COSS during
12 IXFB30N120P MOSFETs are utilized for each converter, the turn-off period of the MOSFET causes power dissipation.
i.e., CBC and BIC, whereas eight IXFX80N60P3 MOSFETs Another source of switching losses is reverse recovery losses
are used for TLC. of the MOSFETs body diode, which is related to the reverse
recovery duration trr . At the turn-off transition of the diode,
IV. L OSS M ECHANISM AND E FFICIENCY A NALYSIS the excess charge Qrr stored in the drift region is removed
before the diode junction become reverse biased. This excess
To estimate the energy loss and efficiency of the converters
Qrr causes undesirable reverse recovery voltage, i.e., Vrr . The
over the entire drive cycle, efficiency models representing dif-
switching losses in CBC, BIC, and TLC are expressed by the
ferent losses, including conduction, switching, inductor core,
following expression:
and winding losses, are presented and evaluated for each con-
verter using 5605 data points, with the determined converter PSW = fs NPS NPC (0.5 VDS (t)ID (t)(tr + tf )
specifications as summarized in Table III. The expressions 
+ 0.5 VDS
2
(t)COSS + Qt VG + Vrr (t)Qrr (18)
for the switch and diode currents of CBC are given for each
operation mode considering CCM/DCM operation in Table IV. where fs is the switching frequency, and NPS and NPC are the
For BIC, the same equations are valid; however, the average number of paralleled switches and paralleled converters, which
current is halved for each switch. The expressions for the TLC are 6 and 1 for CBC, 3 and 2 for BIC, and 2 and 1 for TLC,
converter are provided in the Appendix. respectively. In addition, tr and tf denote the rise-time and fall-
time transitions of MOSFETs during switching periods.
A. Switching Losses
B. Conduction Losses
The switching losses consist of four major loss components:
1) power losses of MOSFET due to the overlap of current At low switching frequencies, the conduction loss is domi-
and voltage at the instant of switching; 2) gate charge losses; nant, and this loss depends on the ON-state resistance RDS(on)
3) parasitic capacitance losses; 4) reverse recovery losses of and drain RMS current ISRMS of MOSFET. The conduction
body diode of MOSFET. At frequencies above 20 kHz, the loss for CBC and BIC can be estimated using the following
switching losses contribute to a significant amount of power equation:
dissipation. The switching losses of the MOSFET are given
by the area under the waveforms of drainsource voltage VDS PCDCBCBIC = NPC RDS(on) IS2RMS (t)/NPS

and drain current ID . The gate charge loss, i.e., Q , is caused + NPS VF IDavg (t) (19)
DUSMEZ et al.: BIDIRECTIONAL THREE-LEVEL DCDC CONVERTER FOR AUTOMOTIVE APPLICATIONS 3311

Fig. 9. Energy losses at the switching frequency of 20 kHz. (a) Switching losses. (b) Conduction and core losses. (c) Overall loss.

where VF and IDavg are the forward bias voltage drop and the D. Efciency Analysis
average current of diode. For TLC, there is an additional term
Using (18)(24), the losses associated with the converters
in boost mode, which can be expressed as
are calculated considering the dynamic power and voltage
profiles explained in Section III-A. The resultant energy losses
PCDTLCBoost = RDS(on) IS2RMS (t)/NPS +NPS + VF IDavg (t)
 for all three converters at the switching frequency of 20 kHz
+ NPS VF ISavg (t) + IDavg (t) (20) are presented in Fig. 9. The corresponding switching losses
[Wh] are given in Fig. 9(a). The reverse recovery losses in
and for buck-mode operation, we have CBC and BIC converters are higher, particularly due to the
higher drainsource voltage across the diodes at the turn-off
PCDTLCBuck = RDS(on) IS2RMS (t)/NPS VF IDavg (t) instants. On the other hand, the voltage across the diodes in
2  TLC is half the output voltage, which causes lower power
+ RDS(on) ID RMS
(t) + IF2RMS (t) /NPS . (21) dissipation. Similarly, the losses associated with the parasitic
capacitance MOSFETs are significantly lower. Gate losses are
In buck and boost modes, the voltages and currents of diodes/ not significant at the given switching frequency.
switches are substituted accordingly. The conduction losses along with core and copper losses are
plotted in Fig. 9(b). Here, the conduction losses of the switches
in TLC are lower in comparison with the other converters, as
C. Inductor Losses
less number of switches is used in TLC. In contrary, diode
The copper loss of the inductor is calculated using winding losses are higher since the current flows through more diodes.
ohmic resistance, which can be calculated as The overall losses of each converter are presented in Fig. 9(c).
TLC exhibits slightly higher efficiency than others, although the
Cu 2
PCu = MLT NI (t) (22) overall conduction losses are higher. The results for the switch-
Aw LRMS ing frequency of 100 kHz are plotted in Fig. 10. As shown, the
where Cu is the copper resistivity constant. efficiency improvement is considerably higher in this case. This
In addition to copper losses, core losses should be con- also proves that TLC becomes more advantageous at higher
sidered, which are mainly due to hysteresis losses. Since all switching frequencies.
cores have been selected from Magnetics Inc. with relative Using the charts given above, the overall efficiencies of the
permeability of r = 60, the core loss of the inductor can be converters considering the full drive cycle can be determined
calculated by multiplying core volume Ve with core loss density through
Pe (Bac , fs ), which is a function of the ac magnetic flux density
of Bac and operating frequency fs , i.e., Po
= . (25)
(Po + PSW + PCD + PCu + PC )
PC = Ve Pe (Bac (t), fs ) . (23)
According to (25), the efficiencies are 98.7%, 98.5%, and
For the aforementioned permeability (r = 60), Pe (Bac , fs ) 98.8% for CBC, BIC, and TLC, respectively. In the case of
has been provided by Magnetics Inc. with the following equa- 100-kHz switching frequency, the efficiencies drop to 96.4%,
tion [32]: 96.3%, and 98%, respectively. As a result of the analysis, it is
 1.29 observed that TLC is the most efficient converter in comparison
2.01 fs to CBC and BIC, both in terms of magnetic component size
Pe (Bac , fs ) = 193Bac (t) . (24)
1000 and the efficiency under given driving cycle conditions. The
3312 IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 62, NO. 5, MAY 2015

Fig. 10. Energy losses at the switching frequency of 100 kHz. (a) Switching losses. (b) Conduction and core losses. (c) Overall loss.

difference in the efficiency is getting more significant as the


switching frequency increases.

E. Control Complexity
Due to the less number of switches, the control of CBC is
relatively simpler than BIC and TLC converters, where only
one carrier signal is sufficient to generate two PWM pulses. In
contrary, CBC and TLC require two carrier signals, which are
180 phase-shifted with respect to each other. Therefore, from
the PWM generation point of view, TLC control is the same as
that of BIC. However, three of the switches require isolated gate
drivers due to the floating ground. This adds to the circuitry as
Fig. 11. Photo of the designed converter. (a) Top view. (b) Bottom view.
well as to the cost.
The main difficulty of TLC control is the inclusion of a
neutral-point voltage-balancing controller. Under ideal condi- (low-voltage side). The voltages applied to the switches and
tions, the drive circuit and the main circuit are completely diodes are different in each case. The voltage spikes across the
symmetrical, and the voltages across the dc-link capacitors switches and diodes are due to the hard-switching and the para-
should be half the dc-link voltage. However, because of the sitic inductances on the printed-circuit-board layout, since their
asymmetry of the switches in series, the drive circuits and ESR contribution to oscillations becomes significant at 200-kHz sys-
of the dc bus capacitors, the dc-link capacitor voltages deviate tem frequency. The switching frequency and doubled inductor
from each other, which causes the switches and the dc-link ca- current frequency can be observed from the results.
pacitor to suffer from different voltage stresses. Therefore, TLC The efficiency curves obtained from the experimental stud-
necessitates an additional voltage balance control. One way to ies under different voltage and power are compared with the
accomplish this is to sense the upper or lower capacitor voltage calculated efficiencies under the same conditions in Fig. 14,
and add a compensating duty cycle to either S2 or S3 in boost to verify the accuracy of the developed loss model. As can
mode and to S1 or S4 in buck mode. With the compensating be seen from efficiency curves, the efficiency curve obtained
duty ratio, the pulse width of the switch is either extended or from the developed model successfully follows that obtained
shortened in comparison to that of its complementary switch. experimentally. The peak efficiency is calculated as 93.2% at
200-kHz switching frequency at 900 W.
V. E XPERIMENTAL R ESULTS
VI. C ONCLUSION
A 1-kW TLC converter prototype has been designed as the
proof of concept, as shown in Fig. 11. The switching frequency In battery/UC hybrid EVs, the selection of the bidirectional
is set to 100 kHz, where the effective ripple frequency becomes dc/dc converter topology, which processes the battery power, is
200 kHz. The inductor is chosen as 100 H. The switches are of great importance as it is one of the major factors contributing
IRFP350, which have a breakdown voltage of 400 V and allow to the size and efficiency of the system. Particularly in higher
16-A continuous drain current. The converter is controlled switching frequencies, the efficiency of this converter decreases
using a low-cost dsPIC30f4011 microcontroller. The inductor due to increased switching losses. This paper has proposed
current is sensed using a LEM CAS-25 sensor. using a three-level nonisolated bidirectional dc/dc converter
The experimental results for three operating points in boost (TLC) as the power electronics interface between the battery
mode are given in Figs. 12 and 13. The figures present the and the UC, instead of a conventional two-quadrant buck/
waveforms for various power and output voltages, i.e., Vo boost converter (CBC), which would increase the conversion
DUSMEZ et al.: BIDIRECTIONAL THREE-LEVEL DCDC CONVERTER FOR AUTOMOTIVE APPLICATIONS 3313

Fig. 12. Experimental waveforms for TLC boost mode when d = 0.25 (Vin = 90 V, Vo = 120 V, Po = 700 W). (a) Inductor voltage and current.
(b) Voltages across S2 and S3 . (c) Input voltage, output current, voltages across S1 and S4 .

Fig. 13. Experimental waveforms for TLC boost mode when d = 0.48 (Vin = 67 V, Vo = 130 V, Po = 500 W). (a) Inductor voltage and current.
(b) Voltages across S2 and S3 . (c) Input voltage, output current, voltages across S1 and S4 .

A PPENDIX
The average and rms current expressions for TLC operating
in CCM boost mode (IB VUC /(32Lf s ) for forward operation
are

d = 1 VB /VUC (26)

 
2 I2L
ISRMS = d ILmin + ILmin IL + (27)
3
(VB VUC /2)d
IL = (28)
Lf s

 
I2L
ISRMS = d I2Lmin + ILmin IL + (29)
3

ILRMS = I2SRMS + I2D1RMS (30)
ISAVG = dIB (31)
Fig. 14. Efficiency curve of the three-level converter operating at
200-kHz switching frequency. ID1AVG = (1 d)IB (32)
ID4AVG = ISAVG + ID1AVG . (33)
efficiency and reduce the size of the magnetic components. In
this regard, the three-level converter was analyzed and com- The average and rms current expressions for TLC operating in
prehensively compared with CBC and interleaved bidirectional DCM boost mode (IB < VUC /(32Lf s )) for forward operation
converter in terms of magnetic component size and efficiency are

considering a UDDS drive cycle where the battery and UC 4Lf s IB (VUC VB )
power are split using three-level wavelet decomposition. The d=
(VUC (VB VUC /2))
results prove that the TLC converter has the smallest size
magnetic component while providing considerable efficiency (34)
improvement under given driving cycle conditions, particularly (VB VUC /2)d
IL = (35)
at high switching frequencies. Lf s
3314 IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 62, NO. 5, MAY 2015


d= 4Lf s IB (VB VUC /2)/ (VUC (VUC VB )) (50)

(VUC VB )d
IL = (51)
Lf s

d
IS1RMS = IL (52)
3
d(VUC VB )
1 = (53)
(VB VUC /2)

(VB VUC /2)2 31 IL (VB VUC /2)21


IDRMS = 2 2
+ 1 I2L
(3L fs ) Lf s

(54)

ILRMS = IS4RMS = I2S1 + I2DRMS (55)
RMS

IS1AVG = dIB (56)

IDAVG = 1 IB (57)


d The average and rms current expressions for TLC operating
ISRMS = IL (36) in DCM buck mode (IB < VUC /(32Lf s )) for reverse operation
3
d(VB VUC /2) are given in (50)(57), shown at the top of the page.
1 = (37)
(VUC VB )

(VBVUC )2 31 IL (VBVUC )21 R EFERENCES


ID1RMS = 2 2
+1 I2L +
(3L fs ) Lf s [1] M. Ortuzar, J. Moreno, and J. Dixon, Ultracapacitor-based auxiliary
energy system for an electric vehicle: Implementation and evaluation,
(38) IEEE Trans. Ind. Electron., vol. 54, no. 4, pp. 21472156, Aug. 2007.
[2] O. Laldin, M. Moshirvaziri, and O. Trescases, Predictive algorithm for
ILRMS = I2SRMS = I2D1 (39) optimizing power flow in hybrid ultracapacitor/battery storage systems
RMS
for light electric vehicles, IEEE Trans. Power Electron., vol. 28, no. 12,
ISAVG = dIB (40) pp. 38823895, Aug. 2013.
ID1AVG = 1 IB (41) [3] I. Aharon and A. Kuperman, Topological overview of powertrains for
battery-powered vehicles with range extenders, IEEE Trans. Power Elec-
ID4AVG = ISAVG + ID1AVG . (42) tron., vol. 26, no. 3, pp. 868876, Mar. 2011.
[4] A. Khaligh and Z. Li, Battery, ultracapacitor, fuel-cell, hybrid energy
storage systems for electric, hybrid electric, fuel cell, plug-in hybrid
The average and rms current expressions for TLC operating electric vehicles: State-of-art, IEEE Trans. Veh. Technol., vol. 59, no. 6,
in CCM buck mode (IB VUC /(32Lf s )) for reverse opera- pp. 28062814, Jul. 2010.
tion are [5] D. Rotenberg, A. Vahidi, and I. Kolmanovsky, Ultracapacitor assisted
powertrains: Modeling, control, sizing, the impact on fuel economy,
IEEE Trans. Control Syst. Technol., vol. 19, no. 3, pp. 576589,
d = VB /VUC 0.5 (43) May 2011.
[6] W.-S. Liu, J.-F. Chen, T.-J. Liang, R.-L. Lin, and C.-H. Liu, Analysis, de-
(VUC VB )d sign, control of bidirectional cascoded configuration for a fuel cell hybrid
IL = (44)
Lf s power system, IEEE Trans. Power Electron., vol. 25, no. 6, pp. 1565

 
1575, Jun. 2010.
I2L [7] J. Jia, G. Wang, Y. T. Cham, Y. Wang, and M. Han, Electrical character-
IS1RMS 2
= d ILmin + ILmin IL + (45) istic study of a hybrid PEMFC and ultracapacitor system, IEEE Trans.
3 Ind. Electron., vol. 57, no. 6, pp. 19451953, Jun. 2010.

 
[8] M. H. Todorovic, L. Palma, and P. N. Enjeti, Design of a wide input
range DCDC converter with a robust power control scheme suitable for
I2L
IDRMS = (1 d) I2Lmax ILmax IL + (46) fuel cell power conversion, IEEE Trans. Ind. Electron., vol. 55, no. 3,
3 pp. 12471255, Mar. 2008.
[9] A. Emadi, K. Rajashekara, S. S. Williamson, and S. M. Lukic, Topo-
logical overview of hybrid electric and fuel cell vehicular power system
ILRMS = I2SRMS + I2D1 (47) architectures and configurations, IEEE Trans. Veh. Technol., vol. 54,
RMS
no. 3, pp. 763770, May 2005.
ISAVG = dIB (48) [10] J. Bauman and M. Kazerani, A comparative study of fuel-cellbattery
fuel-cellultracapacitor and fuel-cellbatteryultracapacitor vehicles,
ID1AVG = (1 d)IB . (49) IEEE Trans. Veh. Technol., vol. 57, no. 2, pp. 760769, Mar. 2008.
DUSMEZ et al.: BIDIRECTIONAL THREE-LEVEL DCDC CONVERTER FOR AUTOMOTIVE APPLICATIONS 3315

[11] P. Thounthong, V. Chunkag, P. Sethakul, B. Davat, and M. Hinaje, Serkan Dusmez (S11) received the B.S.
Comparative study of fuel-cell vehicle hybridization with battery or (Hons) and M.S. degrees in electrical engineer-
supercapacitor storage device, IEEE Trans. Veh. Technol., vol. 58, no. 8, ing from Yildiz Technical University, Istanbul,
pp. 38923904, Oct. 2009. Turkey, in 2009 and 2011, respectively, and
[12] U. R. Prasanna and A. K. Rathore, Extended range ZVS active-clamped the M.S. degree in electrical engineering from
current-fed full-bridge isolated DC/DC converter for fuel cell applica- Illinois Institute of Technology, Chicago, IL, USA,
in 2013. He is currently working toward the
tions: Analysis, design, experimental results, IEEE Trans. Ind. Electron.,
Ph.D. degree at the University of Texas at
vol. 60, no. 7, pp. 26612672, Jul. 2013. Dallas, Richardson, TX, USA.
[13] P. Xuewei and A. K. Rathore, Novel bidirectional snubberless naturally From 2012 to 2013, he was a Faculty Re-
commutated soft-switching current-fed full-bridge isolated DC/DC con- search Assistant with the the Power Electronics,
verter for fuel cell vehicles, IEEE Trans. Ind. Electron., vol. 61, no. 5, Energy Harvesting and Renewable Energies Laboratory (PEHREL),
pp. 23072315, May 2014. Electrical and Computer Engineering Department, University of
[14] A. Khaligh and A. Emadi, Mixed DCM/CCM pulse adjustment with Maryland, College Park, MD, USA. He is the author/coauthor of over
constant power loads, IEEE Trans. Aerosp. Electron. Syst., vol. 44, no. 2, 35 journal and conference papers. His research interests include de-
pp. 766782, Apr. 2008. sign of power electronic interfaces and energy management strategies
[15] D. Yu, Z. Xiaohu, B. Sanzhong, S. Lukic, and A. Huang, Review of for renewable energy sources, integrated power electronic converters
non-isolated bi-directional DCDC converters for plug-in hybrid electric for plug-in electric vehicles, and real-time fault diagnosis of power
vehicle charge station application at municipal parking decks, in Proc. converters.
IEEE APEC, 2010, pp. 11451151.
[16] R. M. Schupbach and J. C. Balda, Comparing DCDC converters for Amin Hasanzadeh (M11) received the B.Sc.
power management in hybrid electric vehicles, in Proc. IEEE IEMDC, degree from Sharif University of Technology,
2003, pp. 13691374. Tehran, Iran, in 1999, the M.Sc. degree from
[17] A. K. Rathore, Interleaved soft-switched active-clamped L-L type Khajeh Nasir Toosi University of Technology,
Tehran, in 2001, and the Ph.D. degree from
current-fed half-bridge dcdc converter, Int. J. Hydrogen Energy, vol. 34, Sharif University of Technology in 2009, all in
no. 24, pp. 98029815, Dec. 2009. electrical engineering.
[18] A. K. Rathore and U. R. Prasanna, Analysis, design, experimental results Between 1999 and 2010, he was a Senior
of novel snubberless bi-directional naturally clamped ZCS/ZVS current- Research and Development Engineer in several
fed half-bridge dc/dc converter for fuel cell vehicles, IEEE Trans Ind. industrial projects for different corporations in
Electron., vol. 60, no. 10, pp. 44824491, Oct. 2013. Tehran, where he was involved in the design and
[19] P. J. Grbovic, P. Delarue, P. L. Moigne, and P. Bartholomeus, A bidirec- production of analog/digital electronic circuits and power electronic con-
tional three-level DCDC converter for the ultracapacitor applications, verters in industrial systems. In 2011, he joined the Center for Advance
IEEE Trans. Ind. Electron., vol. 57, no. 10, pp. 34153430, Oct. 2010. Power Systems, Florida State University, Tallahassee, FL, USA, first as
[20] A. Shahin et al., High voltage ratio DCDC converter for fuel-cell ap- a Postdoctoral Research Associate and later as Research Staff. In 2013,
plications, IEEE Trans. Ind. Electron., vol. 57, no. 12, pp. 39443954, he joined the Power Electronics, Energy Harvesting and Renewable
Dec. 2011. Energies Laboratory (PEHREL), Department of Electrical and Computer
[21] L. Po-Wa, Y.-S. Lee, D. K. W. Cheng, and L. Xiu-Cheng, Steady-state Engineering, University of Maryland, College Park, MD, USA, as a
analysis of an interleaved boost converter with coupled inductors, IEEE Postdoctoral Research Associate. His research interests include power
electronics, application of control systems in power electronics systems,
Trans. Ind. Electron., vol. 47, no. 4, pp. 787795, Aug. 2000.
and hardware-in-the-loop realization of power electronics systems.
[22] H. Kosai, J. Scofield, S. McNeal, B. Jordan, and B. Ray, Design and
performance evaluation of a 200 C interleaved boost converter, IEEE
Alireza Khaligh (S04M06SM09) is an As-
Trans. Power Electron., vol. 28, no. 4, pp. 16911699, Apr. 2013. sistant Professor and the Director of the Power
[23] S. Dusmez and A. Khaligh, Wavelet-transform based energy and power Electronics, Energy Harvesting and Renewable
decoupling strategy for a novel ultracapacitorbattery hybrid power split Energies Laboratory with the Electrical and
gear powertrain, in Proc. IEEE ITEC, 2013, pp. 17. Computer Engineering (ECE) Department and
[24] B. Vural, S. Dusmez, M. Uzunoglu, E. Ugur, and B. Akin, Fuel con- the Institute for Systems Research at the Uni-
sumption comparison of different battery/ultracapacitor hybridization versity of Maryland, College Park (UMCP), Col-
topologies for fuel-cell vehicles on a test bench, IEEE J. Emerging Sel. lege Park, MD, USA. Prior to joining the UMCP,
Topics Power Electron, vol. 2, no. 3, pp. 552561, Sept. 2014. he was an Assistant Professor with Illinois In-
[25] A. S. Samosir and A. H. M. Yatim, Implementation of dynamic evolution stitute of Technology (IIT), Chicago, IL, USA,
control of bidirectional DCDC converter for interfacing ultracapacitor and a Postdoctoral Research Associate with
energy storage to fuel-cell system, IEEE Trans. Ind. Electron., vol. 57, the Grainger Center for Electric Machinery and Electromechanics,
no. 10, pp. 34683473, Oct. 2010. University of Illinois at UrbanaChampaign, Urbana, IL. He is an
[26] S. Dusmez and A. Khaligh, Generalized technique of compensating low- author/coauthor of over 120 journal and conference papers. His major
frequency components of load current with parallel bidirectional DC/DC research interests include modeling, analysis, design, and control of
power electronic converters for electric and plug-in hybrid electric vehi-
converter, IEEE Trans. Power Electron., vol. 29, no. 11, pp. 58925904, cles, energy harvesting, microrobotics, and renewable energy systems.
Nov. 2014. Dr. Khaligh was a recipient of various awards and recognitions,
[27] J. Dixon, I. Nakashima, E. F. Arcos, and M. Ortuzar, Electric vehicle including the 2013 George Corcoran Memorial Award from the ECE
using a combination of ultracapacitors and ZEBRA battery, IEEE Trans. Department of the University of Maryland, the 2013 and 2012 Best Ve-
Ind. Electron., vol. 57, no. 3, pp. 943949, Mar. 2010. hicular Electronics Awards from the IEEE Vehicular Technology Society
[28] A. Khaligh and S. Dusmez, Comprehensive topological analysis of con- (VTS), the 2010 Ralph R. Teetor Educational Award from the Society
ductive and inductive charging solutions for plug-in electric vehicles, of Automotive Engineers, and the 2009 Armour College of Engineering
IEEE Trans. Veh. Technol., vol. 61, no. 8, pp. 34753489, Oct. 2012. Excellence in Teaching Award from IIT. He is the Program Chair of
[29] J. Moreno, M. E. Ortuzar, and J. W. Dixon, Energy-management the 2015 IEEE Applied Power Electronics Conference and Exposition
system for a hybrid electric vehicle, using ultracapacitors and neural (APEC). He was the General Chair of the 2013 IEEE Transportation
networks, IEEE Trans. Ind. Electron., vol. 53, no. 2, pp. 614623, Electrification Conference and Exposition (ITEC), the Assistant Program
Apr. 2006. Chair of the 2014 APEC, the Program Cochair of the 2012 ITEC, and
[30] A. A. Ferreira, J. A. Pomilio, G. Spiazzi, and L. de Araujo Silva, Energy the Program Chair of the 2011 IEEE Vehicle Power and Propulsion
management fuzzy logic supervisory for electric vehicle power supplies Conference. He is an Editor of the IEEE T RANSACTIONS ON V EHICULAR
T ECHNOLOGY (TVT), an Associate Editor of the IEEE T RANSACTIONS
system, IEEE Trans. Power Electron., vol. 23, no. 1, pp. 107115, ON T RANSPORTATION E LECTRIFICATION , and a Guest Associate Edi-
Jan. 2008. tor for the Special Issue of the IEEE J OURNAL OF E MERGING AND
[31] S. Dusmez and A. Khaligh, A supervisory power splitting approach S ELECTED TOPICS IN P OWER E LECTRONICS on Transportation Elec-
for a new ultracapacitorbattery vehicle deploying two propulsion ma- trification. He was a Guest Associate Editor for the Special Issue of
chines, IEEE Trans. Ind. Informat., vol. 10, no. 3, pp. 19601971, the IEEE T RANSACTIONS ON P OWER E LECTRONICS on Transportation
Aug. 2014. Electrification and Vehicle Systems, for the Special Section of the
[32] Kool mu Cores, Datasheet. [Online]. Available: http://www.mag-inc.com/ IEEE TVT on Sustainable Transportation Systems, and for the Special
PRODUCTS/POWDER-CORES/KOOL-MU/KOOL-MU-MATERIAL- Section of the IEEE TVT on Vehicular Energy Storage Systems. He is
CURVES an IEEE VTS Distinguished Lecturer.

You might also like