This article has been accepted for publication in a future issue of this journal, but has not been

fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/JESTPE.2017.2761763, IEEE Journal
of Emerging and Selected Topics in Power Electronics
1

Analysis, Design and Demonstration of a

25-kW Dynamic Wireless Charging System for
Roadway Electric Vehicles
Reza Tavakoli, Student Member, IEEE, and Zeljko Pantic, Member, IEEE

 Secondary Side (Vehicle Side)

Abstract— Dynamic wireless charging of Electric Vehicles
(EVs) can significantly extend the EVs’ driving range and Battery Conv- Sec.
consequently, the prospect of electrified transportation. In this Pack -erter Comp.
paper, a comprehensive study is conducted to elaborate the Secondary coil
constraints of real driving conditions and propose a solution that
could cope with misalignment problem and the dynamics imposed EVDS Primary
by the charging process and by EVs passing over road-embedded
charging pads. A dual-loop primary controller is proposed to EVDS Secondary Primary coils
regulate primary side power and current. The controller allows
sequential and timely activation of segmented primary coils; it Primary
Current
controls the primary coil current at the reference value under no- AC Main Inverter Pri.
load and loaded conditions, compensates for power transfer 60 Hz 20-100 kHz Comp.
reduction caused by the vehicle lateral misalignment, and prevents Primary Side (Road Side)
primary overloading. The primary of the dynamic wireless
charger is modeled using the generalized state space averaging Fig. 1: An illustration of DWPT system with segmented primary coils.
method and the model is verified through simulations and
experiments. After that, a controller has been designed and
considerable reduction in battery size and capacity [1].
implemented and its operation is evaluated through simulations DWPT expands the concept of stationary charging by
and experimental tests. A 25-kW charging system with two allowing an EV to be charged while driving over the primary
primary coils is built and tested in a real environment. The unit. Advancements in power electronics have made it possible
measured energy efficiency is 86% for the laterally aligned vehicle, to transfer tens of kilowatts of power wirelessly between
with the possibility to be increased over 90% using enhanced stationary road-embedded pads and vehicle-attached modules.
schemes for coils’ activation and deactivation. The system is
delivering an equal amount of energy for all lateral misalignments
However, the motion of the secondary unit (receiver) imposes
in the range of ±15 cm, which improves the expected value of a new set of challenges to the design and operation of the
transferred energy by more than 30%. primary unit (transmitter) which is due to the constantly varying
load of the system. To maintain the maximum power transfer
Index Terms— Battery Charging, Dynamic Wireless Power and high operating efficiency, an advanced level of control and
Transfer, Electric Vehicle, Roadway Electrification, Misalignment intelligence needs to be integrated into the system.
Compensation.
In Fig. 1, a typical DWPT system with segmented coil structure
at the primary is shown. On the primary side, the 60-Hz grid
voltage is converted to a DC voltage and further to a high-
I. INTRODUCTION
frequency current (20-100 kHz), which is delivered to the
In the past decade, there has been a significant interest in primary coil. This power conversion is achieved through an
emerging technologies for transportation electrification and intermediate DC link, a phase-shift controlled full-bridge
Electric Vehicles (EV). Today’s vehicle energy storage units inverter, and a properly selected compensation tank. On the
still suffer from low energy and power density, high cost and secondary side, there is another compensation tank, which is
size, and limited lifetime [1–3]. Limited energy capacity and connected to the secondary coil. These tanks supply the reactive
power of on-board battery modules are identified as the main power for the coils inductances and filter harmonics injected by
challenges for successful implementation of EVs. Dynamic the inverter or the rectifier units. High-frequency AC power at
Wireless Power Transfer (DWPT) offers an alternative solution the secondary side is rectified and then through a buck or boost
for limited on-board energy resources through hybridization converter delivered to the vehicle battery. In order to detect the
between vehicles battery storage and the power grid. This position of a moving EV, some form of EV Detection System
technology not only extends the EV range but also results in a (EVDS) is typically applied. Fig. 1 depicts the EVDS that has

2168-6777 (c) 2017 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.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/JESTPE.2017.2761763, IEEE Journal
of Emerging and Selected Topics in Power Electronics
2

been used in this project and will be discussed further in Section loaded and unloaded conditions, explain the dynamics of the
IV. Primary coils are energized consecutively based on the primary unit, and finally, design a dual-loop power and current
position of the moving vehicle. In other words, only the coil that controller. The controller will be implemented in a segmented-
has the EV’s receiving pad over it will be connected to the coil 25-kW DWPT system to charge an electric bus. More
primary compensation tank and energized. specifically, for the primary side, the Small-Signal (SS) model
Research on WPT systems can be divided into three broad of the LCC-compensated WPT system is derived using the
categories: i) coil/pad designs, ii) compensation topologies and Generalized State Space Averaging (GSSA) method. This
iii) power converters and control methods [4]. Regarding the model is verified through simulations and experimental tests.
coil/pad design, a lot of work has been done and several coil Then, a DWPT controller is designed which performs the
structures are proposed both for stationary and dynamic following actions: (i) energize the primary coils successively
charging systems. For stationary charging, circular coil and brings about high-dynamic current responses under no-load
structures are studied and optimized in [5], while in [6] the condition, (ii) keeps the primary current at the reference value
bipolar pad structure, DD, and DDQ pad structures are studied. under no-load and loaded condition, (iii) compensates LaTeral
For the primary side of a dynamic charging system, elongated Misalignment (LTM) by “offering” to transfer more power
loops are suggested in [7], [8]. In such schemes, due to the under misaligned conditions, and (iv) prevents overloading of
limited size of the receiver coils, a long portion of the primary the primary side. In this system, EVDS not only notifies about
loop is energized even though only a small portion of it the presence of the EV but also determines the degree of LTM
contributes to power transfer. That may result in low efficiency between primary and secondary coils with high accuracy. Based
and significant field leakage. Alternately, authors in [9] propose on the EVDS signal and the amount of transferred power, the
a design based on segmented coils where only the optimally controlling module continuously adjusts the primary coil
positioned coil is energized ensuring higher system efficiency. current during the DWPT window. The DWPT window finishes
The other DWPT infrastructure, presented in [10], uses a when the mutual inductance between primary and secondary
double-coupled system. drops below 5% of its peak value.
The role of a compensation tank in a WPT system is to i) This paper is a developed and extended version of [28] and it
minimize VA rating and maximize power transfer capability of is organized as follows: Section II describes the main
the system, ii) allow voltage source to current source challenges of a DWPT system and the effects of some remedial
conversion, and finally iii) filter the harmonics injected by the strategies; section III is dedicated to system modeling and
inverter or rectifier units [11]. Four basic compensation model verification; Section IV, deals with the controller design
topologies, namely series-series, series-parallel, parallel-series, and discusses the integration of the controller and EVDS;
and parallel-parallel, are analyzed in [12]. In [13], an LCL_T Sections V and VI present simulation and experimental results,
based resonant converter is presented and in [14] a double-sided respectively and are followed by some concluding remarks
LCC topology is discussed. given in Section VII.
Regarding the power conversion at the primary side, some
authors propose single-stage AC-AC converters to directly II. PROBLEM STATEMENT
generate a high-frequency voltage from 60-Hz grid voltage In a lossless and perfectly tuned DWPT system, the
[15], [16]. However, the two-stage power conversion with transferred power can be described as [29]:
intermediate DC link and a single-phase full-bridge converter
is more common due to easier control and lower harmonic
M2 2
content. Pout  s I QL
There are various studies exploring control techniques for L2 (1)
power delivery in stationary WPT systems. The objective is to
increase the system efficiency or provide control over the where M is mutual inductance between primary and secondary
secondary output voltage [17]–[20]. Several papers have also coils, I is the RMS value of the primary current, QL is the quality
addressed zero voltage/current switching of a full-bridge factor of the loaded secondary, and ωs is the operation
inverter for an efficient primary operation [21]–[24]. However, frequency determined by the switching frequency of the
few studies have been conducted regarding the modeling and primary inverter. From (1), one can see that three parameters,
operation dynamics of DWPT systems. In [20], dynamics of a namely ωs, QL or I could be controlled to regulate power
series-series compensated WPT system is analyzed to control transfer. In DWPT systems, constant frequency operation is
the output voltage. In [25], the authors use variable frequency preferred over the variable frequency, particularly when
to regulate delivered power. In [26], a small-signal modeling of multiple receiving units are supplied simultaneously from a
an LCL_T-based WPT system is presented to control the RMS single primary source.
value of the primary current under no-load conditions. Finally, Operation with a very high secondary quality factor QL
in [27], the power transfer of DWPT systems are considered and causes high reactive power in the secondary and hence, tuning
an algorithm is proposed to control the amount of energy issues, while a very low QL does not provide enough filtering
transferred to the receiver. and requires a larger secondary coil. Therefore, extensive QL
The objective of this research is to study and develop an variation is not a preferred method to control the amount of
accurate model of the primary side of a DWPT system for transferred power. In the testbed used in this research, the

2168-6777 (c) 2017 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.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/JESTPE.2017.2761763, IEEE Journal
of Emerging and Selected Topics in Power Electronics
3

50
50 y=0 cm
Trajectory 2 (x,y2,h) Experiment y=0 cm
Secondary Coil y=10 cm
y=10 cm y=15 cm
Trajectory 1 (x,y0,h) h 40 y=20 cm 40 y=15 cm
y=20 cm
h
30
Ground Clearance

Pout (kW)
Trajectory 3 (x,y3,h) 30
y2

M (µH)
h 20 20
R
y3 Primary
(x0,y0,0)
Coil 10 10

0 0
-70 -50 -30 -10 10 30 50 70 -70 -50 -30 -10 10 30 50 70
LNM (cm) Longitudinal Misalignment (cm)
-10
Fig. 2: An illustration of lateral and longitudinal Fig. 3: Mutual inductance between primary and Fig. 4: The amount of transferred power as a
misalignments between primary and secondary secondary coils as a function of LNM for different function of LNM for different LTMs.
coils. LTMs.

quality factor can vary between 3.84 and 6.38 which allows for
a reduction of power flow up to 40% by controlling the quality
factor.
The third option is to control the current I of the primary coil,
which allows a wide range of power transfer regulation. In this
section, three main challenges of DWPT are discussed and the
justification of current controller over the other controlling
methods is provided. These three issues include: i) LTM
between primary and secondary coils, ii) limited power rating
of the primary, and iii) transition from no-load to full- load
Fig. 5: Three-dimensional plot of mutual inductance as a function of LTM and
condition. For the sake of numerical calculations, I is assumed LNM.
to be 75 A, ωs is 2π×20 kHz, and QL is 6.38.
Misalignment: When the EV (secondary coil) passes over the
primary coil, mutual inductance M changes dramatically. M
depends on both LoNgitudinal Misalignment (LNM) and LTM
between the aforementioned coils. Fig. 2 illustrates the
secondary coil in three degrees of LTM approaching the
primary coil. Considering Fig. 2, normalized misalignments
between two coils can be defined as:

 x  x0
 x  R ,
Fig. 6: Three-dimensional plot of transferred power as a function of LTM and
LNM
(2) LNM.

  y  y0 ,
mutual inductance profiles.
LTM
 y R In DWPT systems, the time window to transfer power is
limited so it is essential to use the full system capability during
this period. In other words, while the vehicle is passing over the
where R is the radius of the primary coil, x0 and y0 are the
primary coil, it is desired to send the maximum amount of
coordinates of the center of the stationary primary coil, and x
energy to the secondary. As it is shown in Fig. 4, the power
and y are the center coordinates of the moving secondary coil.
transfer capability of the system drops when having LTM≠0
For the analysis in this section, the actual dimension of the
causing the amount of transferred energy to decrease, as well.
experimental setup (R = 35 cm) is used. LNM originates from
Assuming a constant vehicle speed, the transferred energy to
the nature of dynamic charging and models the forward motion
the secondary can be calculated for each LTM by integrating
of EV. However, the LTM is a Random Variable (RV); it is
the curves in Fig. 4. In Fig. 7, transferred energy for a wide
related to driver’s driving experience and habits and describes
range of LTMs and the vehicle speed of 50 km/h is shown. As
how aligned the EV is while passing over the primary coil.
one can see, having even 10 cm of LTM reduces the amount of
Fig. 3 shows mutual inductance between two coils as a
transferred energy by 22 %. As it was mentioned before, LTM
function of LNM for four different LTMs. As the degree of
is an RV and it is best described through its Probability Density
LTM increases, the mutual inductance profile drops which
Function (PDF).
results in the reduction of power transfer capability, as shown
In [30], the authors investigated the lateral position of
in Fig. 4. The 3D plots of mutual inductance and transferred
vehicles in a targeted lane for different driving conditions. The
power as a function of LTM and LNM are shown in Fig. 5 and
PDF for such RV is given as (3), where SD is the standard
Fig. 6, respectively. The curves are derived for circular coils,
deviation and y is the LTM. Fig. 8 shows the final probability
but the analysis can be easily generalized to any type of coil and

2168-6777 (c) 2017 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.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/JESTPE.2017.2761763, IEEE Journal
of Emerging and Selected Topics in Power Electronics
4

700 0.02 Driver is not aware

Driver is aware 110 with controller
600
without controller
500 0.015 100
600
e (J)

400 90

PDF

I* (A)
0.01
300 400

e (J)
80
200
0.005 70 200
100
0 60 0
0
-70 -50 -30 -10 10 30 50 70 0 0.2 0.4 0.6 0.8 1 -2 -1 0 1 2
-150-100 -50 0 50 100 150
|LTM| (Normalized)
LTM (cm) LTM (cm) LTM (Normalized)
Fig. 7: Transferred energy as a Fig. 8: PDF of vehicle’s LTM at the Fig. 9: Primary coil current Fig. 10: Comparison of transferred
function of LTM for the vehicle speed of 50 km/h. reference values as a function of energy with and without the current
speed of 50 km/h. LTM. controller.

density functions when the drivers are aware of testing (SD is 26 energy can be transferred (680 J). For LTMs greater than 43%,
cm) and when they are not aware of that (SD is 46 cm) [30]. In the transferred energy is also improved but it is less than the
both cases, the vehicle speed is 50 km/h. peak value (680 J). From (4), the expected value for the
transferred energy with the current controller is 410 J, showing
a 30 % increase.
y2
 Primary Power Ratings: When multiple receivers are
1
f ( y) 
2
2 sD
e (3) supplied simultaneously, it is possible that the primary side
sD 2
power exceeds its rating and could eventually damage the
system. Instead of completely terminating power transfer, in
Let us assume that one primary coil has been installed at the that case, a power controller is proposed to monitor the primary
center of a lane and that the drivers are aware of that. Now, it is power and when the rated power is exceeded, it reduces the
possible to calculate the expected value for energy transfer at primary current to maintain the power at its rated value.
each desired speed as follows: Transition from no-load to full-load condition: The open

loop control is typically unable to maintain a constant current
 e(Y )   e( y) f ( y)dy (4)
in the primary coil for the full range of load variation from no-

load to full-load. For example, in the existing system, in order
to maintain constant primary coil current from no-load to full-
where Y is the random variable representing LTM, e(y) is the load conditions, it is necessary to vary the phase shift angle of
function modeling transferred energy as shown in Fig. 7, and the inverter up to 15%. It is also worth mentioning that the metal
f(y) is the PDF as in Fig. 8. For the speed of 50 km/h and using body of the vehicle and the reflected reactance from the
(4) the expected value for transferred energy would be 316 J. It parallel-compensated secondary would detune the primary
is inferred that even though we could have had 618 J transferred resonant circuit, causing variation of the primary current.
to the secondary, due to LTM, the expected value has dropped Therefore, to cope with the dynamics change of the
by 54%. environment and operating conditions, a current controller
As it was mentioned before, through controlling the RMS needs to be applied to dynamically control the inverter’s phase
value of primary current, one can control the amount of power shift angle and provide constant primary current.
transferred to the secondary. The nominal value of the primary
current is selected such that the aligned vehicle can receive III. MODELING PRIMARY SIDE OF THE WPT SYSTEM
nominal power. In Fig. 7, it was assumed that the nominal In order to design proper controllers for the primary of the
primary current is Inom=75 A, but if the current can be controlled WPT system, an accurate model of the WPT system is needed.
and increased up to Imax=100 A, it would be possible to In this section, the primary side model is derived using GSSA
compensate for LTM and consequently, transfer more energy. and then verified through simulations and experimental tests.
Selection of the nominal value is a tradeoff between power Finally, the proposed controller structure is presented.
“offered” to the secondary and losses in the primary.
A. System description
Additionally, the ability of the secondary side controller to
reduce the power is limited; therefore, if the primary current is Fig. 11 shows the equivalent circuit of the entire WPT
too high and the vehicle is completely aligned, then the system. In this section, the primary side dynamics is studied.
transferred power could exceed the power rating of the Fig. 12 shows the schematic of the primary side containing a
secondary. From Fig. 4 and (1), one can calculate how much full-bridge inverter, an LCC compensation tank, and a primary
current increase is needed to transfer the same amount of energy coil, all of them supplied from a constant DC source. Using the
while having various degrees of LTM. Fig. 9 presents the full-bridge inverter, the constant DC voltage is inverted to an
reference current at different LTMs. The current is limited to AC square-wave voltage whose fundamental harmonic content
Imax=100 A since the ratings of our experimental system do not is controlled by means of phase-shift control of the inverter. The
allow for more current. Fig. 10 shows the effect of having such
a controller where for LTMs less than 43% the same amount of

2168-6777 (c) 2017 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.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/JESTPE.2017.2761763, IEEE Journal
of Emerging and Selected Topics in Power Electronics
5

Idc IS1 VCs1 VCs2 IS2 Ibat

RLdc RS1 RS2
Ldc LS1 IPri ISec LS2 LB
CS1 CS2
LPri LSec Rbat
Cdc VCp1 CP1 VCp2 CP2 CB
VDC VCB
Vbat
RCdc RP1 RL1 M RL2 Rp2

DC Grid Full-Bridge Primary Secondary Full-Bridge Buck Vehicle

Inverter Comp. Tank Comp. Tank Rectifier Converter Battery
Fig. 11: The equivalent circuit of the entire WPT system.

Idc LCC Comp. Tank series inductance and CP1 and CS1 are parallel and series
IPri capacitances at the primary side, respectively. LPri models the
LS1 CS1 inductance of the primary coil and Vinv,f is the fundamental
Primary
CP1 Coil
harmonic of the inverter output voltage. In order to have an
VDC Vinv accurate model, the parasitic resistances of the components are
taken into account. RS1 and RP1 are Equivalent Series
Zr=Rr+jX r
Resistances (ESRs) of LS1 and CP1, respectively. All
Full-Bridge compensation and coil parameters are defined for the secondary
Inverter side similarly. Furthermore, Rac shown in Fig. 14 (b) is the
resistance seen from the input of the full-bridge rectifier and it
Fig. 12: The primary of the WPT system.
depends on the power delivered to the battery and duty cycle of
LCC compensation is employed between the output of the full- the Buck converter. M is the mutual inductance between the
bridge inverter and the primary coil to provide a sinusoidal coil primary and secondary coils. RL1 combines and models three
current. In Fig. 12, the effect of the secondary is represented by resistive components:
the reflected impedance Zr. The resistive part of Zr models the
active power transfer, while the reactive part quantifies the RL1  RLpri  RCs1  Rr (5)
uncompensated part of the receiver reactance.
Fig. 13 shows the block diagram of the Dual-Loop Controller
(DLC). As one can see, the reference value of the primary coil where RLpri is the ESR of the primary coil, RCs1 is the parasitic
current I* can be determined in two ways, based on the value of resistance of the series capacitance, and Rr is the reflected
power drawn from the DC link (PDC). If PDC is less than the resistance of the secondary at the primary. Unlike RL1, RL2 at the
reference power P*DC, then EVDS is the one that determines I*. secondary does not include Rr, but only the parasitic resistance
In this case, if LTM is zero, I* would be Inom=75 A. However, if of the coil and series capacitor.
the lateral misalignment is present (LTM≠0), then the value of GSSA method is based on complex Fourier series
I* will be determined based on the look-up table obtained from representation of a waveform. Due to the band-pass filtering
Fig. 9. The second case would be when PDC is greater than P*DC, effect of the compensation tank, the harmonics of current Is are
which shows that the power source in the primary is being filtered, allowing only the first harmonic of the coil current to
overloaded. In this condition, the power controller decreases I* be considered for modeling. Vinv is the output voltage of the full-
in order to keep PDC at P*DC. P*DC is compared with PDC and the bridge inverter and it is illustrated in Fig. 15 for two different
error is passed to the Power Controller module. Based on the values of the phase-shift control signal θ. By controlling the
power error value and the primary coil current IPri, this module width of Vinv, the magnitude of its fundamental harmonic (Vinv,f)
determines I*. The Current Controller module employs the error is controlled. The relationship between the width of Vinv (θ) and
between I* and the measured IPri and computes the reference the RMS value of its first harmonic is shown as:
value of the inverter output voltage. Two steady-state SS 4VDC 
models are shown in Fig. 13. The first one is Giv(s) which relates Vm  sin( ) (6)
2 2
the RMS value of IPri to the fundamental harmonic of the
inverter output voltage Vinv,f. The second one is Gpi(s) that
where VDC is the DC source voltage. The fundamental
relates PDC to IPri.
harmonic of inverter output voltage can be described as:
B. GSSA modeling
GSSA method is proposed in [31] for modeling resonant vinv, f (t )  2Vm sin(t ) (7)
converters. In [26], a low power LCL-T circuit for WPT is
modeled using GSSA. Here, a similar approach is adopted to In Appendix I, the entire system shown in Fig. 11 is modeled
derive the two SS models mentioned above. and A matrix needed for modeling the primary side in no-load
1) Deriving Giv(s) model and loaded conditions is presented. Fig. 16 compares the open
Fig. 14 (a) and (b) show the model of the LCC resonant tank loop gains of the primary system at no-load and loaded
and the primary coil in no-load and loaded conditions. LS1 is the conditions. From this figure, one can see higher dynamics of the

2168-6777 (c) 2017 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.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/JESTPE.2017.2761763, IEEE Journal
of Emerging and Selected Topics in Power Electronics
6

Look up table
EVDS I* SS model of SS model of
signal LCC & Primary coil LCC & DC link
Yes Vinv,f IPri,RMS PDC
PDC  PDC
* I* Err Current
Controller
Giv (s) Gpi (s)
No
* Err Power I* IPri
PDC
Controller
IPri
PDC

Fig. 13: The block diagram of the dual loop controller.

IS1 VCs1
RS1  I DC I S1
LS1 G pi ( s)  RL1VDC  I  V  k  G1 ( s)  G2 ( s)
CS1
LPri IPri  S1 Load

k  R V , G ( s)  I DC  s  , I s
Vinv,f VCp1 CP1 (10)
G2 ( s)  S1
I S1  s  VLoad  s 
RP1 RL1  L1 DC 1

(a)
IS1 VCs1 VCs2 IS2 Considering (10), Gpi(s) consists of three parts; first is the
RS1 RS2
LS1 IPri ISec LS2
constant part RL1VDC. The second part is G1(s), which represents
CS1 CS2
LPri LSec the relation between IDC and IS1. Considering Fig. 17, one can
VCp1 CP1 VCp2 CP2 Rac
Vinv,f derive that:
RP1 RL1 M RL2 Rp2

I DC 2 2 
Primary  sin  G1  s  (11)
Comp. Tank
Secondary
Comp. Tank
I S1  2
(b)
Regarding G2(s), it resembles the model in Fig. 14 (a), with
Fig. 14: The model of the LCC resonant tank and the primary coil in (a) No-
load condition, (b) Loaded condition. the exception that the input voltage and the output current have
exchanged their positions. This is illustrated in Fig. 18 (a) and
loaded system, which is the reflection of the secondary (b). According to reciprocity theorem, one can see that G2(s)
dynamics at the primary. However, for the design of the and Giv(s) are then two identical transfer functions.
controller and the anticipated bandwidth being significantly
below 1 kHz, the difference in the high-frequency dynamics is C. Model verification
mostly irrelevant. Therefore, one can infer that for designing In order to verify the accuracy of the developed analytical
the controller, the model of the no-load primary system shown model for Fig. 14, the step response of the SS model is
in Fig. 14 (a) is sufficient. compared with the response of the experimental system at two
2) Deriving Gpi(s) model conditions, namely no-load and loaded condition at 15 kW. To
In this part, the objective is to find the effect of a variable load this aim, a 112° to 120° step change of phase-shift control signal
on the power drawn from the DC source. To do so, three θ is applied to the system. Fig. 19 shows how changing θ would
assumptions are adopted: i) the effect of the secondary on the change RMS value of the first harmonic, used as the input of
primary is modeled as a variable voltage source (VLoad), ii) the the SS model.
full-bridge inverter is modeled as two dependent voltage and In the experimental system, the inverter’s IGBT switches and
current sources, and iii) all the directions for currents and the DC link capacitor demonstrate significant damping effect
voltages are considered with the premise that the input is VLoad on the step response, which has not been modeled in Fig. 14.
and the output is PDC, as shown in Fig. 17. The final goal is to To correct for this additional damping, Rs1 in Fig. 14 (a) is
find Gpi(s): increased by 0.7 Ω. In Fig. 14 (b), which represents the loaded
P s condition, M and Rac are set to M=37 µH and Rac=9 Ω. These
G pi ( s)  DC
I s (8) values are identified from experiment measurements at steady
Pri
state resulting in 15-kW load to the primary, and by using the
known values of the battery voltage and duty cycle of the Buck
Assuming a constant DC voltage source VDC and a constant converter. Fig. 20 compares the step response of the SS model
reflected resistance RL: with the experimental response in the two conditions.
Considering this figure, one can see close resemblance between
 PDC  VDC  I DC the two responses. The differences are mainly due to the
 (9)
VLoad  RL1  I Pri inverter dynamics and non-idealities of IGBT switches. In the
loaded condition, step response indicates more oscillations
and substituting (9) in (8) gives: which originate from the dynamics of the secondary reflected
to the primary. However, as it was shown in Section III.B, the

2168-6777 (c) 2017 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.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/JESTPE.2017.2761763, IEEE Journal
of Emerging and Selected Topics in Power Electronics
7

VDC 2Vm 2 IDC IS1 VCs1

RS1
2Vm1 vinv,f LS1 CS1 LPri
vinv IPri
θ1 VDC VCp1 CP1
s(t) Is1 s(t) VDC RP1
θ2 vLoad

T/2=π/ω
Fig. 17: The primary side model consisting of the DC power supply, the full-
Fig. 15: The inverter output voltage and its fundamental harmonic for two bridge converter, the LCC resonant tank, the primary coil, and the secondary
different values of the phase-shift angle θ. reflected impedance.

IPri IS
LS CS LS CS
LPri LPri
A CP C A CP C

V Vinv,f I I VLoad V

B D BD
Network Network
(a) (b)
Fig. 16: Magnitude Bode plots of the open loop system in loaded and no-load Fig. 18: Illustration of reciprocity theorem: (a) Giv(s)=IPri(s)/Vinv,f(s),
conditions. (b) G2(s)=IS(s)/VLoad.
dynamics of the secondary does not affect the system loop gain
for the frequency range of interest. Thus, the system model Appendix I and after linearizing the system around θ0=120°
derived for no-load condition will be used for the current operating point and nominal load conditions of 25 kW.
controller design below. Equation (6) is used to relate the calculated large-signal value
of the phase-shift angle and the first harmonic of the inverter
IV. CONTROLLER DESIGN output voltage. The nominal power (25 kW) at the nominal
In this section, the control algorithm is explained and primary coil current (75A) is achieved when the reflected
controllers are designed. The structure of DLC is shown in Fig. resistance is Rr=25,000/752=4.44 Ω. Based on the transfer
13. The outer loop controller (regulates the DC power) should function and targeted 400 Hz bandwidth, the linear controller
be fast enough to generate current reference value under the has been designed:
dynamic load changes. The time profile of the variable load is
similar to the power profile given in Fig. 4. For the DWPT 1.69 104
window of 126 cm, and the vehicle speed of 50 km/h, the Gc1 
s (12)
s(1  )
variable load would have a fundamental harmonic of 11 Hz. 2 1500
The spectral analysis of the variable load shows that the first
four harmonics are enough to provide an authentic load profile Fig. 22 shows the magnitude and phase Bode plots of the
representation. Therefore, 80 Hz crossover frequency is compensated and uncompensated loop gains. It is seen that the
selected for the power PI controller. The inner current loop compensated loop gain has a crossover frequency of 400 Hz,
should be significantly faster in order to finalize the current the phase margin is 80 degrees, and the gain margin is 16 dB.
change before a new current reference is generated. Therefore,
the current loop is designed to be five times faster with the B. Power Controller
crossover frequency of 400 Hz. The block diagram of the power controller is presented in Fig.
23. The power controller is only needed when the measured DC
A. Current Controller
power PDC exceeds the maximum allowed level P*DC. It
Fig. 21 shows the block diagram of the large-signal model of prevents further power increase by decreasing IPri. The
the proposed current controller. The time sequence of controller operation is terminated when the measured power
controlling primary current has two distinct phases: i) the coil- naturally drops below P*DC. When PDC is less than P*DC, I* is
energizing phase and ii) the current regulating phase. When the specified by the EVDS block (75 A for the aligned vehicle).
vehicle is detected, the start-up procedure is activated and the To cope with the load dynamics, the bulk of the current
coil is energized by increasing the inverter’s phase-shift angle reference I* is generated through an adaptive proportional
linearly from zero to 120°. This process takes approximately 5 regulator that relates the reference primary coil current and its
ms and by the end of that interval, IPri would be around 71 A. measured (actual) value:
When 120° is reached, DLC is activated to adjust and control
the coil current at I* which is specified by the outer control loop *
PDC
as depicted in Fig. 13. I *  kI Pri  I Pri (13)
In Section III.B, an SS model has been developed for the inner PDC
current loop. Using that model and system parameters listed in
Table I, the transfer function Giv(s) is obtained. The transfer
function is derived from the state space matrices developed in

2168-6777 (c) 2017 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.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/JESTPE.2017.2761763, IEEE Journal
of Emerging and Selected Topics in Power Electronics
8

Vinv 1st Har. 73 68

800 500
112° 120° Model
72

1st harmonic (V)

Experiment 66

IPri (RMS) (A)

IPri (RMS)(A)
400 71
450
Vinv(V)

50 µs 70 64
0
69
400 62 Experiment
-400 68 Model
120° 67 60
-800 350 0 1 2
0 0.5 1 1.5 2 2.5
0 25 50 75 100 125 Time (ms) Time (ms)
Time (ms)
(a) (b)
Fig. 19: Variation of the first harmonic of the Fig. 20: The step responses of the experimental system and its model, (a) no-load condition, (b) loaded
inverter output voltage due to a step change in condition at 15 kW.
phase-shift control signal θ, from 112° to 120°.

θ1
Start-up

IPri

I* Err Controller θ2 θ
Gc1(s)
Eq. (6) Vinv,f
Fig. 22: Magnitude Bode plots of uncompensated and compensated current
Fig. 21: The block diagram of the current controller large-signal model. loop gains.
Table I: System Parameters
Parameter Unit Value
I Pri
PDC I1*
Operating frequency kHz 20 Eq. (14)
*
DC source voltage V 600 PDC
Primary coil inductance, LPri μH 173.7
Primary coil ESR, RL mΩ 55 I 2*
Err Controller
Series inductance, LS μH 54.5 I*
Gc2(s)
Series inductance ESR, RS mΩ 38
Fig. 23: Block diagram of the power controller large-signal model.
Series capacitance, CS μF 0.51
Parallel capacitance, CP μF 1.22
Parallel capacitance ESR, RP mΩ 7

where k is an adaptive constant which is derived from equating

the DC and transferred power after neglecting the primary side
losses PLoss:

 PDC  Rr I Pri
2
 PLoss *
PDC
Fig. 24: Magnitude Bode plots of uncompensated and compensated power
loop gains.
 *  I *
 I Pri
 DC
P  Rr  I 
* 2
 P *
Loss
PDC
(14) phase margin of 80 degrees, and gain margin of 26 dB.
C. EVDS Operation
Fast dynamics of the proportional controller allows the The EVDS installed on the vehicle indicates the presence of
power controller to track PDC variation caused by the changes the EV and measures its degree of LTM. The EVDS consists of
in Rr(t) and prevents overloading of the primary. However, due a primary coil mounted underneath the vehicle and three
to neglected losses in (14), inaccurate regulation can cause secondary coils embedded just before the main primary coil
discrepancy between PDC and P*DC and in some cases small (LPri) in the road, which are shown in Fig. 25 (a) and (b),
system overloading. This overloading is critical if the system is respectively. The EVDS generates an analog voltage signal in
forced to operate for a long time as a stationary charger when the range of 0-5.1 V. When the vehicle is detected, the signal
the vehicle is stopped. To eliminate the steady-state error, an level is changed from inactive zero level to a corresponding
integral controller is designed using the system transfer value in the range of 0.1-5.1V. The EVDS output signal is
function Gpi(s) and targeted crossover frequency of 80 Hz: delivered to the DLC where it is interpreted as shown in Fig.
26. The output voltage of 2.6 V indicates complete alignment
2.16 (LTM=0), while the voltages above or below 2.6 V indicate
Gc 2  (15) misalignment to the left or right directions, respectively. The
s
Fig. 24 shows the compensated and uncompensated loop
gains of the controller. From this figure, one can see that the
compensated loop gain has the crossover frequency of 80 Hz,

2168-6777 (c) 2017 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.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/JESTPE.2017.2761763, IEEE Journal
of Emerging and Selected Topics in Power Electronics
9

LTM (Norm.)
0

-1
0 1 2 3 4 5 6
(a) (b) Voltage (V)
Fig. 25: EVDS structure: (a) primary coil mounted underneath the bus, (b) three detection coils Fig. 26: The relationship between EVDS output
embedded in the road. voltage and the degree of LTM.

EVDS is able to accurately measure LTM in the range +/- 30 150

cm. Using the measured LTM, the DCL block sets the primary 50

IPri (A)
current reference based on the current profile given in Fig. 9. -50
For the voltage of 2.6 V, the reference value is 75 A. As the
-150
degree of LTM increases, the current will be raised up to the 0 20 40 60 80 100
maximum value of 100 A to compensate for undesired LTM Time (ms)
(a)
impact.
80

IPri (RMS) (A)

60 Phase 2 Phase 2
V. SIMULATIONS 40 Phase 3
A single-coil DWPT system is designed and the 20
Phase 1 Phase 4
characteristics of the system are tested through simulations in 0
0 20 40 60 80 100
Matlab/Simulink. The topology of the simulated system is Time (ms)
presented in Fig. 12 and the system parameters are specified in (b)
Table I. The impact of the secondary is modeled through the 160
120
variable load profile Rr(t). The vehicle speed is assumed to be

θ (deg.)
80
50 km/h and the power profile presented in Fig. 4 is adopted for 40
these tests. The reference value for primary current is assumed 0
to be I*=75 A and the primary power limit is set to P*DC=20 kW, 0 20 40
Time (ms)
60 80 100
which is less than the maximum power (25 kW) that the system (c)
can transfer for perfect alignment. 6
Fig. 27 (a) and (b) present the primary coil current and its 4
RL (Ω)

RMS value, respectively. Fig. 27 (c) depicts how the phase-shift

angle is being varied to get the desired primary coil current, and 2

part (d) shows the time profile of the load. Fig. 27 (e) shows the 0
DC power which is drawn from DC source. Four phases are 0 20 40 60 80 100
Time (ms)
identified and labeled in Fig. 27 (b). During Phase 1, the phase- (d)
shift angle is increased linearly up to 120 degrees ramping the 30
primary coil current to 71 A. Phase 1 is followed by a 2 ms
PDC (kW)

20
delay to stabilize the current level. In Phase 2, DLC is activated
to control the current at 75 A. During this phase, the vehicle 10
would approach the primary pad and the power transfer would 0
start. At some point during Phase 2, the primary DC power 0 20 40 60 80 100
Time (ms)
reaches the 20-kW limit, which activates the power loop of
(e)
DLC to regulate the power at this value, as is shown in Fig. 27
(d). In phase 4, the DWPT window is over, so the controller Fig. 27: Simulink simulations of the DLC operation: a) primary coil current,
reduces the phase-shift angle until the current is zero. b) the RMS value of primary coil current, c) the inverter phase-shift angle, d)
time profile of the load e) power drawn from the DC source.
The simulation results confirm that the proposed DLC is able
to control both current and power with a good dynamics for the partially shared compensation tank. Two controlled contactors
EV speed of 50 km/h. In Section VI, the proposed control (SW1 and SW2) are used to connect the right coil and the
algorithm is implemented to control the charging of an electric unshared part of the compensation tank in a timely manner.
bus from two primary pads embedded in the road. Secondly, the series inductance LS1 and series capacitance CS1
are split into two parts in order to provide a balanced voltage at
VI. EXPERIMENTAL VALIDATION AND OUTDOOR the coil. Ldc in this figure models the inductance of the 50 m line
DEMONSTRATION from the power supply in the building to the electric panels by
the test track.
A. System description
Fig. 29(a) and (b) show overview of the 400-m test track and
The proposed controller is implemented on the experimental the electric panels installed by the track, respectively. DC link
system shown in Fig. 28. This topology is similar to what is capacitor, full-bridge inverter, and the compensation tanks are
shown in Fig. 12 with two modifications. Firstly, two coils are mounted inside one of those panels, as shown in Fig. 30. A
sequentially energized from the same inverter through a

2168-6777 (c) 2017 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.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/JESTPE.2017.2761763, IEEE Journal
of Emerging and Selected Topics in Power Electronics
10

SW1 2×CS1

Ldc
2× S1

LS1 /2 SW2
Pri. Pri.
VDC Cdc
Coil2 Coil1
CP1 P1

2× S1
2×CS1

Full-Bridge LS1 /2
Inverter
(a) (b)
Fig. 28: Schematic of the experimental system primary side. Fig. 29: (a) Overview of the test track, (b) electric panels installed by the track.

(a) (b)
Fig. 30: DC link capacitor, full-bridge inverter, Fig. 31: (a) Structure of the primary coil with the radius of 35 cm, (b) Position of two primary pads
and the compensation tanks mounted in the panel. embedded in the road.

circular coil structure with a radius of 35 cm has been adopted Considering Fig. 33 and using (16), the energy efficiency of
for the primary pads as shown in Fig. 31 a). Fig. 31 b) shows the system is calculated as 86% for the entire DWPT window
the position of two primary pads where one coil is inserted and LTM=0. However, the energy efficiency decreases for
beneath the fiberglass grates and the other coil is embedded in misaligned vehicles, as shown in Fig. 35. Using (4) and the PDF
the concrete. The secondary pad is attached under the bus and in Fig. 8, the expected value for energy efficiency is calculated
secondary compensation tank and buck converter are installed to be 72%.
at the back of the bus, as shown in Fig. 32 a) and b), From Fig. 34, it is seen that, as the EV approaches the central
respectively. positions, power efficiency increases and there is a zone of 30
cm where the power efficiency is over 90%. In order to increase
B. Efficiency test
energy efficiency, one can use a portion of DWPT window
The efficiency of the system for the full range of around the center while sacrificing the energy outside that zone.
misalignment is measured assuming the DC voltage source as Table II shows how reducing the active portion of DWPT
the input and the battery terminals as the output. For this test, window would increase the energy efficiency and how it affects
the primary coil current was maintained at 75 A, the secondary the energy transfer capability for the specific coils used in this
side quality factor was set to 3.84, the electric bus was laterally research. Although the values presented in the table would be
aligned, and the longitudinal alignment was changed with different for different types or shapes of WPT coils, it clearly
increments of 1.26 cm. Fig. 33 shows the power drawn from the demonstrates that even in DWPT systems, high energy
DC source and the power delivered to the battery. At the peak efficiency can be achieved without sacrificing too much of the
value, more than 23 kW is drawn from the DC source. energy transfer capability. In this research, 75% of DWPT
Using the values in Fig. 33, the system power efficiency is window has been adopted and used.
calculated and plotted in Fig. 34 as a function of LNM for three During the interval of strong coupling between two coils, the
different LTM values. It can be seen that the power efficiency coil losses (at the primary and the secondary) dominate the loss
changes for a DWPT window from very low values up to 95% model. However, if the primary is energized before the full
for aligned positions. It is also worth mentioning that as the coupling is reached, an additional loss component appears,
LTM increases, the system efficiency decreases. In a DWPT caused by eddy currents induced in the vehicle chassis and
system, the energy efficiency is more important than the power body. Furthermore, the vehicle metal surface impacts the tuning
efficiency, since it is calculated over the entire DWPT window conditions at the primary by changing the self-inductance of the
and allows comparison among different DWPT systems: primary coil. The text below sheds some light on these two
Table II: The effect of using a portion of DWPT window on system energy
 Pbat (t )
 P  P (t ) Power Efficiency efficiency and energy delivered
 DC
 Description Portion of Energy Delivered
  Pbat (t )dt (16) DWPT window efficiency energy
  DWPT window Energy Efficiency Entire DWPT window 100% 86% 100%
 E
  PDC (t )dt Null-to-null window 52% 91% 88%
 DWPT window ηp > 80% window 39% 93% 85%
ηp > 90% window 24% 94% 70%

2168-6777 (c) 2017 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.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/JESTPE.2017.2761763, IEEE Journal
of Emerging and Selected Topics in Power Electronics
11

LTM=0 cm LTM=10 cm
25 LTM=15 cm
Pdc 100
20 Pbat
80

Power (kW)
15
60

ηP (%)
10 40

5 20
(a) (b) 0
0
-70 -50 -30 -10 10 30 50 70
-70 -50 -30 -10 10 30 50 70
LNM (cm) LNM (cm)
Fig. 32: (a) Secondary pad installed under the bus, (b) secondary Fig. 33: The power drawn from the DC Fig. 34: System power efficiency as a
compensation tank and the buck converter installed at the back of source and delivered to the battery. function of LNM.
the bus.

100 30 800 184

Pri. coil is energized
600 182
80 but there is no

Loss (W)
20
Idc (A)

180
ηE (%)

LPri (µH)
coupling between Pri. Coil+Vehicle Body
60 coils. 400 178
Vehicle Body
10
40 200 176
174
20 0 0
0 100 200 300 0 40 80 120 160 172
0 20 40 60
LTM (cm) Time (ms) Time (ms) 0 30 60 90 120
Distance (cm)
(a) (b)
Fig. 35: Energy efficiency of the Fig. 36: Losses in the system: (a) DC-Link current as the vehicle approaches Fig. 37: FEM results for the self-
system as a function of LTM. to and passes over the primary coil, (b) Losses in the primary coil and the inductance of the primary coil as the
vehicle metal structure for the shaded zone shown in (a). vehicle approaches to the primary
pad.
phenomena. vehicle approaches to the primary coil. The simulation
Fig. 36 (a) depicts the profile of DC-link current as the considers the metal body and chassis being solely made of steel.
vehicle approaches the primary coil. After 160 ms in the figure, Distance axis begins where the front of the vehicle is at the edge
the system starts transferring some noticeable power, and from of the primary pad and finishes when the vehicle has the most
that point forward, it would not be possible to distinguish coverage over the primary pad. Being a ferromagnetic material,
between the primary losses, eddy current losses inside the steel increases the value of LPri. The increase is about 5%,
vehicle metal structure, and secondary losses. However, in the which has a negligible impact on the operation of the primary
shaded zone shown in Fig. 36 (a), the primary coil is energized compensation tank and the controller design.
but the mutual inductance between primary and secondary coils
C. DWPT algorithm for EV charging
is too low for some significant power transfer. During that
interval, the system generates losses in the primary coil as well Fig. 38 shows the high-level flowchart of the DWPT
as in the vehicle chassis and body. Fig. 36 (b) shows the algorithm. With a general segmented-coil system at the
experimentally measured losses in the vehicle metal structure primary, the coil nearest the vehicle is connected and ready to
as the vehicle approaches the primary pad. It can be observed be energized first. The controlling system reads the EVDS
that this loss is about 150 W at the maximum value, reached signal constantly, and when it is greater than the specified
when the pad is completely covered by the vehicle metal body. threshold, it is interpreted as an approaching EV capable of
Comparing the eddy current losses and the rated power (25 receiving power. Based on EVDS misalignment signal the
kW), the vehicle body loss constitutes only 0.6 % at its reference value for the current is determined. Furthermore, a
maximum, which seems insignificant. However, one should speed counter starts counting for EV speed estimation. The
keep in mind that eddy current loss could exist much longer if controller increases phase shift angle up to 120° linearly and
the primary is energized all the time, since the vehicle can be when the phase reaches 120°, after 2-ms delay, the DLC is
significantly longer than the receiving pad. That elevates the activated to adjust the current to the reference value. During the
significance of eddy current loss in terms of energy loss and DWPT window, the DLC controls current and power, and when
further reiterates the importance of the strict control over the the amount of transferred power becomes zero, it indicates that
charger activation and deactivation time with respect to the the EV has passed over the first coil and that the coil should be
vehicle position. Considering the size of the vehicle body and de-energized. Using the time counter, EV speed is estimated
its heat dissipation potential, energy dissipated inside the and the time delay for energizing the next coil is calculated. As
vehicle body does not create a significant threat to the onboard a result, each primary coil is energized just prior to the EV
equipment or the people inside. reaching it. The same procedure is followed until all the primary
The chassis and the body of the vehicle also affect the system coils are energized successively.
by changing the self-inductance LPri of the primary coil. Fig. 37 D. Start-up procedure
shows the Finite Element Modeling (FEM) results for LPri as the Fig. 39 (a) shows the experimentally measured start-up

2168-6777 (c) 2017 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.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/JESTPE.2017.2761763, IEEE Journal
of Emerging and Selected Topics in Power Electronics
12

Start sent 2.9 V and 3.3 V signals which were interpreted as 8% and
20% misalignments, respectively (see Fig. 40). Based on the
graph in Fig. 9, the controller set the current reference values to
Connect 1st coil Turn off the coil
75.6 A and 79.2 A, respectively. Fig. 41 shows the resulting
RMS values of the primary coil current for these two tests. It is
seen that the controller is able to maintain the primary current
at the reference value under no-load and loaded conditions.
No Yes Is it the Here, energy drawn from the DC source is calculated. To this
EVDS voltage> Vth final coil?
aim, the DC source current is multiplied by the DC source
Yes No voltage (600 V) and then integrated over the DWPT window.
Calculated energies between null points for two tests are 1.29
Connect the next coil kJ and 1.11 kJ, respectively. Unfortunately, it was not possible
Energize the coil
to replicate speed during the two tests. However, the amount of
transferred energy is inversely proportional to the vehicle
speed, hence assuming that the null points occur at the same
Wait until the bus
reaches the coil longitudinal EV position, the time intervals shown in Fig. 40
DWPT completed?
No can be used as vehicle speed indicators. From Fig. 40 a) and b),
it is seen that the ratio between the periods matches the ratio
Yes between energies, which indirectly validates the LTM
Fig. 38: Flowchart of the DWPT algorithm in a segmented-coil system. compensation concept and implementation.
procedure (the primary coil current and the inverter output F. Segmented-coil operation
voltage) with a section of it being magnified and shown in Fig. Fig. 42 (a) and (b) show the operation of a two-coil DWPT
39 (b). In Fig. 39 (c), the RMS value of the primary coil current system tested at two different speeds, 22 km/h and 50 km/h. As
is shown, and its distinct intervals are emphasized: an initial described in Section VI.A, the controller is designed to energize
period of linear current ramping, a 2 ms delay, and then multiple primary coils successively based on the vehicle
activation of the DLC. It can be seen that the whole process position. Considering Fig. 42, it is seen that as soon as the
takes around 14 ms, though the current reaches 90% of its final EVDS sends its signal to the controller, the first coil is
value in less the 7 ms. Assuming an EV speed of 100 km/h, the energized and it stays in that state until the DWPT window is
primary coil should be energized when the secondary coil is at over. During the transition interval, when both coils are de-
least 40 cm away from the primary coil. Due to the limitations energized, the controller connects the second coil. Based on the
of IGBT switches and driver, the response of the inverter output signal from the EVDS and the turn-off moment of the first coil,
voltage is not linear at the beginning of the start-up procedure, the controller estimates the bus speed and uses the estimated
but that does not affect the overall process. speed to predict the right moment for energizing the next coil.
E. LTM compensation As long as the speed does not change significantly between two
Key roles of the DLC are to maintain the primary current at pads, this approach is capable of predicting activation moments
the reference value and compensate for LTM. Fig. 40 (a) and for a series of coils. It can be observed that regardless of the bus
(b) show primary coil current, DC source current, and the speed, the controller always energizes the second coil at the
EVDS signal recorded during two DWPT tests. In both tests, same longitudinal position. Furthermore, by using an additional
the bus speed was between 20 and 25 km/h. The current drawn sensor mounted in front of the EVDS unit, one can estimate the
from the DC source indicates how much power is transferred to vehicle speed even for the first coil.
the EV augmented by the primary power losses. When EVDS From Fig. 42 (a) and (b) one can recognize that the speed
detects the bus, it sends a 10 ms pulse to the controller to does not affect the power profile. This implies that the induced
indicate the presence of the EV and the degree of LTM. The voltage in the secondary is dominantly caused by the
controller reads this signal and, after a 20-ms delay, energizes transformer effect between the two coils, while the motion-
the primary coil with a reference current proportional to the induced component of the voltage can be neglected. As it was
measured LTM. In the two tests presented in Fig. 40, the EVDS mentioned in Section VI.A, the primary pads are not identically

IPri (A) 80
Linear
IPri (A)
IPri (RMS) (A)

60 ramp up
DLC is activated
40
2 ms
Vinv (V) delay
Vinv (V) 20

0
0 5 10 15 20
Time (ms)
(a) (b) (c)
Fig. 39: Experimentally measured start-up waveforms: (a) primary coil current and inverter output voltage, (b) colored part shown in (a), (c) the RMS value of the
primary coil current.

2168-6777 (c) 2017 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.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/JESTPE.2017.2761763, IEEE Journal
of Emerging and Selected Topics in Power Electronics
13

LTM=8% LTM=20%
80

IPri (RMS) (A)

60
40 Iref=75.6 Iref=79.2

20
0
0 50 100 150 200 250
Time (ms)
(a) (b)
Fig. 40: LTM compensation – experimental waveforms: primary coil current, DC source current, EVDS Fig. 41: RMS value of the primary coil current for
signal for (a) normalized LTM of 8%, (b) normalized LTM of 20%. two LTMs, namely 8% and 20%.
embedded in the road so the amount of transferred power in Fig.
42 is different between them.
G. Power control
Finally, the last task of the DLC is to control DC power when
it exceeds the primary power rating. In order to demonstrate the
operation of the power loop controller, the controller power
limit is set to 18 kW and the response is tested for three different (a)
bus speeds, namely 11, 23, and 30 km/h. Fig. 43(a) – (c) show
the DC link current and primary coil current. The DC link
current is used to represent the input DC power since the DC
link voltage is constant. It is seen that when the DC power goes
over the 18 kW limit (30 A of DC current), the controller
reduces the primary coil current in order to keep the power at
its reference value. As the vehicle speed increases, the
controller needs to overcome higher dynamics and therefore it
(b)
faces higher overshoots. However, in all cases, the power
controller was able to reduce the DC power significantly. Fig. 42: Segmented-coil operation; (a) bus speed is 22 km/h, (b) bus speed is
50 km/h.
VII. CONCLUSION
In this paper, a comprehensive study of dynamic WPT is APPENDIX I
conducted and the results are implemented to the design a 25- The entire WPT system is shown in Fig. 11. In order to model
kW segmented-coil dynamic charging system for an electric the system, all the switching devices are assumed to be ideal, D
bus. This study takes into account stochastic elements of real is the duty cycle of the buck converter and θ is the phase shift
driving that impact the expected value of delivered energy to angle of the inverter. An existence function s(t) and a sign
the vehicles. The measurement of energy efficiency, rather than function sgn(is2) are employed to model the full-bridge inverter
the power efficiency, is introduced to quantify the performance and rectifier, respectively. The operation of these functions are
of the DWPT system. The GSSA method is employed to model given below:
the LCC compensation tank and primary inverter and
simulation and experimental tests were used to verify the 0 n  0, 2, 4 . . .

model. A dual-loop controller is designed to control the primary s(t )   2 j  
power and current. The energy efficiency of the system is
n
 n sin  2  n  1,3,5 . . . (17)
  
measured as high as 86% when there is no lateral misalignment,
2
sgn  x  1 
j x
with the possibility to be increased over 90% with some e 1

modification in coil energizing and de-energizing algorithms. A  (18)

method is proposed that uses the information from a
misalignment detection system to compensate for lateral The state-space equations can be defined as follows:
misalignment and equalizes the amount of delivered energy to
the vehicles. This correction resulted in a 30% increase in the
expected value of transferred energy. By using the proposed
controller, the primary can transfer constant energy with an
energy efficiency of over 80% for lateral misalignments
ranging from -15cm to +15 cm. An outdoor demonstration is
conducted and, in turn, proved the effectiveness of the proposed
dual-loop controller.

2168-6777 (c) 2017 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.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/JESTPE.2017.2761763, IEEE Journal
of Emerging and Selected Topics in Power Electronics
14

(a) (b) (c)

Fig. 43: Operation of the power loop controller when the primary power limit is set to 18 kW: (a) bus speed of 11 km/h, (b) bus speed of 23 km/h, (c) bus speed
of 30 km/h.
 diS 1 RP1  RS 1  r1 RS 2  RP 2  r4
 LS 1 dt  s (t )vCdc  vCp1  RP1 (iS 1  iPri )  RS 1iS 1
 RP1  RL1  r2 RLdc  RCdc  r5
 L diPri  R (i  i )  v  R i  v  M diSec
 Pri dt P1 S 1 Pri Cp1 L1 Pri Cs1 RP 2  RL 2  r3 det  LPri LSec  M 2
dt

 2
(21)
sin 
4
C dvCp1  i  i k
 P1 dt S1 Pri 

C dvCs1  i
 S 1 dt Pri As it can be seen, A matrix is 20×20 with state variables
 defined in (20). In order to model the primary system in no-load
 L diS 2   sgn  i  v  v  R (i  i )  R i
 S 2 dt S2 CB Cp 2 P 2 Sec S2 S2 S2 condition only first eight variables are needed which reduces A
 diSec di matrix to 8×8. For modeling in the loaded condition in this
L   RP 2 (iSec  iS 2 )  vCp 2  RL 2 iSec  vCs 2  M Pri paper, the dynamic of the DC power supply is neglected and
 Sec dt dt
 only the first 16 variables were needed, reducing the size of
C dv
 iSec  iS 2
Cp 2
(19) matrix A to 16×16. The B matrix is 20×2 where two cells are
 P 2 dt
 not zero and are defined as follows:
C dvCs 2  i
 S 2 dt Sec
1 1
 B 18, 2   , B  20,1  
Cdc dvCdc  idc  s (t )iS 1 Ldc LB (22)
 dt
 di u matrix is defined as below:
 Ldc dc  Vdc  vCdc   RLdc  RCdc  idc
 dt
 dv V 
CB CB  sgn  iS 2  iS 2  Dibat u   dc 
 dt Vbat  (23)
 dibat
 LB  DvCB  Rbat ibat  Vbat
 dt
REFERENCES
Using GSSA and the following state variables, the A matrix [1] S. Li and C. C. Mi, “Wireless Power Transfer for Electric Vehicle
is obtained: Applications,” IEEE J. Emerg. Sel. Top. Power Electron., vol. 3, no. 1, pp. 4–
17, Mar. 2015.

 iS 1 1  x1  j x2 iPri  x3  j x4 [2] V. Etacheri, R. Marom, R. Elazari, G. Salitra, and D. Aurbach, “Challenges

in the development of advanced Li-ion batteries: a review,” Energy Environ.

1

 vCp1  x5  j x6 vCs1 1  x7  j x8 Sci., vol. 4, no. 9, pp. 3243–3262, Aug. 2011.

 1
[3] S. J. Gerssen-Gondelach and A. P. C. Faaij, “Performance of batteries for
 iS 2 1
 x9  j x10 iSec 1
 x11  j x12 electric vehicles on short and longer term,” J. Power Sources, vol. 212, pp. 111–
 129, Aug. 2012.
 vCp 2  x13  j x14 vCs 2 1
 x15  j x16 (20) [4] Z. Bi, T. Kan, C. C. Mi, Y. Zhang, Z. Zhao, and G. A. Keoleian, “A review

1
of wireless power transfer for electric vehicles: Prospects to enhance
 vCdc 0
 x17 idc 0
 x18 sustainable mobility,” Appl. Energy, vol. 179, pp. 413–425, Oct. 2016.
 [5] M. Budhia, G. A. Covic, and J. T. Boys, “Design and Optimization of
 vCB 0
 x19 ibat 0
 x20 Circular Magnetic Structures for Lumped Inductive Power Transfer Systems,”
IEEE Trans. Power Electron., vol. 26, no. 11, pp. 3096–3108, Nov. 2011.

In the A matrix: [6] A. Zaheer, H. Hao, G. A. Covic, and D. Kacprzak, “Investigation of Multiple
Decoupled Coil Primary Pad Topologies in Lumped IPT Systems for
Interoperable Electric Vehicle Charging,” IEEE Trans. Power Electron., vol.
30, no. 4, pp. 1937–1955, Apr. 2015.
[7] S. Y. Choi, S. Y. Jeong, B. W. Gu, G. C. Lim, and C. T. Rim, “Ultraslim S-
Type Power Supply Rails for Roadway-Powered Electric Vehicles,” IEEE
Trans. Power Electron., vol. 30, no. 11, pp. 6456–6468, Nov. 2015.

2168-6777 (c) 2017 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.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/JESTPE.2017.2761763, IEEE Journal
of Emerging and Selected Topics in Power Electronics
15

 r1 RP1 1 
 L  0  0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
LS 1 LS 1
 S1 
 r1 RP1 1 k 
   0 0  0 0 0 0 0 0 0 0 0 0 0 0 0 
 LS 1 LS 1 LS 1 LS 1 
 RP1 LSec r2 LSec LSec LSec MRP 2 Mr3 M M 
 0   0  0 0  0  0  0 0 0 0 0 
 det det det det det det det det 
 RP1 LSec rL LSec L MRP 2 Mr M M 
 0   2 Sec 0 0  Sec 0 0  3 0  0  0 0 0 0 
 det det det det det det det det 
 1 1 
 C 0  0 0  0 0 0 0 0 0 0 0 0 0 0 0 0 0 
CP1
 P1 
 1 1 
 0 0   0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
 CP1 CP1 
 1 
 0 0 0 0 0 0  0 0 0 0 0 0 0 0 0 0 0 0 
 CS 1 
 1 
 0 0 0 0 0  0 0 0 0 0 0 0 0 0 0 0 0 0 
 CS 1 
 
 2 

r4

RP 2 1 
 0 0 0 0 0 0 0 0
LS 2 LS 2
0
LS 2
0 0 0 0 0
LS 2
0 
 
 r4 RP 2 1 
 0 0 0 0 0 0 0 0   0 0 0 0 0 0 0 0 
LS 2 LS 2 LS 2
A 
 MRP1 Mr2 M M RP 2 LPri r3 LPri LPri LPri 
 0  0 0  0 0    0  0 0 0 0 0 
 det det det det det det det det 
 MRP1 Mr2 M M RP 2 LPri r3 LPri LPri LPri 
 0 0  0 0  0   0  0  0 0 0 0 
 det det det det det det det det 
 1 1 
 0 0 0 0 0 0 0 0  0 0 0  0 0 0 0 0 0 
CP 2 CP 2
 
 1 1 
 0 0 0 0 0 0 0 0 0  0  0 0 0 0 0 0 0 
 CP 2 CP 2 
 1 
 0 0 0 0 0 0 0 0 0 0 0 0 0 0  0 0 0 0 
 CS 2 
 1 
 0 0 0 0 0 0 0 0 0 0 0 0 0  0 0 0 0 0 
 CS 2 
 k 1 
 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
 Cdc Cdc 
 r5 
 0 1
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0   0 0 
 Ldc Ldc 
 
 4
 D 
 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0  
 CB CB 
 D R 
 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0  bat 
 LB Lbat 

[8] S. Choi, J. Huh, W. Y. Lee, S. W. Lee, and C. T. Rim, “New Cross- [16] M. Moghaddami, A. Anzalchi, and A. Sarwat, “Single-Stage Three-
Segmented Power Supply Rails for Roadway-Powered Electric Vehicles,” Phase AC-AC Matrix Converter for Inductive Power Transfer Systems,” IEEE
IEEE Trans. Power Electron., vol. 28, no. 12, pp. 5832–5841, Dec. 2013. Trans. Ind. Electron., pp. 1–1, 2016.
[9] J. M. Miller et al., “Demonstrating Dynamic Wireless Charging of an [17] H. H. Wu, A. Gilchrist, K. D. Sealy, and D. Bronson, “A High Efficiency
Electric Vehicle: The Benefit of Electrochemical Capacitor Smoothing,” IEEE 5 kW Inductive Charger for EVs Using Dual Side Control,” IEEE Trans. Ind.
Power Electron. Mag., vol. 1, no. 1, pp. 12–24, Mar. 2014. Inform., vol. 8, no. 3, pp. 585–595, Aug. 2012.
[10] L. Chen, G. R. Nagendra, J. T. Boys, and G. A. Covic, “Double-Coupled [18] B. X. Nguyen et al., “An Efficiency Optimization Scheme for
Systems for IPT Roadway Applications,” IEEE J. Emerg. Sel. Top. Power Bidirectional Inductive Power Transfer Systems,” IEEE Trans. Power
Electron., vol. 3, no. 1, pp. 37–49, Mar. 2015. Electron., vol. 30, no. 11, pp. 6310–6319, Nov. 2015.
[11] W. Zhang and C. C. Mi, “Compensation Topologies of High-Power [19] K. Colak, E. Asa, M. Bojarski, D. Czarkowski, and O. C. Onar, “A Novel
Wireless Power Transfer Systems,” IEEE Trans. Veh. Technol., vol. 65, no. 6, Phase-Shift Control of Semibridgeless Active Rectifier for Wireless Power
pp. 4768–4778, Jun. 2016. Transfer,” IEEE Trans. Power Electron., vol. 30, no. 11, pp. 6288–6297, Nov.
[12] C.-S. Wang, G. A. Covic, and O. H. Stielau, “Power transfer capability 2015.
and bifurcation phenomena of loosely coupled inductive power transfer [20] Z. U. Zahid et al., “Modeling and Control of Series-Series Compensated
systems,” IEEE Trans. Ind. Electron., vol. 51, no. 1, pp. 148–157, Feb. 2004. Inductive Power Transfer System,” IEEE J. Emerg. Sel. Top. Power Electron.,
[13] M. Borage, S. Tiwari, and S. Kotaiah, “Analysis and design of an LCL- vol. 3, no. 1, pp. 111–123, Mar. 2015.
T resonant converter as a constant-current power supply,” IEEE Trans. Ind. [21] B. Sharp and H. Wu, “Asymmetrical Voltage-Cancellation control for
Electron., vol. 52, no. 6, pp. 1547–1554, Dec. 2005. LCL resonant converters in Inductive Power Transfer systems,” 2012, pp. 661–
[14] S. Li, W. Li, J. Deng, T. D. Nguyen, and C. C. Mi, “A Double-Sided LCC 666.
Compensation Network and Its Tuning Method for Wireless Power Transfer,” [22] H. H. Wu, A. Gilchrist, K. Sealy, P. Israelsen, and J. Muhs, “Design of
IEEE Trans. Veh. Technol., vol. 64, no. 6, pp. 2261–2273, Jun. 2015. Symmetric Voltage Cancellation Control for LCL converters in Inductive
[15] H. L. Li, A. P. Hu, and G. A. Covic, “A Direct AC-AC Converter for Power Transfer Systems,” 2011, pp. 866–871.
Inductive Power-Transfer Systems,” IEEE Trans. Power Electron., vol. 27, no. [23] S. Bai, Z. Pantic, and S. Lukic, “A comparison study of control strategies
2, pp. 661–668, Feb. 2012. for ZVS resonant converters,” 2010, pp. 256–262.

2168-6777 (c) 2017 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.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/JESTPE.2017.2761763, IEEE Journal
of Emerging and Selected Topics in Power Electronics
16

[24] Z. Pantic, S. Bai, and S. M. Lukic, “ZCS LCC-Compensated Resonant

Inverter for Inductive-Power-Transfer Application,” IEEE Trans. Ind.
Electron., vol. 58, no. 8, pp. 3500–3510, Aug. 2011.
[25] J. M. Miller, O. C. Onar, and M. Chinthavali, “Primary-Side Power Flow
Control of Wireless Power Transfer for Electric Vehicle Charging,” IEEE J.
Emerg. Sel. Top. Power Electron., vol. 3, no. 1, pp. 147–162, Mar. 2015.
[26] H. Hao, G. A. Covic, and J. T. Boys, “An Approximate Dynamic Model
of LCL- -Based Inductive Power Transfer Power Supplies,” IEEE Trans. Power
Electron., vol. 29, no. 10, pp. 5554–5567, Oct. 2014.
[27] N. Hasan, H. Wang, T. Saha, and Z. Pantic, “A novel position sensorless
power transfer control of lumped coil-based in-motion wireless power transfer
systems,” in 2015 IEEE Energy Conversion Congress and Exposition (ECCE),
2015, pp. 586–593.
[28] R. Tavakoli, A. Jovicic, N. Chandrappa, and Z. Pantic, “Design of a Dual-
Loop Controller for In-motion Wireless Charging of an Electric Bus,” in 2016
IEEE Energy Conversion Congress and Exposition (ECCE), Milwaukee, 2016.
[29] G. A. Covic and J. T. Boys, “Modern Trends in Inductive Power Transfer
for Transportation Applications,” IEEE J. Emerg. Sel. Top. Power Electron.,
vol. 1, no. 1, pp. 28–41, Mar. 2013.
[30] “Typical Values for Driving Performance with Emphasis on the Standard
Deviation of Lane Position: A Summary of the Literature,” Volpe - The
National Transportation Systems Center, 06-Jan-2014. [Online]. Available:
https://www.volpe.dot.gov/safety-management-and-human-factors/surface-
transportation-human-factors/typical-values-driving. [Accessed: 03-Mar-
2017].
[31] S. R. Sanders, J. M. Noworolski, X. Z. Liu, and G. C. Verghese,
“Generalized averaging method for power conversion circuits,” IEEE Trans.
Power Electron., vol. 6, no. 2, pp. 251–259, Apr. 1991.

Reza Tavakoli (S’13) received his B.S.

from Iran University of Science and
Technology in 2011 and his M.S. from
Shahid Beheshti University in 2014, both
in electrical engineering, Tehran, Iran.
Since 2015, he was working with Utah
State University Power Electronics lab
where he is pursuing his PhD degree. His
current research interests include wireless power transfer,
transportation electrification and dynamic charging of electric
vehicles.

Zeljko Pantic (S’10–M’13) received the

B.Sc. and M.Sc.degrees from the School
of Electrical Engineering, Belgrade
University, Belgrade, Serbia, in 1998 and
2007. In 2001, he joined the group for
Power Converters and Motor Drives as a
Teaching and Research Assistant for
motor drive and electric-vehicle-related
courses and was a Coinstructor. In 2009,
he became a Research Assistant at the NSF-funded Future
Renewable Electric Energy Delivery and Management
(FREEDM) Systems Center, North Carolina State University in
Raleigh, NC, USA. As a Ph.D. student at NC State, he explored
the emerging topic of wireless inductive power transfer
(WIPT). After graduation in 2013, he joined the ECE
Department, Utah State University, Logan, USA, as an
Assistant Professor. His current research interests include
systems for wireless inductive power transfer, power
electronics, electric and hybrid electric vehicles, and motor
drives.

2168-6777 (c) 2017 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.