Novel Transformerless Multilevel Inductive Power Transfer System

Previous high power on-line inductive power transfer (IPT) systems use line-frequency (LF) transformers and high-frequency (HF) transformers to be compatible with the ratings of commercial power switches. However, LF and HF transformers make the IPT systems bulky and expensive. This study proposes a new transformerless multilevel on-line IPT system using multilevel rectifiers, inverters, and excitation coils. Instead of the LF transformers, a series-connected multilevel rectifier and inverter are used. Instead of the HF transformers, excitation coils, which have strong magnetic couplings with a transmitter coil, is connected to each of the output terminals of the multilevel inverter. The magnetic fields generated by the multiple excitation coils induce a high voltage and current in the transmitter coil. The induced current generates the augmented magnetic field in the transmitter; therefore, a receiver coil that is loosely coupled with the transmitter can collect power from the mains. The input impedance, voltage and current gains, and efficiency of the proposed IPT system are investigated using an equivalent circuit model. The impact of the excitation coil shape on the magnetic couplings is explored using finite element analysis. Based on the investigation, a design for the excitation coil is presented. The feasibility of the proposed system is evaluated using a 4-level experimental test-bed. Measured coil-to-coil and DC-to-load efficiencies of the proposed system are observed to be 88% and 84%, respectively, over a 7 cm air-gap.


I. INTRODUCTION
IPT systems can deliver power from energy sources to loads without any physical contact [1]- [4]. Because they are more convenient, safe, and reliable than conventional power supplies, many researchers have developed stationary or in-motion IPT systems for electric cars, buses, and trains [5]- [9]. Fig. 1 shows a block diagram of a high power on-line IPT system in [9]. Line-frequency (LF) transformers are used at the front end of the IPT system to make the medium-voltage (MV) input down to 750 VDC, a reliable level for commercially available power semiconductors. Parallel-connected H-bridge inverters are connected to the rectifier's DC bus to expand the power level. The output voltages of the inverters are connected to primary windings of high-frequency (HF) transformers, which have seriesconnected secondary windings. The output powers of the inverters are aggregated by the transformers' series- connected secondary windings, which are supplying a net voltage of 3 kVrms to the transmitter coil. It should be noted that this study uses the LF and HF transformers to build the MV system. The structure of this IPT system is straightforward and easy to implement. However, the power density and cost-effectiveness of the IPT system are not good because it requires bulky transformers.
Li et al. proposed a multilevel IPT system with active rectifiers and cascaded H-bridge inverters to eliminate the LF transformer [10]. In this study, a transmitter coil and a compensation capacitor are connected in series with the multilevel inverter. This topology is conducive to supplying high voltage output to the transmitter coil at high switching frequencies. However, high-voltage insulations of the transmitter coil and the resonant-tuning capacitors are drawbacks of this system. It is very challenging to apply the proposed idea to the MV system. Park et al. proposed a multilevel inverter with HF transformers with secondary windings connected in series like Fig. 2(a) [11]. However, the high voltage insulation between the primary windings and the cores of the HF transformer makes the system bulky and expensive. In addition, a failure of one winding of the transformers results shutdown of the entire system. Ayyanar et al. used HF transformers with parallel-connected secondary windings at the outputs of the multilevel inverter as Fig. 2(b) [12]. It should be noted that the parallel-connected secondary windings can suffer from a large circulation current because of the voltage imbalance between the transformers. HF transformers that have two or more primary windings with a single secondary winding can be employed for the multilevel system as Fig. 2(c) [13]- [15]. Augmented magnetic fields of the primary windings can induce high voltage and current at the secondary winding terminal. However, it is required to replace the entire transformer even for a failure of a single winding of the primary side.
A new multilevel IPT system was proposed in [16], which uses multiple excitation coils (see Fig. 2(d)) instead of the HF transformers. Like the multiple primary winding HF transformer, synchronized magnetic fields generated by the excitation coils are augmented and induced high voltage and current to the transmitter coil. Then, the magnetic field generated by the current of the transmitter coil induces voltage and current to the receiver coil, which is loosely coupled with the transmitter coil. The excitation coils are galvanically isolated and have strong magnetic couplings with the transmitter coil. However, the excitation coils do not have couplings with the other excitation coils. This is the unique feature of the study. Because the excitation coils are isolated to each other, this system has the following four advantages compared to the conventional multilevel IPT systems: 1) ease of modularization, 2) fail-safe operation, 3) ease of high voltage insulation, and 4) high power density. Since the excitation coils have no magnetic coupling between them, the excitation coil and its driving inverter can be easily modularized without interfering with other excitation coils and inverters. In addition, a failure of one excitation coil results in a 1/N reduction of the induced voltage at the transmitter and the receiver coil (N is the total number of excitation coils). Therefore, the failure of a few excitation coils or a few converters does not significantly affect the performance of the proposed IPT system. Furthermore, because there are large air gaps between the transmitter coil and the excitation coils, the high voltage insulation of the excitation coil can be easily implemented. Finally, because the power level of each excitation coil is very small compared to the total output power, the excitation coil can be implemented in a compact volume. An overall configuration of the proposed multilevel on-line IPT system is shown in Fig. 3. The IPT system includes a multilevel active rectifier, a multilevel inverter, excitation coils, a transmitter coil, a receiver coil, a receiver rectifier, a receiver-side DC/DC converter, and a load. This paper expands the discussion in [16] by including additional circuit analysis on the resonant frequencies and the current and voltage gains of multi-coil IPT system. Also, this study includes the impact of excitation coil shape on magnetic couplings and self-inductances of the coils. Based on the analysis, a design for the excitation coil is proposed as well. In addition, a 4-level IPT system is built to evaluate the proposed configuration experimentally. In the first part of this paper, the proposed multilevel on-line IPT system with excitation coils is demonstrated and analyzed using a Ttype equivalent circuit model. In the following section, an excitation coil design is demonstrated using theoretical and finite element analysis (FEA) results. In the last section, the viability of the proposed IPT system is evaluated using experimental results. Fig. 4 shows an equivalent circuit model of the proposed multilevel IPT system with N excitation coils. The seriesconnected voltage sources represent series-connected rectifiers and inverters at the front end. An inductor and a capacitor are included to transform the input impedance at low-and high-frequency ranges.

II. Analysis of Proposed Multilevel IPT System
is the self-inductance of the excitation coil, and and are the self-inductance and capacitance of the transmitter resonant tank, respectively. and are the selfinductance and capacitance of the receiver resonant tank, respectively. , , and are the equivalent-seriesresistances (ESRs) of the excitation, transmitter, and receiver coils, respectively.
is the coupling coefficient between the excitation coil and transmitter coil, and the is the coupling coefficient between the transmitter and receiver coils. and are currents of the transmitter and receiver coils induced by N excitation coils, respectively. Fig. 5 shows an equivalent circuit model seen from one of the excitation coils.
= is the mutual-inductance between the excitation and transmitter coils and = is the mutual-inductance between the transmitter and receiver coils.
= ⁄ and = ⁄ are the transmitter and receiver coil currents induced by one of the excitation coils, respectively. To find the input impedance seen from one of the excitation coils, the inductances and resistances in the transmitter and receiver circuits are multiplied by N, and the capacitances are divided by N.

A. Input impedance seen from a single excitation coil
Using the equivalent circuit in Fig. 5, the input impedance seen from one of the excitation coils follows (1).
where, is the angular operating frequency, = + + 1/( ) + is the impedance of the excitation resonant tank, = + 1/( ) + is the impedance of the transmitter resonant tank, = + 1/( ) + + is the impedance of the receiver resonant tank, and is the impedance of the combined circuit of the transmitter and receiver resonant tanks (see Fig. 5). N is the number of excitation coils.
Using (1) and selected circuit parameters in TABLE I, frequency responses of with different numbers of excitation coils are plotted in Fig. 6. It should be noted that the impedance plot has three resonant frequencies. The first resonant frequency , which is caused by the series resonance between + and , follows (2). Parameter robustness of the multiple excitation coil system depends on the as parameter errors exist on the , , and . Furthermore, parameter mismatches can generate additional resonances by combining with other resonant tanks (transmitter and receiver). Fig. 7(a) shows the frequency responses of seen from excitation coils of a 4-level IPT system when is 60 kHz (very close to the tuning frequencies of the transmitter and the receiver). The matching inductance of the second level ′ has a 10% error compared to the other levels' matching inductances. It should be noted that the 10% error in significantly VOLUME XX, 2017 9  distorts the input impedances at the tuning frequencies of the coils. Fig. 7(b) shows the frequency responses of seen from excitation coils of a 4-level IPT system when is 15 kHz. The impact of parameter error is decreased significantly compared to the previous case. Therefore, the tuning frequency is selected to be 15 kHz to avoid the effects of parameter errors of , , and . This is much lower than the tuned frequencies of the transmitter and receiver coils (61 kHz and 64 kHz, respectively).
The second resonant frequency is caused by the parallel resonance between the transmitter and receiver tanks. The is the frequency that satisfies a condition (3) and an approximated solution of is shown in (4).
where Re() is an operator that denotes the real part of a complex number, and = 1/(2 ) and = 1/(2 ) are the tuning frequencies of the transmitter and receiver coils, respectively. It should be noted that the peak magnitude of at increases with the increase of N as shown in Fig. 6. Therefore, is not a suitable operating frequency for a multilevel system with the large number of N.
The third resonant frequency is caused by the combined resonance between the excitation, transmitter, and receiver coils. This frequency satisfies the condition (5).
where Im() is an operator that denotes the imaginary part of a complex number. Because of its complexity, its closed form solution is shown in Appendix. Unlike the resonant frequencies and , the frequency moves away as the number of N increases, resulting in magnitude and phase variations of the input impedance.

B. Current and voltage amplifications<
The current gain of the transmitter coil to the excitation coil with N excitation coils can be calculated as (6). with N excitation coils are calculated as (7) and (8). Fig. 8(b) shows a frequency response of the voltage gain of / with different numbers of excitation coils. Unlike the current gain, the maximum voltage gain appears at . Fig.  8(c) shows frequency responses of the voltage gain of / with different numbers of excitation coils. It should be noted that the voltage gain is independent on the number of excitation coils. Fig. 8(d) shows frequency responses of the voltage gain of / . Like the responses of / , the frequency response of / has its peak at .

C. Efficiency
The coil-to-coil efficiency of the IPT system with N excitation coils follows (9).
The , , and are shown in Fig. 4. Fig. 9 shows the  calculated coil-to-coil efficiency using (9). The peak efficiency appears at the tuned frequency of the receiver coil and is insensitive to the number of excitation coils. In this study, the tuned frequency of the receiver coil is selected as the operating frequency to have maximum efficiency at the rated load. As the number of excitation coils increases, the losses of the excitation coils are negligible compared to other losses. The loss of the excitation coil is proportional to N, while the losses of the transmitter and receiver coils increase by a factor of N 2 .

III. Excitation Coil Design
To implement the proposed multilevel IPT system, practical design for the excitation coil is required. A new design for the excitation coil for the proposed IPT system is demonstrated in this section.

A. Transmitter and receiver coils
The transmitter is single-turn (two wires in parallel), coreless, 2 m long, 20 cm wide, and 2 cm turn spacing. The receiver coil is four-turn, ferrite-cored, 50 cm long, 30 cm wide, and 2 cm turn spacing. The lengths of the transmitter and receiver coils are 2 m and 50 cm, respectively, for a laboratory test. The air gap between the transmitter and receiver coils is 7 cm. Fig. 10(a) and (b) show the transmitter and receiver coils, respectively.

B. Proposed excitation coil design
As mentioned earlier, the excitation coil needs a strong coupling with the transmitter coil but very weak couplings with other coils. Fig. 11 shows the design of the proposed excitation coil. Four ferrite plates form a hollow hexahedron, which wraps around the transmitter coil. The excitation coil is wound around the bottom plate in a parallel direction to the transmitter coil. The magnetic field generated by the excitation coil flows through the ferrite VOLUME XX, 2017 9 core and couples with the transmitter coil. Since the ferrite core can efficiently confine the magnetic field inside the core, the proposed design has a strong magnetic coupling with the transmitter coil and a weak coupling with the receiver coil.
As it is shown in Fig. 11, the mean path length (MPL) , air-gap length , cross-sectional area of the core S, and the number of turns of the coil are the design parameters to achieve the required self-inductance and the coupling coefficients in TABLE I. For the core, ferrite cores that are made with PL11 (Samwha electronics) [17]. The sizes of the ferrite cores that are used in the design are 1 cm × 10 cm × 10 cm and 1 cm × 4 cm × 10 cm. Because the permeability of the PL11 core (initial permeability = 2500) is much larger than that of air, the reluctance of the air is a dominant factor for the inductance and coupling. Therefore, the MPL is not a critical parameter in the excitation coil design. The is selected to be 26 cm to ensure sufficient distance from the transmitter coil. The air gap is the critical parameter that are investigated in this study. Using an FEA simulation software, ANSYS HFSS [18], the dependence of the self-and mutual-inductances of the transmitter, receiver, and excitation coils on the opening  are investigated. Fig. 12 shows details of the simulation model. Fig. 13 shows simulated self-inductances of the excitation coil ( ), transmitter coil ( ), and receiver coil ( ) with . When the opening is small, the rate of decrease of is very fast as the increases. It is because total magnetic flux in the core is inversely proportional to the value of + / when the gap is comparable to / . If is much bigger than / , the flux is dependent only on , and decreases slowly. It should be noted that the inductance also decreases as increases. Since the ferrite core encircles the transmitter coil, the inductance of the transmitter coil is increased when the is small. However, decreases as and corresponding reluctance increase. As the excitation coils have very weak couplings with the receiver coil, is independent to the opening . Fig. 14(a) shows the mutual-inductance ( ) and the coupling coefficient ( ) between the excitation and transmitter coils with . The leakage flux of excitation coils increases with an increase in . Consequently, and the also decrease. To achieve a strong coupling between the excitation and transmitter coils, a smaller is suitable. However, and are very sensitive to when is small as shown in Fig. 13. There is a trade-off relationship between the strong coupling and the parameter robustness of and . To enhance the parameter robustness of the excitation and transmitter coils, the air-gap length is chosen to be 1 mm. Then, the number of turns, , is chosen to be 7 to achieve the target self-inductance of the excitation coil in TABLE I with of 1 mm. Selected design results of the excitation coil are listed in TABLE II. Fig. 14(b) shows the mutual-inductance ( ) and the coupling coefficient ( ) between the transmitter and receiver coils with . It should be noted that the is irrelevant to the . Because the increases while the decreases depending on the length of the .
In section II, the circuit analysis assumes that the excitation coils have no magnetic coupling with each other. Therefore, the cross-couplings between adjacent excitation coils with the selected design are investigated.
is the coupling coefficient between adjacent excitation coils. Fig.  15 shows depending on the distance between excitation coils. It should be noted that the cross-couplings between adjacent excitation coils can be neglected when the distance between them is greater than 10 cm.

IV. Evaluation of Proposed Multilevel IPT System
To evaluate the proposed multilevel on-line IPT system, a four-level IPT system was built. The test setup has the same mechanical and electrical parameters as the simulation model in section III and Fig. 12. Figs. 16 and 17 show a configuration and a photo of the setup. Fig. 18 shows a photo of the designed excitation coils. The target IPT system consists of four series-connected full-bridge inverters, four excitation coils, a transmitter coil, a receiver coil (air-gap 7 cm between the transmitter and the receiver), and a load resistor. The input voltages of the inverters were regulated at 100 V. Four SiC MOSFETs (C3M0120090D, Cree/Wolfspeed [19]) were used for the full-bridge inverters. The switching frequency of the inverters was 64 kHz which was the tuned frequency of the receiver coil. The circuit parameters of the setup are same as the parameters in TABLE I. Fig. 19 presents measured voltage and current waveforms for two different operations: 1) when the four inverters' output voltages are in-phase and 2) when an output voltage of the inverters is out-of-phase with the others. Fig. 19(a) shows measured voltage and current waveforms of the inverter, excitation, transmitter, and receiver coils when the inverter output voltages are inphase. It should be noted that all excitation coil currents had nearly the same magnitudes and phases and the input impedances were well balanced. Each inverter delivered the same output power of 68.5 W to the load, and the total output power was 273 W. The average of the root-meansquare (rms) values of the four excitation coil currents were 894 mArms, while the rms of the transmitter was 28.8 Arms. Therefore, the current gain from the excitation coil current to the transmitter coil current was 32, and it agreed very well with the theoretical result in Fig. 8(a). At this operating condition, the sum of the input powers of the four coils was 310 W, and the measured power from the 400 V DC power supply was 325 W. Therefore, coil-to-coil efficiency was 88%, and DC-to-load efficiency was 84% over the 7 cm air-gap. Fig. 19(b) shows the voltages and currents of the system when there is 5° phase mismatch between the inverters. It should be noted that the excitation coil currents were synchronized even if their voltages had 5° phase mismatch, as shown in the first grid. The magnitude of the mismatched inverter's current decreased from 881 to 827 mArms and the other excitation coil currents increased about 2%. Therefore, the small mismatch of the inverter's output FIGURE 16. Configuration of the test-bed. voltage has a negligible impact on the transmitter coil current and corresponding output voltage of the receiver coil. To verify the measured result, the dependence of the currents of the excitation coils on the phase mismatch of the inverter voltages are calculated using the equivalent circuit in Fig 4. Fig. 20 shows the calculated magnitude and phase of the excitation coils on the phase mismatch of . In this figure, is the current of the excitation coil connected to the inverter that is synchronized with the other inverters, and is the current of the excitation coil connected to the inverter that is out-of-phase. In Fig. 20(a), it should be noted that the magnitude of is reduced 5% while the magnitudes of the other excitation coil currents are increased 2% when the phase of is -5°. This result agrees very well with the measured result. Furthermore, the excitation coil currents have the same phases when the phase mismatch is in the range of -45° to +10° as shown in Fig. 20(b). This theoretical analysis explains the experimental result in Fig. 19(b). This result shows the robustness of the proposed multilevel IPT system on phase mismatches between the inverters. Fig. 21 shows the theoretical and measured frequency responses of the proposed IPT system. The theoretical and measured input impedances seen from the four inverters are shown in Fig. 21(a) and (b). The current gain of the transmitter coil to the excitation coil and the voltage gain of the load to the source voltage of the inverters are shown in Fig. 21(c) and (d). This result shows the validity of the circuit analysis in section II.

V. Conclusion
A novel transformerless multilevel on-line IPT system is proposed in this study. The proposed system utilizes multiple excitation coils with strong magnetic couplings to a transmitter coil. It was found that the voltages induced by multiple excitation coils can be augmented on the transmitter coil because all excitation coil currents have the same frequency and phase. The excitation coils significantly help in improving modularization, fail-safe operation, power density, providing galvanic isolation, and ensuring high electrical insulation. Equivalent circuit models were developed to analyze the proposed multilevel IPT system. An example 4-level IPT system was designed to evaluate the proposed topology. Using simulation results, the magnetic couplings between the excitation, transmitter, and receiver coils, depending on the shape of the excitation coils were investigated. The selfinductance of the transmitter coil is influenced by the shape of the excitation coils, but the mutual-inductance between the transmitter coil and the receiver coil is not. The designed shape of the excitation coil is based on a trade-off relationship between the strong coupling between the excitation and transmitter coils and the parameter robustness of the self-inductances of the excitation and transmitter coils.
The viability of the proposed IPT system was evaluated using the experimental results. The experimental results were in good agreement with the theoretical results. Measured coil-to-coil and DC-to-load efficiencies of the proposed system were 88% and 84%, respectively, over the 7 cm air-gap.