More Efficient Inductive Electric Vehicle Charger: Using Autonomy to Improve Energy Efficiency.
Electric vehicles (EVs) provide a zero-emissions solution for transportation in cities. In 2013, sixty percent of over one thousand respondents were open to the idea of purchasing an EV . Unfortunately, 48% of people surveyed do not have the ability to charge at home, due to lack of electrical outlet or space for a charger . A wireless, hands-free charger that is durable enough to withstand environmental factors would be well suited to strategically important uncontrolled parking environments, like apartment parking lots and public garages , for overnight charging .
Modern inductive EV chargers provide low-clutter, ground-mount charging that is impervious to chemicals and electrically safe, as there are no exposed galvanic connections to create a shock or spark hazard. Today's systems are also hands-free. The existing designs available to consumers, though, are not energy efficient. Wall-to-battery inductive charging efficiencies are right at 85% [4, 5], with state-of-the-art research in the low nineties [6, 7, 8]. Galvanic chargers deliver 85 to 95% wall-to-battery efficiency [9, 10], and address hands-free operation by using a robot to make the high precision galvanic connections at waist level .
As vehicles become more autonomous and begin to park themselves with robot-precision , the need for a robot function in the charger is eliminated. The car can precisely drive into a docking station where the charging connections are made. We wondered, "Is there a way to use the precision of autonomous vehicle parking and modern circuit topologies to make ground-level inductive chargers as efficient as hot contact chargers?" GE Critical Power produces rectifier circuits that are 95 to 97% efficient . The goal was to start there and use the transformer as the interface for inductive energy transfer to the vehicle--with single-owner apartment-home night charging as the use model. Figure 1 demonstrates our desired product.
We based our design on experience designing, producing and supporting over a million products using this circuit, references including  -  and two guiding principles.
The first, which is the foundation for the heart of this work is: "Everything is linear if you get close enough." We know this circuit delivers 95-96% efficiency with a separate resonant inductor and highly interleaved transformer. Our goal is to demonstrate a single transformer that can be assembled each time a vehicle parks that delivers conversion efficiencies that are comparable to that highly interleaved transformer that requires 20 minutes of assembly time to construct. To do that we needed to make all the key parameters close to those of our reference circuit.
The second guiding principle is: "Space allocation is key to product success." We had a team working to package the circuit in way that is practically useful. Heat dissipation, alignment, and customer acceptance are all important, if this product is to serve society well. These efforts act as bookends around the theoretical and experimental work that is the heart of this paper. These space allocation efforts form a blueprint for future work and appear only in that section.
Resonant LLC Circuit Topology
Basing our design off of proven GE Critical Power rectifier topologies, we used the resonant LLC converter circuit with a half-bridge and output voltage doubling , Figure 2.
For an environmentally robust design, we wanted to accommodate bumper covers and other plastic elements needed for 10,000 charging cycles of operation. Any separation in the magnetic structure caused by plastic covering prevents ideal coupling, creating leakage inductance. This intrinsic leakage inductance can be treated as beneficial. It allows us to forego the inclusion of a discrete resonant inductor and use the single assembled magnetic structure to establish resonance [21, 22].
With GE's in-house analysis tool, we calculated the First Harmonic Approximation for the circuit. The inputs were our desired input voltage, output power, and power efficiency, gathered from  for market needs. The tool produced the required turns ratio, magnetizing inductance, resonant inductance--in our case, the leakage inductance, frequency range, and expected output voltage range. Table 1 shows our goals for the reference prototype; Figure 3 is an example of the tool's output as we work to hit the corners of the envelope.
Resonant Frequency Range
Efficiency and power loss is optimized when the circuit runs at or slightly above the resonant frequency--where the current running through the transformer is almost a perfect sinewave . Our in-house analysis tool outputs a frequency range, with the minimum determined by the magnetizing inductance and the maximum determined by the leakage inductance. At high frequencies, current does not run efficiently through standard wire--the skin effect causes AC losses, and the proximity effect of the windings on each other reduce the efficiency, too. To decrease the effect of the variation of frequency on our transformer losses, we chose to use Litz wire for all of our windings.
Core Gap Determination
Determining the transformer core gap, Figure 4, was an important decision at the start. The larger the core gap, the less the coupling . For our environmental protection purposes, we wanted a large core gap. To find the optimal core gap, we first measured the effect of core gap on magnetizing and leakage inductance, and then found an operating point where variations in core gap would not create large variance in the inductances. We determined an insulator thickness of 1.5mm on each surface of the ground-side and vehicle-side would be durable and manufacturable with standard injection molding processes.
We used a set-based concurrent engineering approach  to design the integrated transformer, so we wound and evaluated many transformers with a variety of geometries. For each geometry, we measured magnetizing inductance, leakage inductance, and AC resistance on the Keysight E4980A Precision LCR Meter.
The four primary geometries we began with are shown in Figure 5.
Since the objective was to test a wide range of possibilities, we began with both "core" and "shell" type transformers . Since theoretical calculation methods vary depending upon the type of core, we elected to wind and experiment instead of focusing on models and mathematics. The first geometry, Figure 5a, was an L-L core with the primary and secondary windings on the same leg. The primary winding nests into the secondary winding. From a theoretical perspective, this winding should couple well, so the leakage inductance should be relatively low. The downside to this design is the accuracy necessary for the nesting of the windings. The second winding, Figure 5b, also uses an L-L core, but has the primary on one leg and the secondary on the other. Because there is no overlap of the windings at all, this winding was expected to have higher leakage and thus, lower magnetizing inductance. The third geometry, Figure 5c, was a "shell" type, an E-E core, with the primary winding on one end of the middle leg and the secondary on the other. The main benefit of the E-E core is the ability to have the windings close to each other, even though they do not overlap. Finally, we had a toroid core, Figure 5d, that was split in half, with windings on each side. Measurements were taken with the windings at a range of locations, from close together to far apart.
Additionally, we experimented with different configurations of core gap, turns ratio, number of turns per side, and frequency. We experimented from no core gap up to 13mm per leg. From our analysis tool, we theoretically wanted a turns ratio of approximately 1.25:1, but we tested more variations, too. Theoretically, the number of turns on the primary side effects the magnetizing and leakage inductances, so we wanted to see how that varied experimentally. Finally, because we chose to use Litz wire, we expected to see little variation at different frequencies. We tested frequencies from 20kHz to 120kHz to verify the effectiveness of the Litz wire.
Figure 6 shows a compilation of our initial magnetizing and leakage inductance measurements. The orange circle denotes the ideal values of 110[micro]H and 10[micro]H, respectively. This data shows that points from the same geometry were clustered closely, separated from the others. The clusters consist of the 102 configurations we tested. This suggests that core shape affects magnetizing and leakage inductance most.
Evidently, the adjacent windings on the L-L core and the split toroid core are not at all close to the ideal values. Meanwhile, some of the nested L-L core data points are close for magnetizing inductance, but all are too low in leakage inductance. The E-E core data points show a possibly ideal leakage, but low values for magnetizing inductance. W made the decision to focus on improving the nested L-L core and the E-E core .
In the spirit of set-based concurrent engineering, we continued along with both the nested L-L and the E-E, since both had pros and cons. The nested L-L had more promising values, but required a more accurate alignment of the two sides. The E-E did not require nesting, but the low magnetizing inductance and high leakage inductance were concerning.
Improving the Nested L-L Core
First, we wanted to increase leakage inductance in the nested L-L core geometry. E. C. Snelling's method for calculating leakage inductance is shown in (1), where N is the number of primary side turns, [l.sub.w] is the mean turn length, M is the number of section interfaces, Y is the winding overlap width, X is the overall winding height of affected windings, and [summation][X.sub.[DELTA]] is the sum of all interlayer section thickness. Note that all lengths and widths are in millimeters.
[mathematical expression not reproducible] (1)
Meanwhile, (2) is the equation for magnetizing inductance in a core with an air gap, where [[micro].sub.0] is the permeability constant, A is the cross-sectional area of the core leg, and [l.sub.total gap] is the length of the air gap .
Although increasing the number of primary side windings would increase leakage inductance, it would also increase the magnetizing inductance. The nested L-L core already exceeds the desired magnetizing inductance, so we want to look at other factors that affect leakage inductance.
[mathematical expression not reproducible] (2)
To only change the leakage inductance, we decided to shorten the width of the secondary winding. We accommodated this by making the winding taller. This decreases Y and increases X,[summation][X.sub.[DELTA]], and [l.sub.w]. Therefore, we introduced a new design, keeping the L-L core and nested windings, but with a narrower, taller secondary, shown in Figure 7a.
Upon adding the data from variations of the shorter nested L-L to the previous data, we see in Figure 8 that the leakage inductance increased from approximately 3[micro]H to 10[micro]H.
Improving the E-E Core
Next, we wanted to try to increase the magnetizing inductance on the E-E core. From (2), we knew the number of turns on the primary side would increase the magnetizing inductance. Therefore, keeping the same ratio, we wound an E-E core that was completely full with windings. Unfortunately, by filling up the core to maximize magnetizing inductance, the number of layers and thickness of the windings increased. This caused leakage inductance to increase greatly, too. The closest configuration to the desired values had between 80[micro]H and 90[micro]H magnetizing inductance, but at least 40[micro]H leakage inductance. We concluded that changing the number of primary side windings was not the solution.
We wanted to keep the material--thus the permeability--the same, and we did not want to reduce the core gap further. Therefore, we decided to adjust the cross-sectional area of the core, Table 2. We found a large pot core, shown in Figure 7b, which had a larger cross-sectional area, so we measured its magnetizing and leakage inductances at varying core gaps and frequencies. The magnetizing inductance increased by a large margin, reaching over 100|H, while the leakage inductance was approximately 20[micro]H. Although these numbers are not the exact desired values, we used GE's analysis tool to determine that it should still be usable if run at a lower resonant frequency of 85kHz.
Figure 8 also illustrates the performance of the pot core in context of the previous data. Along with the shorter secondary nested L-L core, the pot core was also much closer to the desired magnetizing and leakage inductance.
MEASURING PEAK POWER EFFICIENCY
Moving forward with the shorter secondary nested L-L and the large pot core transformers, we took power measurements in an open loop with a fixed duty cycle. We used an electronic 400V source to simulate the output of the power factor corrected boost stage (PFC). Figure 9 shows our setup.
For our first prototype efficiencies, the shorter secondary nested L-L performed very well, achieving 97% efficiency across the resonant tank DC to DC converter. The pot core reached 91% resonant tank DC to DC efficiency, which did not meet our goal, but may have room for improvement. Additionally, the voltage supply that we used had a maximum output of 2.5kW, so we were not able to test up to the desired 3kW. From this initial data, though, we have already proven that it is possible for an inductive EV charger with a large core gap to be highly efficient. When viewed at a system level, with a 98% efficient PFC, wall-to-battery efficiency would be 95% for the shorter nested L-L core and 88% for the large pot core. This data is shown in Figure 10.
Figure 10. Peak wall-to-battery efficiency data comparison Galvanic Connection BMW i3 @7.2kW  p.13 Reference Separate Tank to 48V @2.5 kW Galvanically Isolated Connection Shorter Nested L-L @2 kW Large Pot @2 kW Split Nested L-L @2.6 kW Plugless Power @3.3 kW  p.6 estimate Note: Table made from bar graph.
Compared to the BMW i3's galvanic charger, Plugless Power's market inductive charger, and a production model transformer with a discrete resonant inductor for a market rectifier, the shorter nested LL core performed exceptionally. The solution does rely on the autonomous vehicle to precisely nest the two windings, though, so we wanted to see if we could understand and increase the margin for variance.
After achieving our 95% efficient design goal, we optimized our design for market-friendliness considering our use model.
Nested Winding Gap
Our first optimization goal was to see how a gap between the nested primary and shorter secondary windings would affect the results. While keeping other factors constant, we tested winding gaps of 3, 4, 6, and 8mm and core gaps of 6, 8 and 12 mm. We used Nomex for 3mm of each gap as the plastic insulation; the rest of the winding gap was air.
Figure 11 shows the results of a parametric study of the influence of radial winding gap and magnetic path length core gap on the magnetizing and leakage inductances. Variables were kept constant except for winding and core gaps. The data show a linear relationship between winding gap and leakage inductance. Experimentally, although a 6mm gap creates a slightly too high leakage inductance, we can interpolate that a 5mm radial winding gap, to provide a leakage inductance of just over 20[micro]H, would be acceptable. This 5mm radial winding gap translates into a 2 mm wall thickness for each of the stationary and vehicle sides of the interface and a 8mm gap in the magnetic path. A 2mm wall should provide good durability and a magnetic path length that delivers acceptable operational parameters for our LLC resonant circuit.
Nested Winding Overlap
The second goal was to get uniform magnetizing inductance independent of the degree of nesting between the two winding elements. Figure 12 shows the apparatus we used. The secondary winding was indexed in 5mm increments from completely overlapped to not overlapped at all. Figure 13 shows the results.
Figure 13 shows that the useful region of overlap is between 0mm and 25mm, delivering leakage inductance that varies in a range from 7[micro]H to 15[micro]H.
Split Nested L-L Transformer
To reduce conduction losses and improve efficiency further, we introduced a transformer design with the nested windings split on both legs of the L-L core, Figure 14. The efficiency of this transformer is almost as high as the single nested L-L, with 96.6% DC-DC efficiency at 2.64kW peak power. Figure 10 shows the wall-to-battery power efficiency data. This transformer topology did perform better than existing galvanic and inductive chargers.
A hands-free, ground-level, EV docking, inductive battery charger can deliver charging efficiencies comparable to galvanically connected chargers--with no degradation in performance or safety when covered with a variety of environmental contaminants. Two different transformer constructions delivered winding and magnetic path length geometries to support large gaps--large enough to accommodate robust insulation on both ground-side and vehicle-side surfaces.
Ninety-five percent efficient, near ground level galvanically isolated inductive charging shows promise for night charging of EVs at apartments, thus providing high public benefit with minimum public infrastructure expense. Autonomous vehicle capability can indeed be applied to deliver hands free inductive charging while leaving a much smaller carbon footprint than do products on the market today.
With an understanding of the parameters that can be applied to deliver very efficient close coupled inductive charging, we can go forward on the size study working with a 2mm wall thickness for each of the insulators. Future focus is toward scaling this work up into a product that addresses market needs--similar to as shown in figure 1--without losing any efficiency.
An accepted upper power level for home charging of Battery Electrical Vehicles is 10kW . Interleaving two of these transformer circuits should deliver between and 5 and 10 kW--based largely on thermal limitations. Applying what we have learned here to redraw the stationary side of figure 4, using a nested approach with two transformers, we obtain an interleaved power train arrangement. This arrangement can be configured with either one or two power trains provisioned depending on customer preference. The primary side of this arrangement is shown in figure 15.
Additional volume is provided to deliver a self-alignment function that can be integrated into vehicles with a variety of industrial design languages. To achieve alignment without subjecting the nesting insulators to abrasion, the flat top surface first puts the docking head in planar alignment with the vehicle mounted power receiver. Then a 30-degree wedge followed by a flat on each side of the head locates the mating parts before nesting occurs. Only one degree of freedom remains for the last 40-50 mm of travel. Optical/Infrared communication ports verify alignment and provide redundant communication about battery bus voltage and current.
With the two nesting areas for windings biased toward the center, the return magnetic path is biased outwards and provided on the vehicle side of the interface. Energy transfer with 5 to 10 kW of power capability is provided by an active area 225mm wide and 50mm tall.
The need for magnetic return paths when applied to Figure 1, with the car subtracted, obtains Figure 16. We decided to configure the power receiver to simulate a dual exhaust - a survival feature that customers less familiar with electric cars will process without too much hesitation. This will be the lead concept and be used as an efficiency benchmark for alternative constructions.
Our goal is 7.2kW delivered with the structure shown in Figure 17.
After confirming that efficiency is maintained in operation, we will confirm EMI performance in the GE Critical Power FCC Certified EMI chamber. Our hypothesis is the pot core will be better for EMI performance.
Another focus, for us to be able to bring this design into production, is cost. The large pot core was extremely expensive, and all the excess winding space is not necessary. The key element that makes it work is the large cross-sectional area. Therefore, we will experiment with stacking E-E cores side-by-side to increase the cross-sectional area. Our thesis is that this will emulate the pot core and will provide the magnetizing inductance we need to deliver good battery charging performance at a much lower cost.
(1.) Union of Concerned Scientists (UCS) and Consumer Reports, "Electric Vehicle Survey Methodology and Assumptions: American Driving Habits, Vehicle Needs, and Attitudes toward Electric Vehicles," 2013, http://www.ucsusa.org/evsurvey, accessed Aug. 2016.
(2.) Kettles, D., "Electric Vehicle Charging Technology Analysis and Standards," Florida Solar Energy Center, FSEC Report Number: FSEC-DR-1996-15, 2015, http://www.fsec.ucf.edu/en/publications/pdf/FSEC-CR-1996-15.pdf accessed July 2016.
(3.) Idaho National Laboratory, "Plug-in Electric Vehicle and Infrastructure Analysis," INL/EXT-15-35708, 2015, https://avt.inl.gov/sites/default/files/pdf/arra/ARRAPEVnInfrastructureFinalReportHqltySept2015.pdf accessed June 2016.
(4.) Onar, O.C. and Jones, P.T., "Wireless Charging of Electric Vehicles," Oak Ridge National Laboratory Presentation at U.S. DOE Vehicle Technologies Office 2015 Annual Merit Review and Peer Evaluation Meeting, Jun. 2015, http://energy.gov/sites/prod/files/2015/07/f24/vss103_onar_2015_o.pdf accessed July 2016.
(5.) Idaho National Laboratory, "PLUGLESS[TM]Level 2 EV Charging Systems (3.3kW) by Evatran Group Inc.--Results from Laboratory Testing off-board the vehicle," INL/MIS-14-33980, 2015, https://avt.inl.gov/sites/default/files/pdf/evse/PLUGLESSEvatranStandaloneTestResultsFactSheet.pdf accessed July 2016.
(6.) WiTricity[R], "WiT-3300: WiT-3300 Development Kit," http://witricity.com/products/wit-3300/, accessed Jul. 2016.
(7.) Lu, F., Zhang, H., Hofmann, H., and Mi, C., "A Double-Sided LCLC-Compensated Capacitive Power Transfer System for Electric Vehicle Charging," IEEE Transactions on Power Electronics 30(11):6011-6014, 2015, doi:10.1109/TPEL.2015.2446891.
(8.) Onar, O.C., Miller, J.M., Campbell, S.L., Coomer, C., "Oak Ridge National Laboratory Wireless Power Transfer Development for Sustainable Campus Initiative," Transportation Electrification Conference and Expo, 2013, doi: 10.1109/ITEC.2013.6574506.
(9.) Carlson, B., "Idaho National Laboratory's Electric Vehicle Charging Infrastructure Laboratory," INL/MIS-15-36953, 2015, https://avt.inl.gov/sites/default/files/pdf/presentations/VSATTOctober2015EVIL.pdf accessed July 2016.
(10.) Deng, J., Li, S., Hu, S., Mi, C.C., "Design Methodology of LLC Resonant Converters for Electric Vehicle Battery Chargers," IEEE Transactions on Vehicular Technology 63(4):1581-1592, 2014, doi:10.1109/TVT.2013.2287379.
(11.) EV Obsession, "Robotic Electric Car Charging Station From...Volkswagen?" http://evobsession.com/robotic-electric-car-charging-station-from-volkswagen/, accessed Jul. 2016.
(12.) Bosch Media Service, "Accident-free parking: Bosch technology makes anyone a professional parker," http://www.bosch-presse.de/pressportal/en/bosch-technology-makes-anyone-a-professional-parker-44808.html, accessed Jul. 2016.
(13.) GE Critical Power, "NE050AC48ATEZ Infinity Rectifier," Fact Sheet NE050AC48TEZ-FS, Rev. 06/2016, http://apps.geindustrial.com/publibrary/checkout/NE050AC48ATEZ-FS?TNR=Data%20Sheets%7CNE050AC48ATEZ-FS%7CPDF&filename=NE050AC48ATEZ-FS%206-3-16.pdf accessed Oct. 2016.
(14.) Fontana, E., "Passengers First Light Truck - A Modern Take on a Narrow Wake," SAE Technical Paper 2016-01-1333, 2016, doi:10.4271/2016-01-1333.
(15.) Dowell, P.L., "Effects of eddy currents in transformer windings," Proceedings of the Institution of Electrical Engineers 113(8):1387-1394, 1966, doi:10.1049/piee.1966.0236.
(16.) Perry, M.P., "Multiple Layer Series Connected Winding Design for Minimum Losses," IEEE Transactions on Power Apparatus and Systems 98(1):116-123, 1979, doi:10.1109/TPAS.1979.319520.
(17.) Vandelac, J.P. and Ziogas, P., "A novel approach for minimizing high-frequency transformer copper losses," IEEE Transactions on Power Electronics 3(3):266-277, 1988, doi:10.1109/63.17944.
(18.) Taguchi, G., "Introduction to Quality Engineering: Designing Quality into Products and Processes," (Quality Resources, 1986), ISBN 978-9283310846.
(19.) Jiang, H., Brazis, P., Tabaddor, M., and Bablo, J., "Safety Considerations of Wireless Charger For Electric Vehicles--A Review Paper," IEEE Symposium on Product Compliance Engineering, 2012, doi:10.1109/ISPCE.2012.6398288.
(20.) SAE International Surface Vehicle Information Report, "Wireless Power Transfer for Light-Duty Plug-In/Electric Vehicles and Alignment Methodology," SAE Standard J2954, Iss. May 2016.
(21.) ST Microelectronics, "An introduction to LLC resonant half-bridge converter," AN2644, 2008.
(22.) De Simone, S., Adragna, C., and Spini, C., "Design guideline for magnetic integration in LLC resonant converters," International Symposium on Power Electronics, Electrical Drives, Automation and Motion, 2008, doi:10.1109/SPEEDHAM.2008.4581225.
(23.) Rhodes, S.A., "The effects of separable cores on high power transformer design," M.S. thesis, Department of Electrical Engineering and Computer Science, Massachusetts Institute of Technology, 1996.
(24.) Ward, A., Liker, J.K, Cristiano, J.J., and Sobek, D.K.II, "The Second Toyota Paradox: How Delaying Decisions Can Make Better Cars Faster," Long Range Planning, 1995, doi:10.1016/0024-6301(95)94306-J.
(25.) Pugh, S., "Concept Selection: A Method That Works," Proceedings of the International Conference on Engineering Design, 1981.
(26.) EV Obsession, "Electric Car Charging Capabilities--Comparison of 27 Models" September 8, 2015, http://evobsession.com/electric-car-charging-capabilities-comparison-of-27-models/, accessed Dec. 2016.
For further information, please contact Ed Fontana by phone at 214-364-3184, or email at email@example.com
We would like to thank many of our GE Critical Power compatriots in Plano for their help on this project: electrical supervisor Raymond Rene, lab technicians Francisco Chavez, Gary Kirkpatrick, and Danny Winget, machinist Hal Williams, electrical engineers Damen Toomey and Mahitha Velagapudi, software engineers Darwin Smith and Juan Zhang, test engineer Angelo Fiorelli, sourcing specialist Rick Skoch, magnetics design intern Madeline Jasper and mechanica engineers Randy Heinrich and Khanh Nguyen.
Edward C. Fontana, Rick Barnett, Robert Catalano, James Harvey, Jiacheng He, George Ottinger, and John Steel
GE Critical Power
Table 1. Design objectives for electric vehicle charging Characteristics Value Input voltage range 390V - 420V Nominal input voltage 400V Output voltage range 280V - 420V Maximum output power 3kW Desired efficiency 95% Desired magnetizing inductance ([L.sub.M]) 110[micro]H Desired leakage inductance ([L.sub.L]) 10[micro]H Table 2. Transformer construction specifications Shape L-L E-E Toroid Pot EPCOS Magnetics DMR40 EPCOS Core Number Stackable Inc. H160X9 PM U-Cores 0R4820EC 0X20P 114/93 Area (m[m.sub.2]) 761 392 665 1720 Magnetic Path 170 184 392 200 Length (mm) Volume (m[m.sub.2]) 129,370 72,300 261,144 344,000 Winding Leg 45 56 392 63 Length (mm)