Development of a High Speed HTS Generator for ... - IEEE Xplore

1656

IEEE TRANSACTIONS ON APPLIED SUPERCONDUCTIVITY, VOL. 19, NO. 3, JUNE 2009

Development of a High Speed HTS Generator for Airborne Applications K. Sivasubramaniam, T. Zhang, M. Lokhandwalla, E. T. Laskaris, J. W. Bray, B. Gerstler, M. R. Shah, and J. P. Alexander

Abstract—General Electric, under contract with the Air Force Research Labs (AFRL), has successfully developed and tested a high speed, multimegawatt superconducting generator. The generator was built to demonstrate high temperature superconducting (HTS) generator technology for application in a high power density Multimegawatt Electric Power System (MEPS) for the Air Force. The demonstration tested the generator under load conditions up to 1.3 MW at over 10,000 rpm. The new MEPS generator achieved 97% efficiency including cryocooler losses. All test results indicate that the generator has a significant margin over the test points and that its performance is consistent with program specifications. This demonstration is the first successful full-load test of a superconducting generator for the Air Force. In this paper we describe the development of the generator and present some key test results used to validate the design. Extrapolation to a higher power density generator is also discussed.

Fig. 1. Schematic of homopolar inductor alternator with HTS field winding.

Index Terms—High power density, high-speed generator, superconducting generator.

I. INTRODUCTION

EVERAL military and commercial applications need 1–5 MW capability in a portable high-power-density electric power generation package [1]–[3]. One approach is to use a high-speed generator directly coupled to a high-speed gas turbine, with high frequency solid-state power conversion. Superconducting technology offers the highest entitlement for power density of the generator, but several engineering challenges remain in making a reliable, light-weight superconducting machine. To address this need, a rugged, high speed, multi-megawatt, HTS generator has been developed by GE for the Air Force Research Lab (AFRL). The generator has been load tested up to 1.3 MW at GE’s high-speed machine test-bed. The generator is based on the homopolar inductor alternator (HIA) [4] topology to obtain power densities greater than 8.8 kW/kg in a robust construction suited for high-speed applications. This paper describes the generator construction and test results.

S

Manuscript received August 16, 2008. First published June 23, 2009; current version published July 15, 2009. This work was supported in part by the U.S. Air Force under Contract FA8650-04-G-2466. The authors are with GE Global Research, Niskayuna, NY 11361 USA (e-mail: [email protected]; [email protected]; [email protected]. com; [email protected]; [email protected]; [email protected]; [email protected]; [email protected]). Color versions of one or more of the figures in this paper are available online at http://ieeexplore.ieee.org. Digital Object Identifier 10.1109/TASC.2009.2017758

II. HIGH SPEED HTS GENERATOR Trade-off studies based on electromagnetic, thermal, mechanical, cryogenic, and reliability considerations have shown that the HTS homopolar inductor alternator is the preferred configuration for a high-speed superconducting machine [5], [6]. The generator comprises a stationary HTS field excitation coil, a solid rotor forging, and an advanced but conventional stator, as shown in the schematic in Fig. 1. The armature consists of liquid-cooled air-gap windings placed within an advanced iron yoke with laminations oriented in three-dimensions to carry flux from one end of the machine to the other. The stationary HTS field coil is placed around the ferromagnetic rotor forging and between two sets of salient poles that are offset circumferentially by one pole pitch at either end. The HTS field coil provides a magneto-motive force (MMF) capability many orders of magnitude higher than a traditional copper coil, enabling an ‘air gap’ armature winding with high flux density in the gap. The HTS coil can either be within the armature winding or between the armature and the stator yoke. Many advantages result from this design as described in earlier papers [5], [6] and repeated here for convenience: • The stationary field coil does not experience the large centrifugal forces that a rotating coil would be subjected to. The coil can now be a simple solenoid around the rotor instead of a more complicated racetrack coil, so the coil support can be much simpler. The thermal insulation between the coil and ambient is also improved because of lack of centrifugal loads and reduced requirements on the coil support. • Without the large forces and resulting strains in the superconducting coil, more reliable HTS coils can be produced based on BSCCO or YBCO coated conductor technology that operates at 30 K or 70 K, respectively, and at peak

1051-8223/$25.00 © 2009 IEEE

SIVASUBRAMANIAM et al.: DEVELOPMENT OF A HIGH SPEED HTS GENERATOR FOR AIRBORNE APPLICATIONS

1657

TABLE I DESIGN PARAMETERS OF AN AFRL HIGH-SPEED HTS HIA

Fig. 2. Homopolar inductor alternator with HTS field winding.

field of 1 Tesla. Further, the reduced ampere-turns required by this machine design would enable considerable reduction in the utilization of superconductor compared to air-core machine designs. • The cryostat of the coil is stationary. There is no need for a transfer coupling to introduce a cooling medium into the rotating cooling circuit. Instead, the coil can be cooled by one of the established, more reliable ways of cooling, including conduction cooling. The vacuum or foam insulation, as required for good thermal insulation, will be stationary and therefore highly reliable. Other parts of the insulation scheme can also be made more reliable without the large ‘g’ forces. • There is no need for a ‘slip-ring’ assembly to transfer current to the coil from a stationary exciter. The voltage across the coil is then more predictable and makes it easier to detect quench and protect the coil with a reliable protection circuit. • There is no need to consider rotating brushless exciters. Fig. 2 shows a CAD model of a prototype design. The cryocoolers and cryogen recondenser unit are mounted on top of the generator in a simple, robust assembly. Table I summarizes the key design parameters of the generator derived from program specifications. A 1 MW demonstrator generator was built to validate key features of this new generator type. The generator has been successfully load tested. Results are summarized in the following sections. III. KEY FEATURES Fig. 3 shows a picture of the demonstrator generator in the high-speed test bed. The power density of the HTS homopolar inductor alternator design is not as high as that of fully air-core designs because it relies primarily on the iron-core rotor and stator yoke to carry the flux, and there is significantly more leakage flux in the interpolar space that needs to be carried through the iron rotor and

Fig. 3. 1 MW generator in test.

stator components. The power density, however, improves significantly with the high flux density of the air-gap armature winding and rotor speed, but moderately with the number of poles, while the stationary field excitation HTS coil is utilized more efficiently than in any other machine design because it excites all pole pairs in parallel. As a result, the field ampere-turns remain constant as the number of poles is increased to enhance the machine power density. The low full-load ampere-turn requirement of the stationary HTS coil greatly simplifies the development of this coil. The coil can be designed with BSCCO or YBCO coated conductor, and our 1 MW demonstrator used BSCCO that operates at current of 150 A at 30 K and peak field of 1 T. The HTS coil is cooled by gravity-fed boiling liquid neon through a cooling tube heat exchanger in contact with the coil outside surface, and the return boiloff neon is re-condensed by a single GM cryocooler. Vacuum is used to thermally insulate the coil, with a total heat load of 40 W. This refrigerator load requires a single-stage GM

1658


coldhead with 75 W capacity at 25 K, but two coldheads may be preferable for increased reliability and serviceability. The air-gap armature winding design utilizes bars which are wound with compacted Litz copper wire turns and cooled by ordinary water or a dielectric fluid flowing inside alumina ceramic cooling tubes. Each bar is wet wound in a precision mold with thermally conductive epoxy and cured. The bars are then assembled and adhesively bonded to the ceramic cooling tubes and G10 cylindrical inner and outer shells using thermally conductive epoxy. The G10 shells on the ID and OD of the armature bars serve as ground wall insulation. The armature winding assembly is bonded to the stator yoke to form a rigid structure capable to withstand fault torques, vibration, and shock loads. The rotor shaft is sealed with ferrofluid seals inboard of the bearings to enable a vacuum of a few torr to be maintained within the active rotor chamber. This is necessary to reduce windage losses in such a high-speed machine. Especially for a salient pole rotor such as we have employed, the rotor windage losses at 10,000 rpm would be too high to sustain without machine damage, and active cooling would have been difficult and heavy. The yoke within the stator consists of laminated blocks of iron-cobalt alloy to enable both lower eddy current losses from the high frequency operation and high magnetic saturation for the high field developed by the field coil. These blocks are also laminated in different directions to build up the total yoke in order to facilitate the transport of flux from the rotor pole radially, axially, and circumferentially, through the armature windings, to reach the opposite pole, which is offset circumferentially from the first pole. Provisions were made for balancing the solid rotor, first at high (full) speed in a balance pit and in vacuum as needed before assembly. Further balancing of the main rotor mass was unnecessary, but balance provisions were made for other portions of the overall drive train after assembly of the machine into the test facility. The generator was fully instrumented for testing: vibration, thermal, electrical. An IR camera with IR window in the stator was employed to read the rotor temperature during operation. Power input to the generator was measured with a torque meter and tachometer at the drive end, and electrical output was measure with voltage, current, and phase readings of the output. A major reason for building and testing our HIA generator has been to verify the models and analysis we employed in design. Because of high nonlinearity of the generator and the complex three-dimensional nature of the flux paths, effects that are considered higher order in conventional machines dominate and characterize the performance of the HTS HIA. Of special concern are the leakage paths, fringing fields, ac losses, ampere-turn requirement, and core losses. A full 3D electromagnetic model has been built to understand the behavior of the machine and optimize the detailed design. Substantial differences in the flux distribution between the linear and non-linear cases, especially the leakage and bucking flux, meant that all analyses had to be performed with detailed non-linear models. For the conceptual analysis, isotropic properties are used throughout, ignoring the effect of laminations. Eddy currents are not considered directly. Analysis was performed in two ways:

1) Static 3D model with imposed armature currents, and field excitation; 2) Time-stepping 3D transient model with coupled external circuit. Power factor and terminal voltage for rated armature currents as a function of field current and load angle are obtained from the magnetostatic analysis. The results are a strong function of saturation. The linear models give vastly different results because of significantly different flux distributions and the effect of the ‘bucking flux’. The results extracted from the magnetostatic runs are then confirmed with the time stepping model. The static and time-stepping models gave initial confirmation of the electromagnetic performance of the preliminary design of the HTS HIA. The saturation levels need to be monitored closely to optimize machine weight. Finite element modeling with coupled external circuit helps identify areas to focus on for optimization. The model also provides the capability to analyse transient conditions associated with load duty cycle, fault conditions, etc. IV. TEST RESULTS A series of standard machine tests were performed on the generator to validate performance and obtained data to use in scale-up studies. Results and findings are summarized here. Throughout the tests, and continuously for several months, the HTS coil temperatures were steadily maintained with the closed cycle neon refrigeration. A. Open Circuit Tests The purpose of this test was to demonstrate the ability of the machine to generate the desired voltage at the terminals and to obtain the no-load saturation curve from test to verify the electromagnetic design of the magnetic circuit. The test simultaneously challenged the ability of: • The superconducting field coil to provide the ampere-turns of MMF to create flux in the generator. • The HTS cryogenic refrigerator to cool the winding including any ac losses penetrating the flux shield. • The rotor permeability and air core flux paths to link the stator winding and provide useful voltage. • The cooling circuits to handle any localized heating effects due to the magnetic field. Open circuit testing up to 300 V line-line rms was performed at 10500 rpm. This voltage would scale to 357 V at 12500 rpm, and 428 V at 15000 rpm, limited by rotor saturation. Fig. 4 shows the open circuit saturation curve from test compared to the predicted curve from the EM models. The test results compare well with prediction up to and beyond the rated voltage. Voltage imbalance between the different phases was less than 1%. At these flux levels, the flux density in the stator yoke is significantly below designed values, resulting in low core losses as shown below. Fig. 5 compares predicted core losses with measured losses. B. Short Circuit Tests The purpose of this test was to determine the short-circuit characteristics of the generator under armature current loading


Fig. 4. Open circuit saturation curve.

1659

Fig. 6. Short circuit characteristics.

Fig. 7. Copper losses during SC test compared to prediction. Fig 5. Core losses during open circuit test compared to prediction.

without power loading, with all the terminals of the generator shorted together. Power input to the shaft under this condition is for overcoming friction and windage, the joule heating losses of the current flowing in the generator armature, and a small level of core losses. Synchronous impedance tests up to armature line currents of 1450 A rms were performed. Fig. 6 shows the short circuit characteristic of the generator obtained from testing compared to model predictions. The results are within a few percent of expected values. Imbalance between the two three phase sets was about 3% even without the use of trim inductors to balance the leakage reactance of the different phases. The maximum imbalance among 9%. all the phases was In conventional synchronous machines, the electrical losses are dominated by the ohmic losses in the copper windings during short circuit runs and core losses in the open circuit runs. Traditionally, loss data from steady state heat runs under ‘zero-excitation’, open circuit, and short circuit are used to segregate losses in the different components using this assumption. In the MEPS HIA generator, significant ac losses in the

armature winding during open circuit conditions and significant core losses during the short circuit runs, coupled with varying cryostat losses, make this procedure inaccurate. An alternate procedure to segregate losses is analysis of heat rejection in the different cooling systems within the generator. The HIA generator has been designed to have dedicated cooling for most of its major loss centers, including armature straight sections, end sections, iron core, and cryostat, with little heat transfer between these sub-systems. Flow and coolant temperature rise from these cooling circuits have been used to obtain the loss breakdown reported here. Any error in this data is due to the actual flow and temperature measurements and cross-talk between the different subsystems. These are assumed to be minimal. Fig. 7 compares copper armature loss obtained from short circuit tests with calculated losses. The results are in general agreement and confirm that the armature has the capability to work at the rated current of 2309 A rms. The cryogenic systems operated without any problems during these runs. One explanation for the lower measured losses at higher current levels is that at these levels the copper windings are hotter than the yoke, resulting in heat transfer from the windings into

1660


Fig. 8. Efficiency of SC generator observed during load test. Fig. 10. Line currents during 1 MW load test.

Fig. 9. Phase voltages during 1 MW load test.

the yoke and out through the yoke cooling. This would also explain the reversed effect at the lower current levels. C. Load Tests The generator was connected to the resistive load bank, ramped up to 10,500 rpm, and the excitation level stepped up gradually till the generator output was 1.3 MW (limited by the test facility). The terminal voltage was 266 V-rms line-line, and line current was 1460 A-rms at the maximum power level. Power factor was 0.985. The generator efficiency computed from the generator output versus the input power plus the rated power of the cryocooler compressor is shown in Fig. 8. The efficiency is about 97% at 1 MW, and approaches 98% (the designed value) as the power is increased towards the designed rating of 4 MW. Steady state heat runs under load were performed up to 1.05 MW. Loss data and temperatures were consistent with those obtained from the no load runs. Voltage and current imbalances were within 2%, and waveforms are as expected and are showed in Figs. 9 and 10.

Fig. 11. Atmospheric windage results to 10,000 RPM.

Minimal windage and pole face heating was observed during the load tests. D. Zero-Excitation Windage Tests Windage tests were performed at five speeds with degraded vacuum in the airgap. Thermal steady-state was achieved at the lower speeds and transient tests were performed at higher speeds due to high rotor and stator temperatures not allowing steady state to be obtained. The torque is measured by the torque meter, and the computed power loss is plotted in Fig. 11. A power curve of the form is fitted through the data. The exponent for windage torque is found as , and since , its exponent of speed is . This is consistent –3 for with our expectation that windage power loss has losses between concentric cylinders with no axial flow. A maximum power loss of 40 kW is measured for a speed of 10,000 rpm in 1 atmosphere pressure, and this shows the need for low pressure for high-speed operation. Also shown in Fig. 11 are plots representing the predictions from two different rotating cylinder loss models [7], [8]. Each


of these models assumes a perfectly cylindrical surface spinning in a thin annular space. The geometry of the MEPS rotor is very different from this, with four salient poles on each end and a lower diameter mid section. The approach here was to use the relationship between the salient pole depth to rotor radius ratio and a smooth cylinder windage multiplying factor, developed by Vrancik [9]. In the case of our rotor, the calculated smooth cylinder windage value determined by the aforementioned windage models, were multiplied by a factor of 5. While the majority of the windage losses occur due to the cylindrical surface, the predictions in Fig. 11 include adding the effects of the two side “disks” that represent the ends of the cylinder. The disk losses (associated with heating due to the air friction on the sides of the spinning cylinder) were calculated using the same model, applicable to a solid circular disk. For the purposes of this calculation, the outer diameter of the salient pole was used as the disk diameter. The data clearly fall within the regime of both windage models, fitting the model of Ren slightly better, especially at the higher speed range. The curve fit also follows the two windage model curves reasonably well, but it is clear that the rate of increase near the higher speeds is not the same as the windage models. Several windage tests were performed at the lowest pressures attainable in the machine, 375–533 Pa. Three of the tests were open circuit heat runs, two were full load tests, and one was a zero excitation test. None were run as an actual windage test, but we were able to take measurements before the field excitation was applied, thus enabling a windage + friction test point at the speed of the particular electrical test. The windage and friction losses varied from about 3.5 to 4.5 kW. The windage and friction measurement includes bearing and ferro-fluidic seal losses. To determine the windage contribution, they must be subtracted because, unlike the 1 atm windage tests, they are not negligible. The ferro-fluidic seal losses were measured by measuring their removal by the ferro-fluidic seal water cooling circuit, and the estimated bearing losses were estimated by scaling from the manufacturer’s stated losses. The losses were quite constant across all the tests. To obtain the windage loss, the bearing and ferro-fluidic seal losses are subtracted from the windage and friction measurement. While there is some scatter in the data, it shows the windage to be about 1.5 kW. The windage and friction measurement includes bearing and ferro-fluidic seal losses. To determine the windage contribution, they must be subtracted because, unlike the 1 atm windage tests, they are not negligible. The ferro-fluidic seal losses were measured by measuring their removal by the ferro-fluidic seal

1661

water cooling circuit, and the estimated bearing losses were estimated by scaling from the manufacturer’s stated losses. The losses were quite constant across all the tests. To obtain the windage loss, the bearing and ferro-fluidic seal losses are subtracted from the windage and friction measurement. While there is some scatter in the data, it shows the windage to be about 1.5 kW. V. CONCLUSION An HTS HIA high-speed generator was designed, built, and tested. It produced 1.3 MW of electrical power under resistive load in a dynamometer test cell at a speed exceeding 10,000 rpm and with a closed cycle neon cryogenic HTS cooling system. This satisfied all specifications of the Air Force contract. Generator operating characteristics were measured and compared to design predictions, leading to the conclusion that the design methods are adequate for a non-linear HIA machine. In addition, valuable data on windage losses of the salient-pole rotor at high speeds and low air pressures, for which no precise models exist, were gathered. ACKNOWLEDGMENT The authors thank the AFRL (Greg Rhoads, project leader) for the support and consultations during the course of this work. REFERENCES [1] P. N. Barnes, G. L. Rhoads, J. C. Tolliver, M. D. Sumption, and Schmeaman, “Compact, lightweight, superconducting power generators,” IEEE Trans. Magnetics, vol. 41, pp. 268–273, 2005. [2] P. J. Masson and C. A. Luongo, “High power density superconducting motor for all-electric aircraft propulsion,” IEEE Trans. Applied Superconductivity, vol. 15, pp. 2226–2229, 2005. [3] P. J. Masson, D. S. Soban, E. Upton, J. E. Pienkos, and C. A. Luongo, “HTS motors in aircraft propulsion: Design considerations,” IEEE Trans. Applied Superconductivity, vol. 15, pp. 2218–2221, 2005. [4] E. Richter, “High Speed Homopolar Inductor Generator With Straight Winding Construction,” US Patent 3,737,696, 1973. [5] K. Sivasubramaniam, E. T. Laskaris, M. R. Shah, J. W. Bray, and N. Garrigan, “HTS HIA generator and motor for marine applications,” presented at the XVII International Conference on Electrical Machines, Chania, Greece, 2006, unpublished. [6] B. Gamble, G. Snitchler, and S. Kalsi, “HTS generator topologies,” in IEEE Power Engineering Society General Meeting, June 2006. [7] W. M. Ren, Windage and Axial Friction Losses of High Speed Generator TGE 2002-073, January 2003, General Electric Internal Report, unpublished. [8] F. M. White, GE Fluid Flow Data Book. Amsterdam, NY: Genium Publishing Corp, April 2003, sec. 408.7, pp. 1–3. [9] J. E. Vrancik, Prediction of Windage Power Loss in Alternators NASA Technical Note D-4849, October 1968.