1MW MULTI-STAGE AIR-CORED PERMANENT MAGNET GENERATOR FOR WIND TURBINES A.S. McDonald1, N. Al-Khayat1, D. Belshaw1, M. Ravilious1, A. Kumaraperumal1 A. M. Benatmane1, M. Galbraith2, D. Staton3, K. Benoit4 & M. Mueller5 1. NGenTec Ltd. , 7/9 North David Street, Edinburgh, 2. Fountain Design Ltd., UK, 3 Motor Design Ltd, UK, 4. Motor Engineer, France, 5 University of Edinburgh, UK CORRESPONDING AUTHOR: Markus.Mueller@ed.ac.uk NGenTec Ltd is developing a novel PM generator for the offshore wind market. The concept has been proven at lab scale, and now it is being scaled up to the MW level, where the technology has greatest potential. Detailed electromagnetic, structural and thermo-fluid modelling is described along with experimental investigation on full scale modules. The experimental work provides confidence in the design and modelling work completed allowing the design to be finalised. & McDonald [..] have demonstrated a PM generator technology, which has the potential to be 30-50% lighter than conventional generators. It exhibits a high degree of modularity and through the use of air-cored windings, there are potential benefits in component manufacture, assembly and O&M compared to conventional PM direct drive generator technology. The generator technology is being commercialised by NGenTec Ltd, a spin out from the University of Edinburgh. The generator concept has been demonstrated at lab scale, but the benefits are mainly applicable at multi-MW level. This paper describes some of the development of a 1MW demonstrator being developed by NGenTec. 1 Introduction 2 Generator Evolution For offshore wind direct drive generators or low-medium speed hybrid systems are being seriously considered by some of the large wind turbine manufacturers, such as Siemens, Nordex, Alstom, Areva, Gamesa and Vestas. The permanent magnet synchronous generator (PMSG), rather than the field wound synchronous generator (FWSG) as developed by Enercon,dominates in these developments. The PMSG topology used by the manufacturers represents an incremental change from the FWSG, in that the electromagnetic poles in the FWSG are replaced by permanent magnet poles. The stator configuration of the PMSG is exactly the same as for a FWSG, namely pre-formed copper coils are inserted into a ferromagnetic laminated structure to form a 3-phase distributed winding. There are two main benefits of the PMSG over the FWSG: an increase in power density and a reduction in rotor losses due to no rotor field current. However, there are significant challenges with the PMSG particularly during component manufacture and generator assembly. The design of both PMSG and FWSG generators has been well documented [1, 2 & 3]. It is well known that such conventional direct drive generators are very heavy, with the mass of the structural support dominating. McDonald [4] showed that the structural mass could be as much as 80% of the total mass. A generator in excess of 6MW will weigh more than 100 tonnes, and even more for a FWSG. Installation of such generators in the offshore environment becomes a challenge. Should a fault occur on the generator, the whole generator will have to be removed, leading to long downtimes and the use of expensive O&M vessels. Mueller The evolution of the generator topology concept is described in reference .. In this section the authors will outline the learning achieved thus far through the demonstration of small scale lab prototypes. Based on this learning a 1MW demonstrator has been designed and is currently being built for testing in March 2012. Keywords: PM generator, wind energy Abstract 2.1 20kW, 100rpm, Radial Flux Prototype In order to prove the concept a 20kW radial flux machine was built. Figure 1 shows how the rotor and stator were assembled for this machine. Mild steel blocks were cut and machined to allow mounting of the permanent magnets. These blocks were then assembled into a c-core, which were mounted to form a rotor leaving an opening between the c-core limbs for the aircored stator. Coil blocks consist of a single concentrated coil wound on a former and potted in epoxy. The coil blocks are mounted to form a ring of coils, which can be inserted manually into the rotor opening using in this case an engine hoist. As there is no iron in the stator, no forces exist during assembly. 2.2 Multi-Stage Axial Flux Machine The generator topology can be both radial and axial. Due to the high degree of modularity with this topology there a number of design options, which are shown in Figure 2. In each of the examples shown in Figure 2, there are 4 separate generators each consisting of a number of PM c-core modules and single coil modules. In Figure 2(A), four generators are mounted at the same radius supported on two discs external to Figure 1: Assembly process of the rotor and stator for a 20kW radial flux machine Figure 3: Rotor modules assembled onto the rotor structure the air-cored stator. Such an arrangement leads to a long axial length and it is also difficult to access the stator coils should a fault arise. The axial length is reduced in Figure 2(B), with pairs of machine mounted at the same radius. However, each pair of machines will have a different design. In addition the problem of access to the internally mounted stator coils is still an issue. Figure 2(C) shows a multi-stage axial machine consisting of 4 generators of the same design, but the stator is mounted internally making access difficult. An outer stator multi-stage axial flux machine is shown in Figure 2(D), in which full access is now available. The topology in Figure 2(D) is the preferred option for the development of multi-MW generators. A 25kW 3-stage machine was designed, built and tested [ref IET RPG] in order to investigate the multi-stage axial flux electromagnetic and thermal characteristics. Rather than build up the machine of individual PM c-core and coil modules, the 4 c-core modules shown in cross section in Figure 2(D) could be merged in the axial and circumferential directions to form a multi c-core module, reducing the number of parts. Such modules could be cast, making them very cheap to manufacture in volume. Figure 3 shows the outline of one such cast module built for the 25kW 3-stage machine. Figure 2: Multi-stage topology options A number of coils could be potted into one mould to form a multi-coil module as shown in Figure 4. Figure 4: Stator coil module In essence the generator topology developed provides a high degree of flexibility to the machine designer, not available in the conventional PMSG. Figure 5 shows the complete machine installed on a test rig at the University of Edinburgh, but with only one stator stage installed. The other two stator stages were installed in-situ on the test rig to demonstrate the ease of assembly due to the lack of magnetic attraction forces between the stator and rotor. Subsequently it was found that two stator modules in one of the stages had to be replaced, which was done by hand. More details of the design and testing of this machine can be found in refs [5,6 & 7] Figure 5: 25kW prototype with only one stator stage installed 2.3 1MW Demonstrator. NGenTec’s generator technology was specifically developed for large diameter low speed multi-MW machines. The concept and specific aspects of the design and manufacture have been demonstrated at laboratory scale. The next stage in the technology roadmap is to scale up to MW levels. Ultimately NGenTec’s generator is being aimed at the offshore wind market, at ratings of 6MW and greater. In order to demonstrate the potential at this rating, a slice of a 6MW generator is being built. The rating was chosen to be 1MW, but at the same rotational speed of a 6MW, 12rpm, and at the same outer diameter as would be expected at 6MW, 6.4m in this case. Figure 6 shows a CAD model of the 1MW, 12rpm prototype with one stator and rotor module highlighted. From this it can be seen that there are 4 stages in the machine, so that each stage is rated at 250kW. Copper losses and PM loss due to eddy currents were calculated from the FEA model. Table 1 outlines the losses in the machine at the full load point, which are then used to feed to thermo-fluid analysis. It is assumed that the iron loss is negligible, as there is no iron in the stator winding. The total machine torque with 3 stages operating at rated load and 1 stage with a full 3-phase short circuit was calculated, Figure 8. The steady state value, approximately 1500kNm, was then used in the structural analysis. . Figure 7(a) 3D FEA model . Figure 6: CAD model of the 1MW, 12rpm prototype . 3. Modelling the 1MW Prototype In previous prototypes the focus was very much on proof of concept and verification of the electromagnetic design tool. As a result of scaling up both in terms of power rating and physical diameter, the design challenge is structural and thermal. Although no magnetic attraction forces exist the stator and rotor structure have to withstand the rated torques and extreme torques produced during fault conditions, such as a short circuit fault. In terms of thermal performance the machine has to remove more heat dissipated in the windings, whilst the airflow due to the rotation of the machine is much less. Detailed structural and CFD modelling has been undertaken, and utilising results of loss distribution and fault torques from detailed electromagnetic modelling. Such detailed modelling is important to build up confidence in the complete electromagnetic, structural and thermal design before committing to manufacture. 3.1 Electromagnetic Modelling Using the design tools and experience from the small scale prototypes, a 1MW demonstrator was designed. In order to verify the design tool further 3D finite element modelling (FEA) was undertaken using the design tool to provide the main dimensions. NGenTec collaborated with Motor Engineer from France on the em modelling [refs], using Flux 3D [ref – CHECK]. Figure 7 shows the 3D FEA model for a 5 stage machine, the magnetic field distribution due to PM excitation only, and the resulting no-load emfs. The latter is for the 1MW 4-stage axial flux machine. Figure 7(b) Magnetic Field Distribution in one rotor module Figure 7(c) No load induced voltage PIN (kW) N (rpm) T (kNm) I rms (A) Joule Loss (kW) Magnet Loss (kW) 1002.8 12 798.2 84.3 51.2 2.71 Table 1: Loss calculation at full load Figure 8(a) Rotor Module Concept 1 Figure 8: Rated torque in 3 stages and a full 3-phase short circuit torque in stage 1 3.2 Structural Modelling The structural modelling was performed in-house by NGenTec engineers using ANSYS [ref]. Both the stator and rotor modules were modelled in order to finalise their structural design, and the overall structure required to support the stator and rotor modules was also modelled. Figure 8(b) Deflection in rotor module due to rated torque and magnetic attraction forces Within a single rotor module the steel plates making up the module experience magnetic attraction forces as well as torque. Figure 8(a) shows the first rotor module concept mounted on the rotor support structure and Figure 8(b) shows the deflections calculated in ANSYS. A maximum deflection of 0.8mm into the airgap is calculated. Manufacturing tolerances will lead to more deflection, so a modified concept was developed to reduce the deflection due to magnetic forces and torque. Figure 9 shows a ribbed structure on the external place. The total module mass has increased slightly (2.5% increase), but the deflection is now only 0.26mm. The stator modules are cantilevered from the stator support structure as shown in Figure 10. There are no normal magnetic attraction forces between the stator modules and the permanent magnet rotor plates, but the modules and structure have to be designed to react fault torques, such as the condition shown in Figure 8. Stress analysis performed using ANSYS indicates acceptable stress levels and deflections. Harmonic analysis showed no modal vibration issues. Figure 11 shows the deflection of the stator modules held fixed at the mounting points when under rated torque conditions, 800kNm. Figure 9: Rotor Module Final Concept The resulting deflection with both torque and gravity combined is shown in Figure 12(b). It should be noted that the dummy stator blocks were modelled as separate blocks with a small gap between each one so as to closely model the real situation. The peak deflection is of the order of 0.5mm, and further modelling showed that this was mainly due to the torque. Gravity had negligible impact. 3.3 Thermo-Fluid Modelling The stator support structure model is shown in Figure 12(a), with the load conditions A-D shown in the figure. The individual stator coil modules were not modelled in this case. Dummy stator blocks exhibiting the same mass and centre of gravity were modelled instead. A moment of 1500kNm was applied to the faces of the dummy stator blocks, which represents the steady state condition shown in Figure 8. The loss in the machine is dominated by copper loss, including eddy current losses in the winding. The coils are potted in epoxy, which does not exhibit good thermal conducting properties. Understanding the thermal performance is therefore very important for this machine. Figure 10: Stator modules cantlivered from stator support structure NGenTec collaborated with Motor Design Ltd [ref] to develop a thermal model of the machine using Portunus [ref].The thermal model was based on the T-network cube element [ref Mellor & Wrobel]. Computational fluid dynamics was used to determine heat transfer coefficients for use in the thermal model. The permanent magnet rotor acts as a very badly designed fan to stir the air in the machine and provide some natural cooling effect. At rated load the Joule losses are expected to be 51kW. This translates to 2.32kW per module consisting of 4 stator blades. Using ANSYS Fluent the heat flux was applied to the stator modules. The resulting temperature on the blade surface and within the coils is shown in Figure 13 indicating a maximum temperature of 64°C. An airflow of 0.3m3/s is assumed to flow across the coils within the machine. Figure 13: Thermo-fluid modelling of a full scale 1MW stator blade. Figure 11: Stator module deflection under rated torque conditions. The heat transfer coefficient was calculated from the CFD and used in the Portunus model, resulting in a maximum winding temperature at the bottom of the stator coils of 64°C. 4. Experimental Investigation Figure 12(a) Stator support structure load conditions 4.1 Structural Performance Once the 1MW generator is complete deflections will be monitored during testing. The structural design of the rotor module was finalised based upon the ANSYS modelling. The materials used in the rotor module are well understood in terms of deflection under the expected forces. However, the stator module consists of copper coils in epoxy in which the temperature increases with loading. The stator module has to be able to react the torque producing force. There is no magnetic attraction force acting on the stator modules because it is supported in a non-magnetic material. Current flowing in the winding produces heating, which will affect the structural properties of the support material. Figure 14 shows an experimental arrangement of a stator module undergoing deflection tests under a fault condition. The measured deflection is well within the safe deflection specified, and the analysis provides a good estimate of the expected deflection. 4.2 Thermal Performance A test rig of a full scale module was built as shown in Figure 15. The coils were instrumented with thermocouples, and a fan was installed to investigate forced cooling. Figure 12(b) Stator support structure deflection under fault conditions. investigations on full scale modules provide confidence in the final design, which is now being built, with testing expected in February and March 2012. Figure 14: Deflection investigation of stator modules under fault conditions. With a current flowing in the coils corresponding to full load and a module dissipation of 2.2kW the maximum temperature without any forced cooling was measured to be 104°C. Table 2 shows the maximum temperature for different flow rates provided by the fan. At the time of the test the ambient temperature was 20°C. The maximum temperature was observed at the bottom of the coils in the same area as observed in the CFD modelling. Flow rate (m3/s) Maximum Temp (°C) 0.3 62.3 0.55 55.0 0.65 51.8 Table 2: Maximum temp at different flow rates for 2.2kW/module Figure 15: Full scale module test rig. At rated load the winding temperature for the various flow rates shown in Table 2 are all well below the peak operating temperature for electrical machines. However, the winding has to withstand fault condition also. Figure 16 shows the temperature results for an emulated fault condition lasting 100s, with and without forced cooling. For forced cooling a flow rate of 0.3m3/s was chosen. I both cases the temperature rise does not exceed 100°C. In the final 1MW prototype demonstrator the performance of the complete machine will be investigated with and without forced cooling. 5. Conclusion The authors have described the learning that has taken place to develop a 1MW prototype generator. Experimetnal Figure 16: Short circuit test; coil temperature with and without cooling. Acknowledgements The authors would like to thank NGenTec Ltd for permission to publish, DECC for funding some of the work, SMART Scotland for funding the 25kW prototype, Scottish Enterprise for funding the original concept, and the University of Edinburgh for providing facilities.. References 1. CJA Versteegh. Low speed direct drive PM generator for application in the Zephyros Z72wind turbine. IEE Seminar on Electric Aspects of Offshore Renewable Energy Systems,NaREC, Blyth, Northumberland, UK, February 2004. 2. A Grauers. Design of direct-driven permanent-magnet generators for wind turbines. PhD dissertation, Department of Electric Power Engineering, Chalmers University of Technology, Göteborg, Sweden, 1996. 3. H. Polinder, F.F.A. van der Pijl, G.-J. de Vilder, and P. Tavner, Comparison of direct-drive and geared generator concepts for wind turbines, Proc. 2005 IEEE International Conf. onElectric Machines and Drives, San Antonio, TX, May 15, 2005, pp. 543-550. 4. AS McDonald, Structural analysis of low speed, high torque electrical generators for direct drive renewable energy converters. PhD dissertation, School of Engineering, University of Edinburgh, 2008. 5. Y Chong, D Magahy, J Chick, M Mueller, A McDonald, D Staton, Numerical Modelling of an Axial Flux Permanent Magnet Machine for Convection Heat Transfer, 1st IET Renewable Power Generation Conference, Edinburgh, Sept 2011. 6. M Mueller, A McDonald, M Benatamane, A Multi-stage Axial Flux Permanent Magnet Machine for Direct Drive Wind Turbines, 1st IET Renewable Power Generation Conference, Edinburgh, Sept 2011. 7. Chong et al – PEMD paper