This document describes improvements made to the design of a low-speed, high-torque, direct-drive permanent magnet generator for wind turbines. Finite element analysis and optimization techniques were used to reduce the starting torque and improve performance. A prototype was built and tested, confirming satisfactory operation. Compared to the previous prototype, the new design has lower mass, lower starting torque, and improved efficiency.
This document describes improvements made to the design of a low-speed, high-torque, direct-drive permanent magnet generator for wind turbines. Finite element analysis and optimization techniques were used to reduce the starting torque and improve performance. A prototype was built and tested, confirming satisfactory operation. Compared to the previous prototype, the new design has lower mass, lower starting torque, and improved efficiency.
This document describes improvements made to the design of a low-speed, high-torque, direct-drive permanent magnet generator for wind turbines. Finite element analysis and optimization techniques were used to reduce the starting torque and improve performance. A prototype was built and tested, confirming satisfactory operation. Compared to the previous prototype, the new design has lower mass, lower starting torque, and improved efficiency.
For Wind turine! W" Wu #"S" $am!den CSIRO Telecommunications & Faculty of Engineering, UTS Industrial Physics PO Box !", Broad#ay PO Box !$, %indfiled &S' !()( &S' !(() *ustralia *ustralia T" %raw&ord, G. Hi'' 'est#ind Tur+ines Pty %td !, O#en Road, -elmscott 'estern *ustralia . ( *ustralia A!tract-There i! a mar)et &or !ma'', e&&icient and co!te&&ective wind generator! &or mini-grid and remote power !*!tem!" Direct-drive permanent magnet generator! have ecome ver* attractive &or thi! app'ication" Thi! paper de!crie! the improvement! achieved in an outer-rotor direct-drive permanent magnet generator * u!ing &inite e'ement ana'*!i! and optimi!ation technique!" The !tarting torque o& the generator i! !tudied" An optimi!ation routine &or the de!ign, inc'uding magnetic &inite e'ement ana'*!i! and 'umpedparameter therma' mode', i! pre!ented" A protot*pe &or +, )W, +(( rpm generator wa! ui't" The te!t re!u't! with a re!i!tive 'oad con&irm the !ati!&actor* operation o& the generator" %ompared with the previou! protot*pe, the new de!ign ha! 'ower ma!!, 'ower !tarting torque and improved e&&icienc*" ("-.T$/D0%T-/. Traditionally #ind tur+ine generators ha/e used gear+oxes and 0itch control to allo# constant high1s0eed generation under /arying #ind s0eed conditions2 In recent years contem0orary 0o#er electronics of high efficiency, high relia+ility and decreasing cost offers the o0tion to change the 0o#er fi1e3uency out of the generator to match the system fre3uency, #hich leads to the idea of /aria+le s0eed direct1dri/e generators2 * num+er of alternati/e conce0ts ha/e +een 0ro0osed for direct1dri/e elecmcal generators for u!e in gridconnected or stand1alone #ind tur+ines 4,!52 Com0ared to a con/entional gear+ox1cou0led #ind tur+ine generator, a direct1dri/e generator has reduced o/erall si6e, lo#er installation and maintenance cost, a flexi+le control method and 3uic7 res0onse to #ind fluctuations and load /ariation2 * direct1dri/e generator must +e light and efficient to minimise the re3uirements for the to#er structure and to maximise elecmcal 0o#er extracted .1om the #ind2 For small #ind turine!, direct1dri/e 0ermanent magnet generators ha/e +ecome /ery attracti/e +ecause of their high efficiency, high 0o#er density and ro+ust rotor structure2 The attracti/eness of direct1dri/e 0ermanent magnet generators is further enhanced +y im0ro/ements of 0ermanent magnet characteristics and decrease of material 0rices2 Some directdri/e exam0les are Enercon 8E!,"( 7'9, Pro/en 8!2: 7'9, %;' 8!2:1( 7'9 and <enco1'est#ind 8!2:1( 7'9 123" * =oint effort to de/elo0 a !( 7' lo#1s0eed, high1 tor3ue, directdri/e 0ermanent magnet generator for #ind tur+ines #as initiated +y the Uni/ersity of Technology Sydney 40TS5 and Common#ealth Scientific and Industrial Research Organisation 8CSIRO9 in con=unction #ith the *ustralian Coo0erati/e Research Centre for Rene#a+le Energy 8*CRE9 and <enco1'est#ind2 * nono0timised, >$10ole, )( r0m 0rototy0e #as constructed +y <enco1 'est#ind earlier 4>52 It features a radial1flux, slotted1stator to0ology #ith outer1rotor and surface1mounted &d1Fe1B magnets, as sho#n in Fig2 2 The magnets are +onded to the inner surface of a steel drum that rotates around a stationary stator #ith con/entional three0hase #indings2 An ad/antage of this arrangement is that the centrifugal force of the rotating magnets a00lies a 0ressure to the +onding media, therefore increasing the relia+ility of the glued =oint2 *lso the +lades of the #ind tur+ine are directly mounted on the fi1?nt surface of the outer1rotor drum, #hich leads to a sim0le assem+ly 0rocess, as sho#n in Figs2 : and .2 This 0a0er descri+es the im0ro/ed design of the second 0rototy0e +y using finite element analysis and com0uter search techni3ues2 Section -( considers starting tor3ue2 The analysis of the direct1dri/e 0ermanent magnet generator is gi/en in Section III. Section I< discusses the design o0timisation and com0ares designs for se/eral num+ers of 0oles, and se/eral lamination and magnet materials2 In Section <, test results of the second 0rototy0e are com0ared #ith 0redicted 0erformance2 The conclusions are summarised in Section I<2 !2ST*RTI&@ TORAUE The starting tor3ue of a 0ermanent magnet generator is the total tor3ue including the 0ea7 cogging tor3ue, hysteresis tor3ue, and the tor3ue necessary to o/ercome the +earing and seal friction of the generator2 Bysteresis tor3ue arises .1om the hysteresis loss of the generator2 The cogging tor3ue is a dorninent com0onent, #hichCis inherently generated from the interaction of the magnets #ith the stator teeth2 For a direct1dri/e #ind generator, the starting tor3ue is an im0ortant design issue +ecause high starting tor3ue 0re/ents o0eration at cut1in #ind s0eed2 *s a conse3uence, it is necessary to reduce the starting tor3ue to acce0ta+le /alues2 Fig2 %ayout of the direct1dri/e 0ermanent magnet generator 0-7803-6401 -5/001$10.00 0 2000 IEEE >) The cogging tor3ue can +e calculated directly for different rotor 0ositions, #hen the stator #inding canies no current and the rernanence of the magnet is 7no#n2 Since the magnet remanence is tem0erature de0endent, the cogging tor3ue /aries #ith the o0erating tem0erature of the generator2 The maximum cogging tor3ue occurs #hen the rotor tem0erature is at room tem0erature2 The cogging tor3ue is affected +y air ga0 length, slot #edge material, s7e#, magnet 0ole arc ratio, and slot o0ening #idth etc2 For the first 0rototy0e, if the air ga0 is increased from (", mm to ("6 mm, the pea) cogging tor3ue is reduced from 7+"7 &m to +7", &m2 By using a magnetic slot #edge, the cogging tor3ue can +e reduced further from +7", &m to 8", &m2 'hen the stator slot is s7e#ed +y one tooth #idth as used in the first 0rototy0e, the 0ea7 cogging tor3ue decreases from .!2. &m to (("9 &m for an air ga0 of (", mm2 The cogging tor3ue is /ery sensiti/e to the 0ole arc ratio as sho#n in Fig2 !8a92 'ith a decrease of (2)! to ,"7 of 0ole arc ratio, the 0ea7 cogging tor3ue increases o/er (, times2 The slot o0ening #idth also affects the cogging tor3ue2 Fig2 !8+9 sho#s the /ariation of 0ea7 cogging tor3ue against the slot o0ening #idth2 It can +e seen that there are a num+er of #ays to reduce cogging tor3ue, ie using a longer air ga0, a magnetic slot #edge, slot s7e#ing, and o0timum magnet 0ole arc ratio and slot o0ening #idth2 For the same out0ut 0o#er, the use of a larger air ga0 may increase the thic7ness of the magnet2 The use of a magnetic slot #edge increases the flux lea7age +et#een ad=acent 0oles and reduces the out0ut 0o#er of the generator2 'hen the slots are s7e#ed +y one slot 0itch, the cogging tor3ue should +e ideally reduced to 6ero, +ut the out0ut is also reduced2 The most effecti/e solutions for cogging tor3ue reduction are the ad=ustment of the magnet m #idth and the slot o0ening #idth2 Pole arc rat10 Therefore, o0tirnising magnet arc and slot o0ening #idth to gi/e a minimum cogging tor3ue #as used in the im0ro/ed design2 The calculation of the o0erating characteristics of the generator is +ased on a finite element analysis of the magnetic field and a lum0ed10arameter circuit analysis of the thermal +eha/iour2 The characteristics of the generator are 0redicted for fixed or /aria+le s0eed o0eration under a +alanced resisti/e load2 A" Back emf and synchronous reactance The +ac7 emf is calculated from the /ariation of the flux lin7age #ith the stator coil #hile the rotor turns2 Fig2 2 com0ares the 0redicted and measured ernf for the first 0rototy0e2 'ithout s7e#ed slots, the 0hase ernf #a/eform sho#s a small di0 in the middle, #hich is attri+uted to the effect of the slot o0ening and dee0 slots2 The 0hase emf #a/eform +ecomes tra0e6oidal #hen the slots are s7e#ed one tooth #idth2 The measured emf sho#s a distorted #a/eform, #hich indicates that one side of the magnets is of lo#er magnetisation than the other side2 The unsymmehical strength of the magnets may +e due to 0artial demagnetisation from a hea/y load test2 The ernf can +e resol/ed into a Fourier series com0rising a fundamental com0onent and a series of odd harmonics, gi/en +y #here 0 is the num+er of 0oles and w is the rotational angular s0eed of the rotor in radls2 E, is the 0ea7 am0litude of the jth harmonic of the emf #ithout s7e#ed slots2 kg is the s7e# factor for the fi harmonic2 Line to line. unskewed, nn! 287 2 V Line to line. skewed. rmr 279.5 V Phase. unskewed, m 166.2 V Phase, skewed. nns 161 4 V 4a5 Pea) c o p g toque again!t po'e an: mtio Slot opening wldth (10'm) 45 Pea) cogging torque again!t !'ot opening width Fig2 ! Pea) cogging torque 0 60 120 180 240 300 3 Rotor poslstlon ele!tr"!al de#ree$ 4a5 ;cted ac) em& 8+9 Mea!ured pha!e em& Fig2 2 %ompari!on o& predicted and mea!ured em& &or the &ir!t protot*pe The synchronous reactance, D,, consists of the armature reactance and the lea7age reactance2 The armature reactance can +e found from incremental finite element analysis [5], #hile the lea7age reactance can +e calculated +y an em0irical formula2 In terms of the 0hase emf, epk synchronous reactance, X,, and 0hase resistance, R,, the generator can +e re0resented +y an e3ui/alent circuit on a 0er10hase +asis2 Thus, for a star1connected 2- 0hase symmetric resisti/e load, the load current, iph, is deri/ed from #here and R, is the load resistance 0er 0hase2 Thus, therms /alue of the load current is =!"- j=1,5,7,1 '""" B. Losses and load characteristics The losses #hich affect the efficiency of the generator are the #inding co00er loss, stator core loss, mechanical loss 0redominantly from earing and seal friction, and !tra* loss, due to edd* current losses in the #inding and magnets2 The co00er loss, 0,, is the 0rinci0al loss in the generator under most o0erating conditions2 It can readily +e calculated from the #inding resistance at the o0erating tem0erature, and is gi/en +y *ssuming a uniform sinusoidal flux density in the lamination, the core loss, p#, can +e ex0ressed as #here p,, is the classical hysteresis loss, p, the classical eddy current loss and 0, the anomalous loss2 These are gi/en +y #here kh, k, and k, are the s0ecific hysteresis, eddy current, and anomalous losses, res0ecti/ely, #hen the 0ea7 flux density B* is T and the fre3uency f is :( B62 '0 is the lamination mass2 The core losses in the stator tooth and yo7e are calculated se0arately2 The mechanical loss, 0, is gi/en +y 2?f Pmec = Tbaa #here Twi! the +earing and seal friction in &m2 The stray loss, pm, #as found to +e "E of the out0ut in the first 0rototy0e2 This /alue is used in the design o0timisation2 From the resultant losses, the efficiency of the generator, v, can +e o+tained from #l= PO xl(( PO $ Pcu $ Pfe $ Pmec $ Pstray #here P, is the out0ut 0o#er and is gi/en +y For a fixed s0eed and o0eration under a +alanced resisti/e load, the load characteristics of the generator can +e o+tained from e3uation 8>9 to 8(92 It is found that the out0ut 0o#er reaches its maximum /alue #hen the load resistance is a00roximately e3ual to Since the generator is directly cou0led to the #ind tur+ine, the o0erational s0eed /aries o/er a range, and hence affects the no1load emf and synchronous reactance2 The efficiencies for /aria+le s0eed o0eration can also +e o+tained fiom the a+o/e e3uations2 C. Thermal analysis The generator is a totally enclosed non1/entilated machine2 *s the generator is located on a to#er out of doors in hot en/ironments, its cooling de0ends not only u0on natural con/ection +ut also radiation and solar a+sor0tion2 The thermal 0erformance is a +alance of heat in0uts from machine losses and solar a+sor0tion, and cooling +y natural con/ection and radiation2 A 0reliminary calculation sho#ed that the heat transfer of radiation and solar a+sor0tion is a00roximately +alanced2 Thus for sim0licity the thermal analysis does not include radiation and solar a+sor0tion2 Ignoring radiation and solar a+sor0tion, the heat generated +y the total loss flo#s to the rotor surface through conduction of the s0ider structure inside the stator, and con/ection across the air ga0 from the stator outer surface, and then dissi0ates +y natural con/ection into the surrounding air2 A lum0ed10arameter model for steady1state analysis is 0ro0osed to re0resent the com0lex distri+uted thermal 0arameters of the generator, as sho#n in Fig2 >2 The lum0ed0arameter circuit can +e sol/ed to gi/e the tem0erature rise at different locations2 . Performance of the non!optimised prototype The 0erformance of the first 0rototy0e #as 0redicted +y the methods descri+ed a+o/e and is gi/en in Ta+le I2 It can +e seen that there is good agreement +et#een the measured results and 0redicted 0erformance, in terms of /oltage, current, tem0erature rise and starting tor3ue2 earth magnets is only a+out E of the total cost92 The running cost #as also found to +e small com0ared #ith the ca0ital cost2 Pcu Pfer Rg2> %um0ed10ammeter model for steady1state thermal analysis R, Fthermal resistance +ehyeen stator #indug and stator tooth R, = thermal resistance +et#een stator #indmg and stator yo7e "# = thermal resistance +et#een stator tooth and yo7e R, $thermal resistance +et#een stator tooth and outer surface Rs = t h d resistance +et#een stator outer surface and mtor inner surface through con/ection and conduction of the air in the air gap R, = % & a l resistance +et#een stator yo7e and inner surface ", = thermal resistance +et#een s0ider structure and +earing Rb = thermal resistance +eh/een +earing and mtor R, = therma' mistance of the mtor outer surface "% Fthermal resistance of the su00orting frame outer surface T*B%E I -C !" TBE F-$ST P$/T/':F; Items ;easured Predicted Pea7 cogging to3ue 8&m9 2$ '' - - ;aximum hysteresis to3ue 8&m9 Starting to3ue 8&m9 R;S cunent 8*9 R;S /oltage ()* Po#er out0ut 8'9 Efficiency (+* Rotor tem0era7m rise 8GC9 Stator #indmg tem0emture rise 8CC9 *cti/e mass 87g9 I<2 HESI@& OPTI;II*TIO& Based on the /alidated modelling 0rocedure, an o0tirnisation routine for the design of a !( 7', ! r0m direct1 dri/e 0ermanent magnet generator #as set u02 It uses ele/en dimensional /aria+les and three cons+1aints2 The three constraints are tem#1ature rise, out0ut 0o#er, and starting tor3ue2 The o+=ecti/e is to minimise the ca0ital cost of the machine including materials, 0o#er loss, su00orting structure and #ind s0eed /ariations2 It is noted that the cost o0timisation effecti/ely minimises the acti/e mass due to the high cost of the to#er and site installation2 The 0arts cost of the It is /ery im0ortant to 0ro/ide an initial feasi+le design for the o0timisation 0rocess, #hich can narro# the design s0ace and therefore sa/e com0utation time2 The design data for the first 0rototy0e #as use# for the initial design2 Hesign o0timisations for se/eral num+em of 0oles, and se/eral lamination and magnet materials #ere 0erformed2 The design results are gi/en in Ta+le ((--#" For a &i<ed num+er of 0oles, eg >$, the design #ith short stac7 length sho#s a minimum acti/e mass or total cost2 'hen the stac7 length is &i<ed, the acti/e mass decreases #ith an increase of the num+er of 0oles2 It has +een found that a +etter grade of lamination material hel0s to meet the starting tor3ue 8hysteresis tor3ue9 re3uirement, #hile a +etter grade of magnet material increases out0ut 0o#er2 The most fa/oura+le design is a .(1 0ole machine #ith %y1core "( lamination and &":SB magnet2 Bo#e/er, this design has a large diameter of (2, m2 A! manufacturing methods 0laced an u00er limit on the outerrotor drum diameter, a ".10ole #ind generator #as selected for the second 0rototy0e2 <alues of the 0rinci0al design detads are gi/en in Ta+le <2 T*B%E n !( -' CiE&FR*TOR HESI@& OFII;R*TIO& FOR HE'JE' &'ER OF PO%ES ($%&!'( U0 %*;I&*TIO& )*+ .nSH %&'()*$ Poles ". >( >$ :> 60 '?ndin?tenumamerise8GC9 : !! !> "( "( . . EfficieK (+* $)2 $.2: $.2) $.2) $.2: Outer diameter 8( m9 .$, ., ),: ,(. ,( OveraU lenK 8(" m) !)> !." !, E ,( *cti/e mass 87g9 : (. ,> $. $! Grade %y1core ++, %y1core 1+, mppn 3,- lamination Cost8*L7g9 ! > > Core loss 8'7g9 !2: 2" 2O Winding &pa%,e rise 8CC9 "( "( "( ?fficieK? (+* $.2: $)2 $)2! Bystensis to3ue 8&m9 ,2! >2 "2> (uterdiameter8(" m9 ,( ," $)) OveraU length 8' m9 ,( $: $) *cti/e mass (-g* $! ), )) Grade &!)SB &""SB &":SB Magnet Cost &-. >( :( .( $emanenceo 2(. 2: 2, Winding temperature rise 8GC9 "( "( "( ? f f i c i e n ? ? (+* $.2: $.2, $.2, Bystemis to3ue 8&m9 ,2! ,2) ,2. Outer diameter 8( m9 ,( () ,). OveraU length 8(" m9 ,( $! $" *cti/e mass (.g) $! $2) $( machine is small # m 0 d #ith the total cost 8eg the cost of rare .otor position me!hani!al de#ree$ Fig2 : C o 0 g to3ue as a function of rotor 0osition T*B%E < P$MPAL D;S-G. HET%U OF TBE SECO&H /'!T!T(/( &um+er of 0oles &um+er of 0hases Outer diameter O/erall axial 'en= Air m %amination material ;agnet &aterial ;agnet thic7ness *cti/e mass 'esis0ance per 0hase at !( /0 Synchronous in#uc0ance per 0hase Synchronous reac0ance at ! r0m &o1load 0hase emf <oltage at U9 .1 out0ut /urrent a0 !( 7' out0ut Efficiency a0 U9 .1 out0ut ;aximum 0o#er out0ut at ! rpm The cogging tor3ue is minimised #hen the magnet arc ratio is (2) and the slot o0ening #idth is "2) mm2 Fig2 : sho#s the cogging tor3ue #ithout s7e#ed slots as a function of rotor 0osition2 'ith s7e#ed slots for one tooth #idth, the 0ea7 cogging tor3ue is :2$ &m2 The hysteresis tor3ue is $2> &m2 Therefore, the starting tor3ue of the generator is .2! &m including ! &m for +earing and seal iiiction2 The measured staring tor3ue of the second 0rototy0e is !( &m2 Fig2 . sho#s the cross section of the generator 1it% its flux lines at no1load2 The 0redicted +ac7 ernf at no1load is sho#n in Fig2 )2 Fig2 $ sho#s the load characteristics of the generator at ! r0m2 For a maximum out0ut 0o#er, the load resistance 0er 0hase is a+out !2!> R2 The /oltage and current #a/eforms for the maximum out0ut 0o#er are sho#n in Fig2 ,2 It is noted that the current for a hea/y load tends to +e inducti/e since the synchronous reactance is much larger that the #inding resistance2 Fig2 ( sho#s the 0erformance as a function of s0eed2 The out0ut 0o#er fiom the #ind tur+ine is 0ro0ortional to the shaft s0eed cu+ed for the s0eed u0 to nominal s0eed2 The generated /oltage is 0ro0ortional to the s0eed s3uared2 Therefore the current is 0ro0ortional to the s0eed s3uared #hen the s0eed is less than the nominal s0eed2 'hen the s0eed is higher than the nominal s0eed, the load current 7ee0s constant2 To test the second 0rototy0e in the la+oratory the direct1dri/e 0ermanent magnet generator #as dri/en +y a ;a6da internal com+ustion engine2 Fig2 sho#s the measured o0en1circuit /oltage #a/eform at no1load, #hich agrees #ell #ith the 0redicted #a/eform sho#n in Fig2 .2 'hile /arying symmetrical "10hase resisti/e load, the /oltage, current, and in0ut 0o#er #ere recorded for different engine s0eeds2 T#o sets of "10hase resistors #ere connected either in star or delta to gi/e ",.,, and $ R for the load resistance2 Fig2 ! and " sho# the measured out0ut 0o#er and efficiency of the generator as a function of s0eed, res0ecti/ely2 The load characteristics at ! r0m can +e extracted 7om the test data and are sho#n in Fig2 >2 'hen the out0ut 0o#er reaches !( 7' at ! r0m, the 0hase /oltage is !(! <, the efficiency ,> + and the load current "" A" The measured /oltage and efficiency are slightly higher than the 0redicted /alues sho#n in Ta+le < for the follo#ing reasonsM $ An a/erage /alue for the remanence of the magnets #as used in the design, #hich may +e less than the actual /alues for the deli/ered magnets2 $ An am+ient tem0erature of :( >% #as assumed in the design as the generator #as assumed on the to#er2 This /alue is higher than that in the la+orato32 Therefore, the o0eration tem0erature of the magnets #as higher than the test condition, #hich results in a lo#er remanence of the magnets2 $ The stray loss #as assumed to +e "E of the out0ut 0o#er, #hich may +e too high for the second 0rototy0e2 The test results of the generator #ith a resisti/e load confirm the satisfactory o0eration of the generator2 The direct1 dri/e generator is currently installed #ith a t#o1+lade tur+ine on a ".m to#er at ;urdoch Uni/ersity, to conduct full site testing2 Figs2 : and . sho# the direct1dri/e generator mounted on the main fiarne and on the to#er, res0ecti/ely2 Fig" & 0lu1 line a0 no1load I . Phase emt. withan skew I Rotor posistion ele!tri!al de#ree$ 6,, . . 4,, Pham em1 2lm skew Phase 2dta#e Load !urrent - 2,, S Line lo ltne em1 mthwt skew Phase em, mthan skew Llne to ltne em3 mth skew 4 + 4 . 56 . 756 . ." 27, *( Rotor poslstlon ele!treal de#ree$ Fig2 ) Redicted no1load ernf #a/eforms at ! rpm Fig2 , <oltage an# a m a t #a/eforms for maximum out0ut 0o## at ! 1 r0m 6utp"t power -. 7nput pawer )8l!9em: ;Volta#e %a#net temp me .mdm# temp rlse - 2. - 175 - - - 15.+ - 2 - 125 g, - - loo 2 Load current &$ Fig2 $ 8 P r e d i c t e d load characteristics at ! ( rpm r 6utput power 7nput power A *ur<lne output power )33l!len!: 0 Load !urrent $ Volta#e %a#net temp. nse r .lndln# temp rlse Fig2 ( Predicted 0erformance again!t speed under resisti/e load Fig" ' - 3easure# opencircuit vo'tage wave&orm - Load re3i3tance 9 $1 2 Lo=d mslstan!e 18 n Y 25 Speed rpm$ Fig" (+ Mea!ured output power again!t spead Fig" (2 ;easured e&&icienc* again!t !peed Direct-drive generator Fig" (8 Direct-drive gemator mcunted on 04e main 4&&e --5,tp,t po1er 5'' Inp,t po1er )33l!len!: 4Volta#e o l . l . l . l " l " l " l " l " l . l " l o 0 10 20 30 40 50 60 70 80 60 100 Load !urrent 4A5 &ig" 14 ;easured 'oad chamcteri!tic! at +( ( rpm 0rototy0e2 Than7s also go to Bo#ard %o/att at CSIRO and Peter 'atterson at 0TS for discussions2 Fig2 . Hirectdi/e generator on 0he to#er <I2 CO&C%USIO& The cogging tor3ue, +ac7 ernf, synchronous reactance and iron loss of a directdri/e 0ermanent magnet #ind generator #ere calculated +y using a finite element analysis2 Com+ined #ith an e3ui/alent circuit and a lum0ed10arameter thermal model, the analysis of the 0erformance under a +alanced resisti/e load #as 0resented and /alidated +y examining the non1 o0timised !( 7' 0rototy0e2 A design o0thisation routine #as a00lied to study a range of designs #ith different num+er of 0oles, different lamination and magnet materials2 The design of a ".10ole machine #as finalised and +uilt, #hich has a larger diameter and shorter axial length than the 0re/ious 0rototy0e2 Test results #ith a resisti/e load ha/e confirmed satisfactory o0eration of the generator2 Its acti/e mass is smaller, it has a lo#er starting tor3ue, and it is more efficient, com0ared to the 0re/ious 0rototy0e2 B2J2 Chalmers, '2 'u, E2 S0ooner, G*n axial1flux 0ermanent1magnet generator for a gearless #ind energy system,G '((( trans. on (ner)y Con*ersion, <ol2 >, &o2 !, June ,,,, 00!:1!:)2 E2 ;ul=ada, C2P2 Butterfield, N2 'an, G*xial1flux modular 0ermanentmagnet generator #ith a toroidal #inding for #ind1tur+ine a00lications,G '((( Trans. on 'ndustry +pplications, <ol2 ":, &o2 >, JulyO*ugust ,,,, 00$"1$".2 <2 S2 Ramsden, G*00lication of rare1earth magnets in high10erformance electric machines,G 15Ih 'nternational ,orkshop on "are!(urrh -a)nets and Their +pplications, Hresden, "( *ugust1" Se0tem+er, ,,$, 00.!"1.>!2 J2N2 Chen, C2<2 &ayar, G* multi10ole 0ermanent magnet generator direct cou0led to #ind tur+ine,G 'nternational Conference on (lectrical -achines, Istan+ul, Tur7ey, .!/ Se0tem+er ,,$, 00))1)!!2 ;2 @yimesi, H2 Ostergaard, GInductance com0utation +y incremental finite element analysis,G '((( Tram -a)netics, <ol2 ":, &o2 ", ;ay ,,,, 00,1!!2 The authors #ould li7e to than7 Jianyi Chen at Curtin Uni/ersity of Technology, for 0ro/iding them 1it% test results of the 0re/ious