This version was downloaded from Northumbria Research Link:

Transcription

1 Citation: Dorrell, D. and Jovanovic, Milutin (8) On the oibilitie of uing a bruhle doubly-fed reluctance generator in a MW wind turbine. In: 8 IEEE Indutry Alication Society Annual Meeting. IEEE Xlore, Picataway, NJ. ISBN Publihed by: IEEE Xlore URL: Thi verion wa downloaded from Northumbria Reearch Link: htt://nrl.northumbria.ac.uk/73/ Northumbria Univerity ha develoed Northumbria Reearch Link (NRL) to enable uer to acce the Univerity reearch outut. Coyright and moral right for item on NRL are retained by the individual author() and/or other coyright owner. Single coie of full item can be reroduced, dilayed or erformed, and given to third artie in any format or medium for eronal reearch or tudy, educational, or not-for-rofit uroe without rior ermiion or charge, rovided the author, title and full bibliograhic detail are given, a well a a hyerlink and/or URL to the original metadata age. The content mut not be changed in any way. Full item mut not be old commercially in any format or medium without formal ermiion of the coyright holder. The full olicy i available online: htt://nrl.northumbria.ac.uk/olicie.html Thi document may differ from the final, ublihed verion of the reearch and ha been made available online in accordance with ubliher olicie. To read and/or cite from the ublihed verion of the reearch, leae viit the ubliher webite (a ubcrition may be required.)

2 On the Poibilitie of Uing a Bruhle Doubly-Fed Reluctance Generator in a MW Wind Turbine David G Dorrell Det of Electronic and Electrical Engineering Univerity of Glagow Glagow, G1 8LT, UK Milutin Jovanovic School of Comuting, Engineering & Information Science Northumbria Univerity Newcatle uon Tyne, NE1 8ST, UK Abtract The aer will ut a forward a decrition of a realitic deign for a bruhle doubly-fed machine that i of ractical value. It take the form a 4-ole / 8-ole machine aimed for ue in a MW wind turbine over a correct eed and torque range. Previou examle dicued laboratory machine and thi tudy take the outcome from thee tudie to formulate a rocedure for izing and deigning the machine. The latet deign and analyi technique are ued, with ractical rotor ducting conidered. Control i alo addreed in the aer, in articular the aement of enorle reactive ower control. Thi aer i aimed at taking the machine from a mall-cale laboratory examle to conideration a a large-ize indutrial generation and therefore it rereent a te-change in the literature on the machine. Keyword- Doubly-fed reluctance generator, wind turbine I. INTRODUCTION Wind turbine ue either a cage-rotor induction generator, a variable eed ynchronou generator or, more recently, a doubly-fed wound-field induction generator (with the field fed from a converter and main winding fed connected to the grid - DFIG). Thee are uually connected to the turbine via a gearbox although large-diameter direct-drive generator do exit. The wound-field induction generator i now very common but it ha reliability iue due to the li ring. Thi may be an obtacle for it wider ue, for examle, in off-hore wind turbine where oeration and maintenance cot can be ignificant [1]. Reearcher are now invetigating bruhle doubly-fed generator where there are two et of 3-hae winding thee will have different ole number (ay and 6 ole, a often ued in the literature, or a 4 and 8 ole combination a ued here) with one connected to the grid (ower winding) and one controlled via a converter (control winding). There are two alternative for thi machine: induction tye (with the rotor formed from bar connected in neted loo) and the bruhle doubly-fed reluctance machine tye (with a alient ole rotor imilar to witched reluctance machine or an axially-laminated rotor) the BDFRM. The BDFRM rotor doe not have a cage or winding. Thi cold rotor i more mechanically robut and allow imle modeling and control a well a higher efficiency [13]. Thi aer reort on a tudy of the electromagnetic analyi of the doubly-fed radially-laminated reluctance generator. A recent aer [1] aeed the iron loe generated in the axially-laminated machine and debated whether the correct otion hould be to ue a radially-laminated rotor to reduce the rotor loe. Thi wa exanded uon in []. The control of thee machine ha already been tudie extenively [3]-[7]. While work on the electromagnetic deign wa carried out in [8]. The correct deign of thi tye of machine i till under invetigation; [9] tudied the converion of an induction machine with a ecially contructed rotor. It wa found that it wa difficult to get the correct electrical loading becaue there wa inufficient lot ace. Another recent aer [1] develoed an algorithm for the correct deign of a machine uing a alient-ole reluctance machine. The deign focued on the radially-laminated nominally 7.5 kw machine. Initially the deign took the axially-laminated machine in [1] and redeigned the rotor with a radially-laminated rotor. The rocedure took the tator and redeigned the lamination to give the correct lot area. Thi wa done by eentially increaing the lot deth and outer diameter. Thi aer highlighted variou iue with thi deign on machine. Firtly that the /6 ole combination i quite a oor combination. Thi i becaue cro couling occur between the tator winding et not only through the rotor ermeance modulation, but alo through 3 rd harmonic aturation of -ole winding. Secondly, while the axiallylaminated rotor in [1] roduce eddy current lo it i very ole-ecific wherea the alient ole rotor i not a generalized and can roduce modulation with further ermeance harmonic. Fig. 1 how thee different rotor combination. The third major deign iue that wa raied wa the lot ize. Previou rototye have all been converted from induction machine deign with ecially contructed rotor. However, thi wa found to be inaroriate becaue the rotor electrical loading, a reent in a cage or DFIG wound rotor, i moved from the rotor to the tator. Therefore more tator lot area i needed eentially the lot deth and tator outide diameter have to be increaed to accommodate thi. Another alternative i to ue a ducted rotor a hown in Fig.. Thi i imilar to the axially-laminated rotor but there are much fewer duct (8 er ole here) ince radial lamination are ued. Thi i a 6 ole rotor rather than the 4 ole rotor illutrated in Fig. 1. Thi i becaue a /6 ole wa found to be oor o that a 4/8 ole combination wa choen here. It alo how increaed lot area. The rotor ha axial duct (not necearily air they may be ued for contruction) and i radially laminated. In thi aer deign for a MW bruhle reluctance generator will be develoed to invetigate it otential a a wind turbine generator. A full electromagnetic imulation will /8/$5. 8 IEEE 1 Authorized licened ue limited to: Univerity of Northumbria. Downloaded on Augut 6, 9 at 5:4 from IEEE Xlore. Retriction aly.

3 be conducted uing finite element analyi. In the econd art of the aer a control trategy i ut forward which i neceary to control the machine. The reult of the control imulation are decribed. Fig. 1. Axially-laminated rotor and alient-ole radially-laminated reluctance rotor. The control of the machine, in articular direct torque control (DTC) and enorle control, i addreed below. To the bet of the author knowledge, the only enorle eed control algorithm for the BDFRM ha been reorted in [14], but without uorting tet reult to demontrate it ractical realization. A DTC cheme wa rooed and imulated in [15], and exerimentally verified in [16]. It wa hown to overcome the uual deficiencie of the traditional DTC aroache and allowed table machine oeration down to zero alied frequency of the inverter-fed (econdary) winding (Fig. ). However, while enorle control of torque and flux wa achieved in [15][16], the eed feedback information required for eed control wa derived from rotor oition meaurement and not etimate. The control ection of thi aer i comlementary in nature to [15] and [16], and can be treated a a comrehenive extenion to thi reviou theoretical work on enorle DTC [17]. Unlike [17], where the maximum torque er inverter amere roerty wa conidered, the condition for maximum rimary ower factor control will be develoed in the aer and it ucceful ractical imlementation verified by exerimental reult. The latter will clearly how how a conventional load model baed oberver [18] can be effectively ued for the machine eed identification from the etimated rotor oition to achieve true encoderle eed control in realtime. ~ u & i calc. i u Primary Flux Etimator Torque Etimator Rectifier Inverter r * Switching Table Driver i calc. Signal i Rotor Poition Etimator Sector Secondary Flux T _ e T e Etimator _ * T + e + * PI Flux Reference Calculator + _ r T r e Angular Velocity Oberver Fig.. Cro ection of 4/8 ole machine with 6-ole ducted rotor and enorle eed and direct torque (ower) control of BDFRM. II. MACHINE DESIGN CHOICES A MW turbine will uually rotate between about 1 rm and rm with the eak ower being reached at about 15 rm. We need to et the eed range of the turbine. Taking the 8-ole winding to be the ower winding, there i a ynchronizing requirement for the non-grid connected control winding: ω = Pω ± ω (1) c r where the ω i a rotational velocity (rad/ec) and c, r and rereent the control frequency, rotor velocity and grid frequency; P i equal to 6 for the 4/8 machine. At low eed we want the control winding to be at Hz u to about 5 Hz. Thi give a eed range from 5 to 1 rm (Fig. 3 olid line) which i a imilar eed range to the equivalent DFIG machine (although lightly lower). At 75 rm eak ower i reached and the machine will then go into effectively a field weakening range, thi correond to 5 Hz control winding frequency. Machine eed [rm] control winding frequency [Hz] Fig. 3. Seed range for 4/8 machine (4-ole winding i control winding). A. Baic izing If the target deign i taken a 75 rm for MW then the torque i 6 knm. From [11] the torque er rotor volume for an integral hore-ower indutrial motor could be u to 3 knm/m 3. Thi target eed i at maximum ower but at thi tage we are imly carrying out baic izing, alo ince thi i a larger machine then the torque er rotor volume hould be high. Therefore the rotor will have volume of 6/3 =.86 m 3. Let u alo aume a diameter to axial length ratio of a half then we can define the rotor diameter a.84 m and the axial length a 1.68 m. Sizing the tator i omewhat more comlex, however, a an initial etimate the outer diameter i taken a double the rotor diameter at 1.68 m. B. Winding arrangement For uch a large machine the number of lot would be high, oibly 7 lot or above. However, to maintain imlicity 48 lot are choen here ince it i the lowet oible number of lot for an 8 ole ymmetrical winding. Thi arrangement i hown in Fig.. In Fig. 4 one hae of the 4 ole and one hae of the 8 ole winding are illutrated. The 8 ole winding ha 3 turn-er-coil and the 4 ole winding ha 3 turn-er-coil. If double layer winding were ued then further refinement i oible with hort itching, articularly in the 4 ole winding. Authorized licened ue limited to: Univerity of Northumbria. Downloaded on Augut 6, 9 at 5:4 from IEEE Xlore. Retriction aly.

4 (a) Fig. 4. (a) 8-ole and (b) 4-ole hae winding. (b) With the geometry ued in imulation and the winding decribed above the hae voltage when the 8-ole winding i excited are hown in Fig. 6 and for 4-ole excitation in Fig. 7. The current in the 4-ole winding in Fig. 7 i high and thi roduce ome lotting effect a can be een in the voltage waveform. However, thi i a baic izing exercie uing current-fed tatic finite element olution more refinement of the rotor flux barrier to lot number ratio would hel eliminate thi. If the flux wave i invetigated then the rile i not a noticeable (a hown later under loaded condition); the voltage i the differential of the flux linkage - an alternative would be to brake down the flux wave into it Fourier comonent and um the differential term. The higher rile can be ignored. C. Oen circuit imulation If the 8-ole ower winding i excited with a hae current of 9.1 A rm (the current at maximum ower at 1 rm) then a flux lot i hown in Fig. 5(a). In addition, a flux lot with the 4-ole winding excited only i alo hown in Fig. 5 (b). Thi i carried out at 41.5 A to maintain the EMF in the ower winding. While it i oible to oberve flux attern that may include 4 and 8 ole together, it i more correct to how the EMF induced into the winding. It can be een that voltage are uitable for u to 5 kv although the turn would be choen to match the ytem (5kW i obviouly high voltage requiring ecialit winding or reduced hae turn can be ued to reduce the voltage then te u via a tranformer). Bear in mind that the 8-ole winding i nominally the ower winding, which i fixed in frequency and voltage wherea the 4-ole winding i the control winding with variable voltage and frequency. Thee waveform aear reaonable for a firt-a deign. In thee imulation the rotor i notionally rotating at 1 rm o that the frequencie in both of the winding are 5 Hz. Thee reult how that the FEA model of the machine roduce reaonable reult. The torque i aeed below uing current-flux denity loo a commonly ued for torque rediction in reluctance and ermanent magnet machine. 1 Voltage [V] (a) FEA voltage from differential of flux linkage Fundamental comoment Current hae [elec deg] (a) ower winding with current FEA voltage from differential of flux linkage Fundamental comoment Voltage [V] (b) Fig. 5. Flux lot with (a) 8-ole winding exited only (9.5 A) and (b) 4-ole winding excited only (41.5 A). -3 Current hae [elec deg] (b) control winding oen circuit Fig ole excitation voltage (a) 8-ole winding with 9.5 A current 11.5 kv rm, and (b) 4-ole oen circuit voltage 1.9 kv rm; 1 degree mechanical rotation at 1 rm. 3 Authorized licened ue limited to: Univerity of Northumbria. Downloaded on Augut 6, 9 at 5:4 from IEEE Xlore. Retriction aly.

5 Voltage [V] Voltage [V] FEA voltage from differential of flux linkage Fundamental comoment Current hae [elec deg] (a) control winding with current FEA voltage from differential of flux linkage Fundamental comoment Current hae [elec deg] (b) ower winding oen circuit Fig ole excitation voltage (a) 4-ole winding with 41.5 A current 4.7 kv rm, and (b) 8-ole oen circuit voltage 1.3 kv rm; 1 degree mechanical rotation at 1 rm. III. SIMULATIONS In thi ection we will examine the oeration of the outlined machine toology uing finite element analyi. Firtly we will addre the baic equivalent circuit and obtain equation for the ower delivered via each winding. Thi will illutrate why the 4 ole winding i ued a the control winding and why the frequency of the control winding hould be le that the ower winding. The haor diagram under load i obtained and then the torque and ower are obtained from the finite element analyi uing current flux linkage loo (I-Pi loo). A. Equivalent circuit and haor diagram The baic teady-tate equivalent i given in Fig. 8. Thi can be ued to redict the oeration in the machine. From thi circuit (note the ue of generating convention for the current) the ower generated i * * Ptotal = PS + P = Re{ ES IS } + Re{ EP IP } () δ * δ * = Re{ SMIe P IS} + Re{ PMIe S IP} If the current i maintained on the reective notional q-axe o that the current i in hae with the winding back-emf E and E then when the ower and control winding have the ame frequencie (i.e., at 1 rm) then the maximum ower in each winding et i the ame. However, a the eed reduce, ay to 75 rm, then the control winding frequency reduce to 5 Hz and now the control winding can only contribute half of the maximum ower of the ower winding becaue of the frequency term in (). The haor diagram are given in Fig. 9 for oen-circuit oeration (i.e., the ower winding i not connected to the bubar) and full load oeration at 1 rm. It can be een that the control winding require much more current on oen-circuit becaue it ha to induce the bubar voltage into the ower winding. The haor diagram under load are at the oint when the machine i under-excited and reactive ower i being drawn from the ulie. To imrove oeration in term of ower factor, o that reactive ower i generated rather than aborbed, then careful control need to be imlemented. Control winding V S I S R S L S SM IPe δ Variable frequency and terminal voltage Fig. 8. Baic teady-tate equivelant circuit. On O/C no EMF induced by ower winding o that: = VS + SLS IS + I R S h V S Bubar voltagev P j je δ = I S 1 S L I Control winding - oen-circuit on ower winding Power winding - oen circuit S S S Power winding L P R P PMIe δ S P I P V P Fixed frequency and terminal voltage (for both O/C and Loaded cae) SM ISe δ I S I P V S Control winding - Loaded V P Power winding - loaded PM IPe δ L I SM ISe δ S S S L I P P P (a) (b) Fig. 9. Oen-circuit on ower winding and loaded haor diagram increaed control winding current required on oen-circuit to maintain bubar voltage on ower winding. B. Load imulation and I-Pi loo at 1 rm. The machine wa imulated at 1 rm with 59 A in the 4 ole control winding and 9.5 A in the ower winding. The frequency in both winding i 5 Hz at thi eed and equation () ugget that for maximum ower at thee current then each winding contribute about the ame ower. In Fig. 1 the current flux denity (I-Pi) loo are hown for all three hae of each winding. The area encloed in each loo rereent the work done. The 4-ole winding contribute. 98 MW of electrical ower and the 8 ole ower winding contribute.97 MW. Thi give a total of 1.95 MW which i cloe to the required (although thi excluded coer loe, iron loe and friction and windage). There i ome rile in the flux linkage and when thi i ued to obtain the hae voltage that the characteritic a hown in Fig. 1 (ower winding) and Fig. 13 (control winding) are obtained. A reviouly mentioned thi arrangement i aimed at teting the oibility of realizing thi machine at thi ize o that a more detailed deign would ue a different rotor/tator arrangement to reduce the voltage rile. 4 Authorized licened ue limited to: Univerity of Northumbria. Downloaded on Augut 6, 9 at 5:4 from IEEE Xlore. Retriction aly.

6 Flux linkage [mv] Flux linkage [mv] Phae current [A] (a) 8-ole ower winding Phae current [A] (a) 4-ole control winding Fig. 11. I-Pi loo for machine with 9.5 A rm in control winding and 59 A rm in ower winding at 1 rm. Power winding voltage [V] relatively narrow with a voltage i 8. kv rm and the hae angle i 79 elec deg. Thi give a ower factor of.. However, thi oint i when the current are almot in hae with the induced back-emf o that the machine i oerating very under-excited a illutrated in the haor diagram in Fig. 9. If the current were made equal then the ower factor can be imroved. If we look at the oen-circuit tet in Fig. 6 and 7 then we can calculate the X P a 39 Ω an M S-P to be 369 Ω from the voltage and excitation current (9.5 A rm) in ower winding. However when the control winding i excited with 41.5 A then X S i 595 Ω and M P-S i 49 Ω. In theory, the cro couling hould mean that M P-S = M S-P but thi i not the cae. Inection of the flux lot in Fig. 5 how that when the control winding i excited with a higher current then there i coniderably more flux in the machine and the rotor teel aear to be cloe to aturation. If the current i reduced to 9.5 A in the control winding when the ower winding i oencircuit then X S increae to 78 Ω and M P-S to 38 Ω. Hence thi illutrate the oint that thi machine i oerating cloe to it limit (a it hould be due to the deign aroach). Flux linkage [mv]. Two loo on to of each other Phae current [A] (a) 8-ole ower winding (Phae 1 only) Current cycle [elec deg] Fig. 1. Power winding hae voltage under load with 59 A rm in control winding and 9.5 A rm in ower winding at 1 rm. 5 Flux linkage [mv] Control winding voltage [V] Current cycle [elec deg] Fig. 13. Control winding hae voltage under load with 59 A in control winding and 9.5 A in ower winding at 1 rm. The ower winding loo i an oval and the hae difference between the voltage and current i 3 deg. Thi give a ower factor of.87. However the loo of the control winding i Phae current [A] (a) 4-ole control winding (Phae 1 only) Fig. 14. I-Pi loo for machine with 41.8 A rm in control winding and 41.8 A rm in ower winding at 75 rm. C. Oeration at 75 rm The eed in the imulation wa reduced thi wa done be carrying out a imulation over 18 mechanical degree of movement. During thi eriod the ower winding will cycle through two comlete current cycle and the control winding (now at 5 Hz) with cycle through one comlete cycle. If the current are et to 41.8 A in both winding and the eed reduced to 75 rm the ower from the 4 ole control winding wa calculated to be.66 MW while the 8 ole ower from the 5 Authorized licened ue limited to: Univerity of Northumbria. Downloaded on Augut 6, 9 at 5:4 from IEEE Xlore. Retriction aly.

7 ower winding wa 1.14 MW. Thi give a total outut ower of 1.8 MW which i cloe the target MW. The I-Pi diagram are hown in Fig. 14. If can be een that the area of the loo are now larger and the voltage of the 8 ole winding i 15.6 V rm and the 4 ole winding i 16.1 V rm. The 4 ole ower factor i now.58 and the 8 ole ower factor i.33. Thee are till relatively low however the machine i till oerating in under-excited mode. D. Dicuion Thi ection ha given a baic arrangement for a oible MW bruhle doubly-fed reluctance machine. Thi i an exercie in caling and illutrate the oeration of the radiallylaminated ducted rotor. Thi eem to give good reult uing a crude firt-te deign when imulated uing D finite element analyi. There eem to be reaonable modulation by the rotor on the MMF to get the required cro-couling of the different ole-number winding. The machine wa heavily loaded to check to ee if there wa any additional harmonic cro couling. Thi wa found to be a roblem in the /6 machine in [1] and []. The cro-couling i very imortant for correct oeration. The voltage rile can be traced to either numerical error or lotting effect and deign and analyi refinement will alleviate thi. The imulation ued the I-Pi loo to obtain the torque and the flux linkage differential i ued to obtain the voltage. The ower factor aear to be low however the imulation ket the current cloe to the hae of E P and E C which i effectively under-excited oeration. IV. CONTROL The model ued in the finite element analyi rereent a imle teady-tate circuit for the machine to enable the baic izing and teting of a 4 ole control winding and 8 ole ower winding machine and thi gave romiing reult for oible ue in a MW wind turbine generator. Here we will look at a detailed review of the control of thi tye of machine. It i teted on a /6 lab-baed machine to validate the algorithm. A. Dynamic Model The ace-vector equation in a tationary reference frame and the fundamental angular velocity relationhi for the BDFRM torque roduction are [19] [1]: dλ dλ u = Ri + = Ri + θ cont = + λ dt dt (3) dλ dλ u = Ri + = Ri + θ cont j = + ( ωr ω ) λ dt dt (4) λ = L i + L * jθ r i e jθ = λ e (5) * jθ jθ r λ = L i + L i e = λ e (6) ω r = dθr / dt = rωrm = ω + ω (7) where L,, rereent the 3-hae inductance of the gridconnected (rimary or ower) and inverter-fed (econdary or control) winding [1][], ω rm i the rotor angular velocity (rad/) at which the machine develo ueful torque, r i the number of rotor ole, ω, are the alied frequencie to the winding. Note that ω > for uer-ynchronou oeration and ω < if the machine i oerated below the ynchronou eed. At ynchronou eed ω =, i.e., the econdary ide i DC ulied a with a claical r -ole ynchronou machine. The negative econdary frequency in the ub-ynchronou mode imly mean the ooite hae equence of the econdary to the rimary winding. The angular oition of variou haor in (3) to (6) are defined in Fig. 7. i i A - axi (tator) L L Fig. 15. Characteritic haor in a tationary reference frame. B. Senorle reactive ower control A detailed decrition and erformance evaluation of the DTC cheme for the BDFRM develoed wa ut forward in [15], [16] and [17]. One of the BDFRM main attribute i it ower factor control caability [4]. The ower factor in the econdary winding i directly related to the inverter ize, but i irrelevant to the outide utility network (ince the inverter effectively iolate the econdary from the main uly). However the ower factor of the rimary winding i of great imortance to the utility grid (eecially in weak network) in the light of reactive ower requirement. To minimize the total current loading (and thu loe) for a given real ower demand, it i therefore deirable to kee the rimary ower factor at or, a cloe a oible to unity. Uing the rimary flux oriented form of (3)-(6) the econdary flux exreion for otimization of the rimary ower factor or any other erformance indicator of the machine can be derived. It wa hown in [3] and [4] that the maximum rimary ower factor (MPPF), i.e., no reactive ower flow through the rimary winding: 3 ω λ Q = ( λ L id ) = L i achieved if i d = λ / L. Under thi condition, the MPPF econdary flux reference for a deired torque (Fig. 15) and thi can be exreed a: * * L σl T λ e = λ + (8) L 1 σ 3rλ λ λ d where σ = 1 L /( LL ) i the leakage factor and i d i the d-axi econdary current aligned with the mutual flux vector, i.e., λ in Fig. 15. It can be een from (8) that λ d cont irreective of the machine loading due to the rimary winding q 6 Authorized licened ue limited to: Univerity of Northumbria. Downloaded on Augut 6, 9 at 5:4 from IEEE Xlore. Retriction aly.

8 grid connection, i.e., λ cont. Thi fact i imortant a it mean that the torque roducing λ q comonent can be controlled indirectly via λ but in a tationary (and not rotating) frame. In other word, the DTC can be otimized ince in thi cae vector control roblem i reduced to a ingle variable effectively becoming calar in nature. For the coe of thi aer, a ignificant benefit of greater control freedom, afforded by the acceibility of both BDFRM winding, i the oibility of enorle eed control [17]. The rotor angle can be retrieved from (5): Im[ ( λ L i ) i ] θ r = tan 1 (9) Re[ ( λ L i ) i ] The raw oition etimate are then inut into a Luenberger tye PI oberver [18] to redict the rotor angular velocity ω r = dθ/dt ued for the eed control a hown in Fig. 1. Excellent low a filtering abilitie of thi oberver, anticiated by imulation in [18], have been exerimentally verified by reult which are reented below. It hould be mentioned here that the rotor oition information i only required for eed etimation and not for torque control, ince the method i tator-frame-baed, a i uual for DTC. C. Exerimental Reult The enorle algorithm in Fig. ha been imlemented and executed in dspace for the MPPF control trategy on a caled-down 6/-ole BDFRM rototye. Detail of the tet ytem and the relevant machine arameter can be found in [15] and [16]. The to lot in Fig. 16 rereent the rotor angle obtained from (9), and their abolute variation from encoder meaurement. The raw etimate are noiy, but deite the error ike, which have been found to be largely due to the ractical effect uch a meaurement noie and enitivity to arameter knowledge inaccuracie, the average etimation error i till reaonably low ( 7º). The effectivene of the oberver a a low-a filter i evident from the ame figure (bottom lot), and a ignificant imrovement in accuracy i achieved by roceing θ r through it. The average error i reduced to aroximately 1.5º with the maximum value being about 3.4º or le. The main reaon for uch a high accuracy i the quality etimate being fed into the oberver by the oition etimator which, imilarly to the latter, work in a cloed-loo fahion a illutrated in Fig.. Fig. 17 how the eed waveform for change in deired eed value between 95 rm, 75 rm and 55 rm. In thi cae, the eed limit correond to f = 13.3 Hz in either mode. It can be een that the machine may be effectively controlled over the conidered eed range, including ynchronou eed (75 rm) when f =. Reliable low-frequency oeration of the BDFRM i an imortant oint of the rooed enorle cheme, and certainly rereent a ignificant advantage over traditional DTC and many other back-emf baed control method. Thee have difficultie (or imly do not work) in thi frequency region even in enor eed mode. It hould be emhaized that the gain of both the eed PI regulator and the oberver mut be lowered and aroriately tuned a intability and divergence of the control algorithm may otherwie occur due to the noiy inut etimate. Thi trade-off reult in low bandwidth control and relatively modet dynamic reone of the machine which i, fortunately, quite accetable for the target alication where teady-tate erformance i of more interet. Poition [degree] Error [degree] Poition [degree] Error [degree] Time [m] Time [m] Time [m] Time [m] Fig. 16. Etimated (to) and oberved (bottom) rotor oition at 85 rm (f = 6.7 Hz). Seed [rm] Time [] Fig. 17. Senorle control erformance down to ynchronou eed Authorized licened ue limited to: Univerity of Northumbria. Downloaded on Augut 6, 9 at 5:4 from IEEE Xlore. Retriction aly.

9 Voltage Current Time [m] Fig. 18. Primary voltage and current waveform for unity ower factor. The ocillocoe trace for rimary winding voltage and current in Fig. 18 clearly demontrate that the intended maximum rimary factor oeration ha been uccefully achieved. Note that the reective waveform are mooth, virtually witching rile-free (due to the relatively weak magnetic couling between the winding being inherent with thi articular machine) and at line frequency (5 Hz). Similar reult could be obtained for unity (or even leading) line ower factor control in which cae the econdary ide would be entirely reonible for the machine magnetization by roviding the neceary reactive ower to the rimary (or to the grid) at the exene of increaed inverter loading. CONCLUSIONS Thi aer ut forward a decrition of a realitic deign for a bruhle doubly-fed machine that i of ractical value. It take the form a 4-ole - 8-ole machine aimed for ue in a MW wind turbine over a correct eed and torque range. Previou examle dicued laboratory machine and thi tudy take the outcome from thee tudie to formulate a rocedure for izing and deigning the machine. A radiallylaminated ducted rotor deign i tudied and thi roduce good modulation and cro-couling between the two different olenumber winding. Finite element analyi i ued in conjunction with I-Pi loo to validate the baic deign. Control i alo addreed in the econd art of the aer. The work illutrate that ower factor control i neceary and thi i a major iue with thi machine. Senorle control i aeed a a oible control trategy. Thi aer i aimed at taking the machine from a mall-cale laboratory examle to conideration a a large-ize indutrial generation. REFERENCES [1] I. Scian, D. G. Dorrell, and P. Holik, Aement of Loe in a Bruhle Doubly-Fed Reluctance Machine, IEEE Tranaction on Magnetic, INTERMAG ecial edition, October 6. [] D. G. Dorrell, I. Scian, E. M. Schulz, R. B. Betz and M. Jovanovic, Electromagnetic Conideration in the Deign of Doubly-Fed Reluctance Generator for ue in Wind Turbine IEEE Indutrial Electronic Conference, IECON, Pari 7-1 Nov. 6. [3] R. E. Betz and M. G. Jovanovic, The Bruhle Doubly Fed Reluctance Machine and the Synchronou Reluctance Machine - A Comarion, IEEE Tran on Indutry Alication, Vol. 36, No. 4, July/Augut. [4] R. B. Betz and M. G. Jovanovic, Theoretical Analyi of Control Proertie for the Bruhle Doubly Fed Reluctance Machine, IEEE Tran on Energy Converion, Vol 17, No 3, Set, [5] M. G. Jovanovic and R. E. Betz, The Ue of Doubly Fed Reluctance Machine for Large Pum and Wind Turbine, IEEE Tran on Indutry. Alication, Vol 38 No 6, Nov, [6] M. G. Jovanovic and J. Yu, An Otimal Direct Torque Control Strategy for Bruhle Doubly-Fed Reluctance Motor, IEEE Power Electronic and Drive Sytem Conference, 17- Nov 3, Vol, [7] M. G. Jovanovic, Control of Bruhle Doubly-Fed Reluctance Motor, IEEE International. Symoium. on Indutrial Electronic, 5, Dubrovnik, Croatia, [8] E. M. Shulz, and R. E. Betz, Otimal Torque er Am for Bruhle Doubly Fed Reluctance Machine, 4th IEEE IAS Meeting, -6 Oct, 5, Hong Kong, [9] D. G. Dorrell, I. Scian, and P. J. Holik, Aement of the Electromagnetic Performance of Doubly-Fed Reluctance Machine for Ue in Variable Seed Generator, ICEMS, Nagaaki, Jaan, -3 Nov 6. [1] D. G. Dorrell, Deign Requirement for Doubly-Fed Reluctance Generator, IEEE PEDS Conference, Bangkok, Thailand, Dec 7. [11] J. R. Henderhot and TJE Miller, Deign of Bruhle Permanent Magnet Motor, Oxford Science Publication, [1] P.Bauer, S. de Haan, C.R.Meyl, and J.T.G.Pierik, Evaluation of electrical ytem for off-hore windfarm, IAS Annual Meeting, Rome, Italy,. [13] F.Wang, F.Zhang, and L.Xu, Parameter and erformance comarion of doubly-fed bruhle machine with cage and reluctance rotor, IEEE Tranaction on Indutry Alication, vol. 38, , Set/Oct. [14] Y. Liao and C. Sun, A novel oition enorle control cheme for doubly fed reluctance motor drive, IEEE Tranaction on Indutry Alication, vol. 3, , Set/Oct [15] M.G.Jovanovic, J.Yu, and E.Levi, Direct torque control of bruhle doubly fed reluctance machine, Electric Power Comonent and Sytem, vol. 3, , October 4. [16] M.G.Jovanovic, J.Yu, and E.Levi, Encoderle direct torque controller for limited eed range alication of bruhle doubly fed reluctance motor, IEEE Tranaction on Indutry Alication, vol. 4,. 71 7, May/June 6. [17] M.G.Jovanovic, J.Yu, and E.Levi, A doubly-fed reluctance motor drive with enorle direct torque control, IEEE International Electric Machine and Drive Conference (IEMDC), Madion, Wiconin, June 3. [18] R. Lorenz and K. Patten, High-reolution velocity etimation for alldigital, ac ervo drive, IEEE Tran. on Indutry Alication, vol. IA- 7, , July/Augut [19] F. Liang, L. Xu, and T. Lio, D-q analyi of a variable eed doubly AC excited reluctance motor, Electric Machine and Power Sytem, vol. 19, , March [] Y. Liao, L. Xu, and L. Zhen, Deign of a doubly-fed reluctance motor for adjutable eed drive, IEEE Tranaction on Indutry Alication, vol. 3, , Set/Oct [1] R.E.Betz and M.G.Jovanovic, Introduction to the ace vector modelling of the bruhle doubly-fed reluctance machine, Electric Power Comonent and Sytem, vol. 31, , Augut 3. [] R.E.Betz and M.G.Jovanovic, The bruhle doubly fed reluctance machine and the ynchronou reluctance machine - a comarion, IEEE Tranaction on Indutry Alication, vol. 36, , July/Augut. [3] L. Xu, L. Zhen, and E. Kim, Field-orientation control of a doubly excited bruhle reluctance machine, IEEE Tranaction on Indutry Alication, vol. 34, , Jan/Feb [4] M. G. Jovanovic and R. E. Betz, Power factor control uing bruhle doubly fed reluctance machine, Proc. of the IEEE-IAS Annual Meeting, Rome, Italy, October. 8 Authorized licened ue limited to: Univerity of Northumbria. Downloaded on Augut 6, 9 at 5:4 from IEEE Xlore. Retriction aly.