Improved sensorless control of a permanent magnet machine using fundamental pulse width modulation excitation

Transcription

1 Pulished in IET Eletri Power Applitions Reeived on 19th April 2010 Revised on 27th July 2010 doi: /iet-ep Improved sensorless ontrol of permnent mgnet mhine using fundmentl pulse wih modultion exittion Y. Hu M. Sumner G. Asher Q. Go K. Sleh Deprtment of Eletril nd Eletroni Engineering, University of Nottinghm, Nottinghm NG7 2RD, UK E-mil: ISSN Astrt: A new method to estimte the rotor position of permnent mgnet synhronous motor (PMSM) drives over wide speed rnge, inluding zero frequeny is introdued. The method mesures the motor urrent derivtive in response to the stndrd pulse wih modultion (PWM) sequene to estimte rotor position. No dditionl signl injetion or seprte test vetors re required lthough modifition is needed when nrrow PWM voltge vetors our. Three ompenstion tehniques re proposed to improve the qulity of the position estimtion nd redue urrent distortion when nrrow vetors re used for position estimtion. Experimentl results for sensorless speed nd position ontrol of 4 kw PMSM verify the effetiveness of the proposed tehniques. 1 Introdution Sensorless ontrol of AC motor drives hs een developed over reent yers to improve reliility nd redue ost of industril vrile speed drives. Shft mounted position or speed sensors dd to system ost s they require oupling to the motor nd lignment, nd this uses high proportion of uild nd drive ommissioning filures. Enoder signl wiring lso redues the reliility of ommissioned drive systems. Sensorless ontrol sed on the fundmentl eletromehnil model performs dly in the low nd zero speed rnge [1, 2] s the mhine s k-emf tends to zero s the rotor speed redues nd informtion is lost in noise nd low mesurement resolution. Therefore sensorless methods whih trk the inherent slienies of motor [3 9] hve een developed. These slienies n e used y mgneti sturtion or y the physil onstrution of the rotor. Most of these slieny trking tehniques fll into two tegories. The first tegory injets ontinuous high frequeny voltge or urrent into the mhine [4 7]. The position informtion resides in the indued high frequeny urrent or voltge signls. However, the position signl n e diffiult to extrt s it is orrupted y hrmonis resulting from other seondry mhine slienies [5] nd the non-linerity of the power onverter [10, 11]. To inrese the signl-to-noise rtio of the estimted position mesurements, the rrier mplitude must e inresed nd this leds to signifint torque ripple nd ousti noise within the drive system. Signl injetion tehniques n only e implemented t very low speeds, unless omined with model-sed pproh [12, 13]. The seond tegory exploits the trnsient urrent responses to pulse wih modultion (PWM) swithing sttes. The INFORM method [4] uses trnsient exittion to quire the rotor position, y using test vetors pplied during the null vetor of PWM period. The use of seprte test vetors results in extr losses, inresed ommon mode urrent nd voltge stress in the motor, nd therefore the integrtion of test vetors into the stndrd PWM wveform is preferred, s proposed in [8, 9], however, oth shemes still require speil test vetors t very low operting speeds. A more ttrtive pproh is to use the norml fundmentl PWM sequene s the test vetor [14, 15] termed here the fundmentl PWM exittion (FPE) method. The urrent derivtives for two tive vetors nd the null vetors re mesured over suessive norml PWM swithing yles. No speil test vetors re required nd dditionlly FPE n e employed over wide speed rnge. However, the min diffiulty of integrting the position estimtion sheme within the fundmentl PWM sequene rises from the prsiti effets within the drive system. Prsiti effets suh s inter-turn pitne in the motor, pitne of the insultion etween the insulted gte ipolr trnsistor (IGBT) nd the hetsink, nd so on result in high frequeny urrent osilltions when the IGBTs swith sttes, nd these tke few miroseonds to die down. The urrent derivtives required for the position estimtion nnot e mesured until these osilltions hve disppered. When the urrent derivtive response to prtiulr swithing vetor must e mesured, tht vetor must hve minimum durtion t min. Vetors shorter thn t min must e extended if position estimtion is required during tht swithing period, nd this results in signifint distortion of the motor urrent [14, 15]. This pper introdues three tehniques to improve the sensorless drive opertion when using FPE. A pulse extension/ompenstion sheme is proposed whih redues IET Eletr. Power Appl., 2011, Vol. 5, Iss. 4, pp doi: /iet-ep & The Institution of Engineering nd Tehnology 2011

2 the distortion introdued into the motor line urrent when nrrow vetor hs to e extended. A simple ompenstion lgorithm is proposed to eliminte ommon mode high frequeny osilltions from the urrent derivtive mesurements. Finlly, differentil mode high frequeny osilltions re redued sustntilly y modifying the IGBT drive iruit, to slow down the IGBT swithing only when nrrow vetor is deteted. Experimentl results re presented to demonstrte the effetive sensorless ontrol of surfe mounted permnent mgnet synhronous mhine (PMSM) using the proposed tehniques. 2 Position estimtion using fundmentl PWM sequene The sttor indutnes of three-phse, str-onneted surfe mounted PMSM re modulted y min flux sturtion nd therefore n e modelled y (1) [3] l s = l 0 + Dl os(2u e ) l s = l 0 + Dl os(2(u e 2p/3)) l s = l 0 + Dl os(2(u e 4p/3)) where l 0 is the verge sttor indutne nd Dl is the indutne vrition due to modultion y the min flux. Thus, the min flux position u e n e trked y monitoring the indutne vrition. This proess is now desried. When PWM wveform is pplied to PM mhine, the sttor phse urrent response is desried y (2) (1) u s = r s i s + l s di s + e (2) where u s nd i s re the sttor voltge nd urrent vetors, r s nd l s re the sttor resistne nd indutne, nd e is the k-emf vetor. The PWM sheme uses the six tive nd two null swithing sttes of three-phse inverter, defined s V 1 (1, 0, 0), V 2 (1, 1, 0), V 3 (0, 1, 0), V 4 (0, 1, 1), V 5 (0, 0, 1), V 6 (1, 0, 1) nd V 0 (0, 0, 0), V 7 (1, 1, 1) in the formt V x (,, ), where V x is the vetor nme nd (,, ) gives the swith stte of eh inverter leg. The phse urrent trnsient responses to the exittion of these voltge vetors n e derived from (2). In str-onneted PMSM, tking V 1 (1,0,0) s the pplied voltge vetor, the three sttor voltges re U DC = i (1) r s + l s 0 = i (1) r s + l s + e (1) i (1) + e (1) i (1) U DC = i (1) r s + l s r s l s r s l s + e (1) i (1) e (1) e (1) r s l s e (1) where U DC is the DC-link voltge, nd the supersript (1) denotes quntities mesured in response to voltge vetor V 1. A similr set of equtions derived from pplying the (3) losest null vetor in the PWM sequene, V 0 is given in (4). 0 = i (0) r s + l s 0 = i (0) r s + l s 0 = i (0) r s + l s + e (0) i (0) r s l s + e (0) i (0) r s l s + e (0) i (0) r s l s e (0) e (0) e (0) As the smpling instnts for V 1 nd V 0 re very lose to eh other (less thn one PWM period), the k-emf nd sttor resistne voltge drop n e ssumed to e pproximtely the sme in (3) nd (4) nd n e eliminted. Comining (3) nd (4) yields l s U DC = l s 0 = l s l s l s U DC = l s l s l s l s l s l s l s Inorporting (1) into (5) nd rerrnging yields di(0) di(0) di(0) = di(10) = di(10) = di(10) (4) (5) = 1 ( g 2 Dl ) os(2u l e ) 0 = 1 ( g 1 + Dl ( ( os 2 u l e 4p ))) 0 3 = 1 ( g 1 + Dl ( ( os 2 u l e 2p ))) 0 3 (6) where g ¼ 3l 0 (1 2 (Dl/2l 0 ) 2 )/U DC. The supersript (10) denotes the omintion of quntities mesured during V 1 nd V 0. The flux position vetor p s n e defined s [3] p s = Dl (p l + p + 2 p ) = Dl ( os(2u 0 l e ) ( ( 0 + os 2 u e 2p )) ( ( + 2 os 2 u 4p ))) 3 3 where ¼ e j(2p/3). The position vetor defined in (7) n now e expressed in terms of the mesured urrent trnsients of (6) s p = 2 g di(10) p = 1 g di(10), p = 1 g di(10), The position ngle nnot e derived from (8) lone. Exittion from nother tive voltge vetor is required. In setor I, the other tive vetor is V 2. Similr position (7) (8) 360 IET Eletr. Power Appl., 2011, Vol. 5, Iss. 4, pp & The Institution of Engineering nd Tehnology 2011 doi: /iet-ep

3 signls n e lulted during V 2 nd V 7 s given in (9). p = 1 + g di(27), p = 1 + g di(27), p = 2 + g di(27) (9) where the supersript (27) denotes the omintion of quntities mesured during V 2 nd V 7. The position vetor n e onstruted y seleting the position slrs with the nellle offsets from (8) nd (9). Suh s p = p + p + 2 p = + 1 g d(10) 1 + g di(27) g d(10) (10) where the offset 21 in the position slr equtions n then e eliminted. A unique estimte of the rotor position n e hieved using only mesurement of urrent derivtives in response to speifi pplied voltge vetors. Other omintions of mesured urrent derivtive in response to speifi voltge vetors n e derived for the other PWM setors, s desried in [14, 15]. 3 Current swithing osilltions The min diffiulty when pplying urrent derivtive-sed position estimtion shemes to rel systems rises from the prsiti effets within the motor, the inverter nd the ling. The high dv/ used y power devie swithing exites prsiti pitnes distriuted in the system, nd uses high frequeny urrent osilltions. Bsed on high frequeny system modelling nd the nlysis of the eletromgneti interferene effets, the swithing osilltions n e tegorised s ommon mode nd differentil mode omponents [16]. The ommon mode osilltion is used y the urrents flowing etween the inverter phse iruits nd the system ground, for exmple, through pitne etween the IGBT nd the grounded hetsink, or pitne etween motor winding nd the sttor iron itself due to the winding insultion. When one phse is swithed ON/OFF, the differentil mode osilltion on the inverter output results from similr prsiti pths (i.e. stry pitne) etween the swithed phse nd the other two phses. To illustrte the ehviour of the experimentl system used in this projet, the output urrent of one inverter phse together with the IGBT hetsink erth urrent nd the AC power supply erth urrent re shown in Fig. 1 for one swithing instnt. Two frequeny rnges re found in the ommon mode osilltion. The lower rnge ( khz) is the urrent whih penetrtes into the sttor windings nd flows through the supply indutors. This omponent exists in the erth ondutor of oth the supply le nd the mhine le. The other omponent is the smll high frequeny osilltion (ove 2 MHz) whih ppers only t the grounding point on the inverter s output side. The ommon mode urrent t this frequeny is loked y the impedne of the supply indutne nd the mhine s sttor windings. The differentil mode urrent flows etween three phses nd so only ppers in the output urrent. It hs high mplitude nd frequeny round 2 3 MHz. If the motor urrent, or urrent derivtive re to e mesured for position estimtion, it n e seen from Fig. 1 tht the mesurement must e delyed y 8 ms (t min ) from the swithing edge, so tht these osilltions do not orrupt the mesurement. However, t min should e kept s smll s possile in order to minimise the inurred distortion in the motor urrent. 4 Swithing vetor extension nd ompenstion Nrrow voltge vetors must e extended to minimum vetor durtion t min, whih is set ording to the high frequeny osilltion dey time. Beuse the voltge time re in eh PWM period should not e hnged, ny vetor extension must e ompensted. Fig. 2 shows simple extension nd ompenstion sheme pplied when the sttor voltge vetor is in setor I [14, 15]. Both swithing edges of phse re shifted to mke V 1 wide enough during the first hlf PWM period, nd the ON time Fig. 1 Current wveforms Devie turn-on Devie turn-off Top the output urrent i ; middle the hetsink ground urrent t the inverter output point ignd1; ottom the ground urrent t the power supply point ignd2 IET Eletr. Power Appl., 2011, Vol. 5, Iss. 4, pp doi: /iet-ep & The Institution of Engineering nd Tehnology 2011

4 Fig. 2 PWM sequene in setor I with tive vetors eing extended to t min on eh phse is kept unhnged. The urrent derivtive signls n then e mesured during V 1 nd V 0. If it does not meet the minimum durtion t min, the other tive vetor, V 2, is extended nd ompensted y moving the swithing edges of phse in the seond PWM period. The position vetor is then otined from the di/ signls mesured during V 1, V 2, V 0 nd V 7, s desried in Setion 2. Modifying the stndrd PWM sequene when neessry [14, 15] n use dditionl motor urrent distortion, nd this results in torque ripple nd ousti noise. These effets should therefore e minimised. A new method is introdued here to study the urrent ripple during one PWM period. In spe vetor PWM voltge soure inverter (VSI), the sttor voltge referene vetor is implemented using voltge vetor swithing sequene. The mgnitude of these swithing stte vetors is normlised with respet to the DC-link voltge. For exmple, the referene voltge vetor u s with mgnitude of U DC /2 nd phse ngle in setor I is shown in Fig. 3. The durtions of the tive vetors V 1 nd V 2, tht is, t 1 nd t 2, re represented y the lengths of the vetors on the vetor plot. In eh PWM period, the null vetor time for V 0 nd V 7, re set to hlf the totl null vetor time. There will e n error etween the instntneous voltge vetor nd the referene vetor. The voltge errors orresponding to the tive vetors V 1, V 2 nd zero vetors V 0 (V 7 ) n e expressed s These voltge errors tht re referred to the referene vetor n e defined s the voltge ripple vetor, represented y u ripple, nd the time integrls of these error voltge vetors determine the distortion in the output voltge [17] nd the sttor urrent. A mesure of the phse urrent ripple n e defined orrespondingly s urrent ripple vetor, i ripple.if the sttor resistne voltge nd the k-emf re negleted, the voltge nd urrent ripple vetors re orthogonl. Tht is, t very low speed, u d L d (di d /), u q L q (di q /), where L d nd L q re the orthogonl totl indutne on the rotting referene frme. Also, in PMSM, the ripple vetors n e resolved long the d nd q xes, in the referene frme synhronous with the rotting referene vetor u s. The q-xis, in ordne with the torque urrent diretion, is ligned with u s. Fig. 4 illustrtes the voltge ripple vetor plot for one fundmentl PWM yle. The swithing sequene in the first hlf yle is indited y the lk line nd tht in the seond hlf yle is indited y the grey line. Aordingly, in Fig. 4 the d nd q-xis voltge nd urrent ripple re plotted ginst time. The totl urrent hrmoni distortion n e evluted y summing the res onfined y the ripple wveforms nd the zero xes. The mgnitudes of oth voltge ripple nd urrent ripple over one fundmentl PWM period therefore depend on the referene vetor nd the employed swithing sequene. DV 1 ref = V 1 u s DV 2 ref = V 2 u s DV 0 ref = DV 7 ref = u s (11) 4.1 Simple vetor ompenstion sheme The urrent ripple used y the PWM vetor extension nd ompenstion shemes used in the slieny position estimtion sed on the FPE n now e studied. When the Fig. 3 Stndrd fundmentl PWM Vetor plot Swithing sequene 362 IET Eletr. Power Appl., 2011, Vol. 5, Iss. 4, pp & The Institution of Engineering nd Tehnology 2011 doi: /iet-ep

5 Fig. 4 Voltge nd urrent ripple of the stndrd fundmentl PWM sequene Vetor plot of voltge ripple Voltge nd urrent ripples in the d q referene frme referene voltge vetor is in setor I nd V 1 needs to e extended for orret di/ smpling, the resultnt voltge nd urrent ripple vetors re plotted in Fig. 5. The dshed line indites the ripple vetor of the stndrd fundmentl PWM sequene (without extension) for omprison. As n e seen, the min disdvntge of this sheme is tht lrge urrent devition from the fundmentl urrent ripple exists during the long intive period of V 7, lthough this offset is ompensted lter. Signifint devitions exist in oth the d nd q-xis urrent omponents. 4.2 Improved vetor ompenstion sheme To redue this dditionl urrent devition, modified ompenstion sheme is proposed where the ompenstion is mde in dvne. When n tive voltge vetor needs Fig. 5 Voltge nd urrent ripples of the PWM with extension of V 1 Vetor plot of voltge ripple Voltge nd urrent ripples in the d q referene frme IET Eletr. Power Appl., 2011, Vol. 5, Iss. 4, pp doi: /iet-ep & The Institution of Engineering nd Tehnology 2011

6 Fig. 6 PWM sequene with extension of V 1 nd V 2 in setor I using improved ompenstion sheme to e extended y time t ext to t min, its opposite voltge vetor should e pplied during the lose null vetor, with wih equl to the extension t ext. Fig. 6 shows opertion for setor I. V 1 (t 1 ) is not long enough nd is therefore extended to t min. The opposite vetor V 4 is then inserted t the eginning of V 7 to ompenste this effet, with durtion equl to the extended prt t ext (t ext ¼ t min 2 t 1 ). In the seond PWM period, if V 2 is extended, its opposite vetor V 5 is dded for ompenstion. The urrent ripple introdued y this method is plotted in Fig. 7. By ompring Fig. 7 with Fig. 5, it n e seen tht the urrent devition is muh lower, tht is, the urrent devition is redued y this improved ompenstion sheme. This ompenstion still voids the swithing of two phses t the sme instnt nd therefore no dditionl voltge stress is put on the winding insultion. To illustrte the urrent distortion quntittively, the sttor urrents mesured on the experimentl system desried in Setion 6 re presented for these two extension/ ompenstion shemes. The PM mhine is ontrolled t 100 rpm without n externl lod. The PWM period is 100 ms nd t min is set s 8 ms (note tht no ommon mode or differentil mode ompenstion hs een inorported s yet). Fig. 8 shows the phse urrent nd the PWM sequene when u s is loted in the middle of setor I nd the simple ompenstion sheme is used. The lrge urrent devition due to the pulse extension n e oserved in the middle of the wveform. The urrent wveform for opertion with the improved ompenstion sheme is shown in Fig. 8. There is signifint redution in the urrent devition when ompred to the simple ompenstion sheme. Bringing forwrd the ompenstion n effetively redue the distortion used y the vetor ompenstion. 5 Redution of ommon mode nd differentil mode effets The vetor extension nd ompenstion shemes ring out dditionl urrent ripple. For this reson, if the effet of the urrent swithing osilltions n e redued, the vetor extension n e minimised, nd thus the urrent ripple n e suppressed. Fig. 7 Voltge nd urrent ripples of the PWM with extension of V 1 nd improved ompenstion Vetor plot of voltge ripple Voltge nd urrent ripples in the d q referene frme 364 IET Eletr. Power Appl., 2011, Vol. 5, Iss. 4, pp & The Institution of Engineering nd Tehnology 2011 doi: /iet-ep

7 Fig. 8 Current ripple for different ompenstion shemes Simple ompenstion sheme Improved ompenstion sheme 5.1 Redution of mesured ommon mode noise Normlly for the AC inverter, if the three-phse lyout is lned nd les re the sme, the ommon mode urrents (i om, i om, i om ) flow eqully through three phses, nd the ground urrent is the sum of them. The motor urrents re i = i 1 + i om i = i 1 + i om (12) i = i 1 + i om where i 1, i 1, i 1 re the urrent omponents without ommon mode noise nd Therefore i 1 + i 1 + i 1 = 0 (13) i om = i om = i om = 1 3 (i + i + i ) (14) eliminted, ut the durtion of urrent swithing osilltions re not signifintly redued. The new method proposed here is to inrese the IGBT gte resistne. The gte resistne hs signifint impt on the dynmi performne of the IGBT swithing. A smller gte resistne will hrge nd dishrge the gte pitne fster, reduing the swithing time nd swithing losses, while using high dv/. A lrger gte resistne will redue dv/ t the output (nd the ssoited high frequeny osilltions) t the expense of dditionl swithing losses. A gte drive iruit providing hngele gte resistne hs een developed nd pplied s shown in Fig. 10. R G1 nd R G2 re onneted in prllel. The i-diretionl swith, whih is relised y MOSFET iruit, is ontrolled s prt of the min ontrol lgorithm. During norml opertion, this swith is turned ON nd the nominl gte resistne for norml opertion of the inverter (the prllel vlue of R G1 nd R G2 ) is onneted in the iruit. When sttor di/ signls re to e mesured for position estimtion, the gte resistne is inresed y turning OFF the MOSFET swith. The turn-on wveforms of the gte driving voltge v GE, the output voltge v CE nd the output urrent i o of the lower A similr reltionship exists for the urrent derivtives, tht is di om = di om = di om = 1 ( di 3 + di + di ) (15) Using these equtions, the ommon mode noise n e removed from the mesured di/ signls y sutrting (15) from eh di/ mesurement. The effet of the ommon mode ompenstion on the estimted position signls in the experimentl system is shown in Fig. 9. Itis ler tht the ommon mode omponents ppering in the position signls re signifintly redued. 5.2 Redution of differentil mode noise hngele IGBT gte resistne To redue oth ommon mode nd differentil mode urrent osilltions, when di/ mesurements need to e mde for position estimtion, it is possile to hnge the rte of hnge of inverter output voltge. One method is to dd snuer iruit, whih n redue dv/ y trnsferring the high frequeny swithing energy from the IGBT output to n energy storge element. For this work snuer pitors hve een dded nd the output voltge spikes re Fig. 9 Estimted position signls Without nelltion of the ommon mode noise With nelltion of the ommon mode noise IET Eletr. Power Appl., 2011, Vol. 5, Iss. 4, pp doi: /iet-ep & The Institution of Engineering nd Tehnology 2011

8 Fig. 10 IGBT gte drive with hngele gte resistne IGBTs in one phse ridge, during one swithing instnt, re shown in Fig. 11. The IGBT type is SKM 50 GB 123D [18]. The reommended gte resistne is R G ¼ 28 V. A vlue of R G ¼ 56 V hs een hosen for use when tking di/ mesurements. This vlue ws seleted s it extends the turn-on time from 420 to 700 ns, nd lso extends the turnoff time from 80 to 120 ns [18]. From onsulttion with [18], the inrese in swithing loss per swithing opertion is eptle, nd it should e noted tht this inresed loss only ours when nrrow swithing vetor is demnded from the PWM unit nd urrent derivtive mesurement is required. The urrent wveforms in Fig. 11 n e ompred diretly. The high frequeny osilltions, oth on the urrent i o nd on the voltge v CE, re muh redued with the lrger gte resistne. When R G ¼ 28 V, v CE flls from 620 to 0 V within less thn 0.5 ms nd the urrent osilltion lsts out 5 ms. When R G ¼ 56 V, v CE settles t its stedy stte in out 1 ms. The urrent settling time is just less thn 3 ms. With redued urrent swithing osilltions, the minimum vetor wih t min n e deresed to 5 ms nd y using this vlue the motor urrent ripple n e redued. For the system developed, the FPE method ws implemented every fourth PWM yle. The test rig ws operted using minimum pulse wih of 5 ms only when the FPE method is implemented. It ws found tht when operting t 1 Hz (eletril), minimum pulse wih ws required for every PWM vetor when the FPE method ws implemented, tht is, for 25% of the totl numer of swithing pulses per fundmentl period. Tht mens tht the lrger gte resistne (nd the improved vetor ompenstion sheme with its dditionl power swith ommuttion) ws used for 25% of the totl numer of swithing pulses per fundmentl period: this is n indition of the inrese in swithing losses ssoited with the proposed improvements for the sensorless ontrol lgorithm. For opertion t 1 Hz nd full lod, this redues to 24.8%. Similrly for opertion t 15 Hz, no lod, the inresed gte resistne is used 21% of the time, reduing to 14% t full lod. For opertion t 50 Hz (1/3 rted speed), the inresed gte resistne is used 8% of the time on no lod nd full lod. This will only e of signifine if the drive is to e operted t low or zero speed for sustined periods of time, in whih se, dditionl ooling my e required. 6 Experimentl setup nd results Fig. 11 R G ¼ 56 V R G ¼ 28 V IGBT turn-on wveforms A 20 kw power onverter ws onstruted for this reserh nd used to drive ommeril surfe mounted PMSM (4 kw, 6 pole, 3000 rpm). The overll struture of oth the power onverter nd the ontrol system is shown in Fig. 12. The ontrol system is relised through DSP-FPGA oordinted ontrol pltform. The speed nd position signls used for sensorless ontrol re otined using the position estimtion sheme sed on FPE. The rw position signl ontins hrmonis, predomintely t 4f e nd 8f e whih re present in the mhine due to higher order geometril nd sturtion slienies [5, 19]. An dptive disturne identifier is employed to eliminte these hrmonis. A mehnil 366 IET Eletr. Power Appl., 2011, Vol. 5, Iss. 4, pp & The Institution of Engineering nd Tehnology 2011 doi: /iet-ep

9 Fig. 12 Overll struture of the sensorless ontrol system oserver is lso used to improve the dynmi response of the estimted position nd speed signls [14]. A finl ompenstion is introdued to ompenste the rmture retion effet, wherey the identified ngle follows the min sturtion slieny rther thn the mgnet position. This ompenstion ngle is derived s funtion of lod from pre-ommissioned lookup tle [15]. The trnsient urrent responses re mesured through speilly designed di/ sensors. Three espoke ir-ore mutul indutors re used [3, 14]. The sensor hs two lyers of losely oupled windings. The sttor urrent psses through the primry winding nd the indued voltge ross the seondry winding whih is proportionl to the sttor urrent derivtive is mesured (As prt of ongoing work, the di/ will in future e otined from the urrent sensor mesurement.). To quntify the improvement to the motor urrent wveform due to the effets desried in this pper, the mesured totl hrmoni distortion (THD) of the sttor urrent is nlysed. To demonstrte the hnge due to the improved vetor ompenstion sheme of Setion 4, Tle 1 shows the THD of the sttor urrent for different ontrol shemes operting t 2 Hz nd full lod. The shemes employed re sensored vetor ontrol (i.e. the enhmrk where t min ¼ 0), sensorless ontrol using the improved INFORM method reported in [20] nd the FPE method proposed in this pper with the two vetor ompenstion shemes desried in Setion 4. The minimum vetor wih t min used for these results ws set to 5 ms, nd the ommon mode orretion ws lso employed. The FPE method with improved ompenstion sheme lerly hs lower THD ompred to the other sensorless methods. To quntify expliitly the improvement due to the use of vrile gte resistne for differentil mode redution, the system ws operted with the improved ompenstion nd the ommon mode nelltion. By reduing the minimum vetor time durtion or the test vetor wih t d min from 8 to 5 ms (y using the modified gte drive iruit), the THD of the motor urrent t 2 Hz nd full lod redues from to 8.19% signifint improvement. Both sensorless speed ontrol nd sensorless position ontrol hve een implemented using the ompenstion methods proposed nd t min of 5 ms. Fig. 13 shows the experimentl results of speed reversl t very low speed. The mhine is elerted from stndstill to 10 rpm, the speed is then reversed to 210 rpm, nd then the motor is rought to stndstill. Throughout the test, rted lod is pplied (i.e Nm). A 6f e periodi flutution ppers in the speed nd the torque urrent; this is produed y the ogging of the PM mhine, nd is of similr level to tht mesured under fully sensored ontrol. It n e seen tht oth the stedy-stte performne nd the dynmi response re exellent. The men speed error is pproximtely zero, nd the pek speed error is pproximtely 10 rpm t the instnt of torque hnge. The mximum speed error is 3 rpm during stedy-stte opertion. The qulity of the sensorless ontrol hs not een redued y the redution of the minimum vetor length. The proposed sensorless ontrol using FPE lso works in the high speed rnge. Fig. 14 shows the experimentl results when the motor is elerted to 1000 rpm without externl lod. The dynmi response meets speifitions nd the estimtion errors of oth the speed nd the eletril position ngle re eptle. The men speed error is zero. The pek position error is 0.15 rd, with men error of 0.06 rd. The ltter illustrtes the diffiulty in otining perfet ompenstion for rmture retion effets, ut is onsidered eptle. Fig. 15 shows the experimentl results for sensorless position ontrol test. The mhine is driven six rottions from stndstill, held sttionry for 5 s nd then driven six rottions in the opposite diretion k to its initil position. In the upper plot, the mesured solute mehnil rotor position ngle follows the ommnd position signl very well. At t ¼ 2.5 s full lod torque is pplied, nd then relesed t t ¼ 4.5 s. Smll disturnes pper on the position wveform t these instnts, ut the Tle 1 THD of sttor urrents with different modifition shemes Position estimtion sheme Sensored vetor ontrol INFORM FPE with stndrd ompenstion FPE with improved ompenstion sttor urrent THD, % IET Eletr. Power Appl., 2011, Vol. 5, Iss. 4, pp doi: /iet-ep & The Institution of Engineering nd Tehnology 2011

10 Fig. 13 Sensorless ontrol results for low speed reversl with full lod ( rpm) v r mesured rotor speed, vˆ r estimted rotor speed, i sq torque urrent, u r mesured rotor ngle ontroller is le to quikly reover position with dynmi response omprle to the fully sensored position ontrol system. 7 Summry This pper desries the fundmentl PWM exittion method for sensorless ontrol, tht is, the use of the voltge vetors employed y the norml PWM lgorithm in vetor ontrolled drive to determine the rotor position. The sturtion slieny is trked y mesuring the sttor urrent trnsient response to these voltge vetors. The ury of the urrent trnsient (i.e. the urrent derivtive) mesurements during nrrow voltge vetors is of onern due to the effets of prsiti ringing. If nrrow swithing vetor exists when di/ mesurements need to e mde, these n e extended to improve the di/ mesurement, nd the extension n e ompensted lter in the PWM period using new ompenstion tehnique whih employs the opposite swithing vetor. A new method to nlyse the dditionl distortion used y the extension/ompenstion of nrrow vetors hs lso een presented. The nlysis demonstrtes the effetiveness of the proposed ompenstion method in reduing the THD of the motor urrent when nrrow swithing vetors re present. The use of the high frequeny prsiti ringing hs lso een disussed in terms of ommon mode nd differentil mode iruits whih respond to the high dv/ imposed on 368 IET Eletr. Power Appl., 2011, Vol. 5, Iss. 4, pp & The Institution of Engineering nd Tehnology 2011 doi: /iet-ep

11 Fig. 14 Sensorless ontrol results for high speed opertion ( rpm) v r mesured rotor speed, vˆ r estimted rotor speed, u r mesured rotor ngle, v r estimted rotor ngle Fig. 15 Results for sensorless position ontrol: u r mesured rotor ngle, v r mesured rotor speed, i sq torque urrent the drive when n IGBT hnges stte. A novel modifition to the IGBT gte drive iruit is proposed to inrese the gte resistne vlue nd thus to suppress the osilltions during swithing periods when di/ mesurement must e mde. This effetively dereses the requirement for the minimum vetor durtion t min. The inrese in gte resistne (only IET Eletr. Power Appl., 2011, Vol. 5, Iss. 4, pp doi: /iet-ep & The Institution of Engineering nd Tehnology 2011

12 when required for FPE mesurement), nd the modified ompenstion method will inrese the onverter swithing losses. The requirement for dditionl inverter ooling will e dependent on the prtiulr pplition of the drive. It ould e rgued tht this pproh trnsfers dditionl motor losses inurred when sensorless ontrol is implemented to the power onverter losses nd further reserh is required here. Common mode ompenstion n lso e hieved through simple softwre lgorithm. Full sensorless speed nd position ontrol hve een implemented using the proposed tehniques on experimentl rig employing 4 kw surfe mounted PMSM. The FPE method with the proposed modifitions hs een shown to hve good stedy-stte nd dynmi performne over wide speed rnge. 8 Referenes 1 Jnsen, P.L., Lorenz, R.D.: Trnsduerless field orienttion onepts employing sturtion indued slienies in indution mhines. IEEE IAS Annul Meeting, 1995, vol. 1, (30), pp Holtz, J.: Sensorless position ontrol of indution motors n emerging tehnology, IEEE Trns. Ind. Eletron., 1998, 45, (6), pp Shroedl, M.: Sensorless ontrol of AC mhine t low speed nd stndstill sed on the INFORM method. IEEE Industry Applitions Soiety Annul Meeting, Pittsurgh, 1996, 30 Septemer 4 Otoer, pp Degner, M.W., Lorenz, R.D.: Using multiple slienies for the estimtion of flux, position nd veloity in AC mhines, IEEE Trns. Ind. Appl., 1998, 34, pp Teske, N., Asher, G.M., Sumner, M., Brdley, K.J.: Suppression of sturtion slieny effets for the sensorless position ontrol of indution motor drives under loded onditions, IEEE Trns. Ind. Eletron., 2000, 47, (5), pp Jng, J.H., H, J.I., Sul, S.K.: Vetor ontrol of surfe mounted permnent mgnet motor without ny rottionl trnsduer. Pro. IEEE Applied Power Eletronis Conf. nd Exposition (APEC), USA, 4 8 Mrh 2001, pp Linke, M., Kennel, R., Holtz, J.: Sensorless position ontrol of permnent mgnet synhronous mhines without limittion t zero speed. Pro. IEEE Industril Eletronis Soiety Annul Conf. (IECON), Spin, 5 8 Novemer 2002, pp Wolnk, T., Mhl, J.: A modified PWM sheme in order to otin sptil informtion of mhines without mehnil sensor. Pro. IEEE APEC, Dlls, TX, 2002, pp Holtz, J., Juliet, J.: Sensorless quisition of the rotor position ngle for indution motors with ritrry sttor windings, IEEE Trns. Ind. Appl., 2005, 41, (6), pp Choi, J., Sul, S.: A new ompenstion strtegy reduing voltge/urrent distortion in PWM VSI systems operting with low output voltges, IEEE Trns. Ind. Appl., 1995, 31, (5), pp Holtz, J., Qun, J.: Sensorless vetor ontrol of indution motors t very low speed using nonliner inverter model nd prmeter identifition, IEEE Trns. Ind. Appl., 2002, 38, (4), pp Bisheimer, G., Sonnillon, M., De Angelo, C., Solson, J., Grí, G.: Full speed rnge permnent mgnet synhronous motor ontrol without mehnil sensorless, IET Eletr. Power Appl., 2004, 4, (35), pp Silv, C., Asher, G.M., Sumner, M.: Hyrid rotor position oserver for wide speed-rnge sensorless PM motor drives inluding zero speed, IEEE Trns. Ind. Eletron., 2006, 53, (2), pp Hu, Y., Asher, G.M., Sumner, M., Go, Q.: Sensorless ontrol of surfe mounted permnent mgnet mhine using the stndrd spe vetor PWM. IEEE IAS Annul Meeting, USA, 2007, pp Go, Q., Asher, G.M., Sumner, M., Mkys, P.: Position estimtion of AC mhines over wide frequeny rnge sed on spe vetor PWM exittion, IEEE Trns. Ind. Appl., 2007, 43, (4), pp Rn, L., Gokni, S., Clre, J., Brdley, K.J., Christopoulus, C.: Conduted eletro-mgneti emissions in indution motor drive system Prt I: time domin nlysis nd identifition of dominnt modes, IEEE Trns. Power Eletron., 1998, 13, (4), pp Nrynn, G., Zho, D., Hrish, K.K., Rjpndin, A., Rngnthn, V.T.: Spe vetor sed hyrid PWM tehniques for redued urrent ripple, IEEE Trns. Ind. Eletron., 2008, 55, (4), pp SEMIKRON: SEMITRANS IGBT Modules Dtsheet SKM 50 GB 123D 19 Holtz, J., Pn, H.: Elimintion of sturtion effets in sensorless position ontrolled indution motors. IEEE IAS Annul Meeting, Pittsurgh, USA, 2002, vol. 3, pp Roeishl, E., Shroedl, M.: Optimized INFORM mesurement sequene for sensorless PM synhronous motor drives with respet to minimum urrent distortion, IEEE Trns. Ind. Appl., 2004, 40, (2), pp IET Eletr. Power Appl., 2011, Vol. 5, Iss. 4, pp & The Institution of Engineering nd Tehnology 2011 doi: /iet-ep