scieee Open visual document viewer

Integrated Control and Modulation for Three-Level NPC Rectifiers

Ventosa Cutillas, Antonio; Montero-Robina, Pablo; Umbría Jiménez, Francisco; Cuesta Rojo, Federico; Gordillo Álvarez, Francisco

Abstract

This paper uses a novel approach for the control of three-level neutral-point-clamped (NPC) rectifiers in order to tackle the capacitor voltage balance problem. A distinctive feature of the new control approach is that it is based on a model which is written in terms of the duty ratios for each phase at each level. Hence, the system model presents nine duty cycle variables. Despite the fact that this formulation is different from the usual ones, it is shown that the control problem of currents and dc-link voltage can be formulated in a similar way to conventional methods. Furthermore, the control of the capacitor voltage balance can be expressed by means of equations that are decoupled from the currents and dc-link voltage dynamics, which results in a specific controller for the voltage balancing that does not affect the previous dynamics. A key point of the proposed approach is that part of the modulation stage is implicit in the formulation. Two particular controllers are compared in this paper. The first one fulfills the different control objectives at the expense of a large number of commutations. This problem is overcome in a new proposed controller, which presents similar performance and a satisfactory number of commutations. Experimental results are performed showing the effectiveness compared with a modified virtual space vector modulation with capacitor voltage balance capabilities.

Full text

ene gies A icle In eg a ed Con ol and Modula ion o Th ee-Le el NPC Rec i ie s An onio Ven osa-Cu illas 1,* , Pablo Mon e o-Robina 1, F ancisco Umb ía 2, Fede ico Cues a 1and F ancisco Go dillo 1 1Depa amen o de Sis emas y Au omá ica, Escuela Técnica Supe io de Ingenie ía, Uni e sidad de Se illa, 41092 Se illa, Spain; [email p o ec ed] (P.M.-R.); [email p o ec ed] (F.C.); [email p o ec ed] (F.G.) 2ASM Assembly Sys ems GmbH & Co. KG, 81379 Munich, Ge many; [email p o ec ed] *Co espondence: [email p o ec ed]; Tel.: +34-95-448-2292 Recei ed: 8 Ap il 2019; Accep ed: 26 Ap il 2019; Published: 30 Ap il 2019   Abs ac : This pape uses a no el app oach o he con ol o h ee-le el neu al-poin -clamped (NPC) ec i ie s in o de o ackle he capaci o ol age balance p oblem. A dis inc i e ea u e o he new con ol app oach is ha i is based on a model which is w i en in e ms o he du y a ios o each phase a each le el. Hence, he sys em model p esen s nine du y cycle a iables. Despi e he ac ha his o mula ion is di e en om he usual ones, i is shown ha he con ol p oblem o cu en s and dc-link ol age can be o mula ed in a simila way o con en ional me hods. Fu he mo e, he con ol o he capaci o ol age balance can be exp essed by means o equa ions ha a e decoupled om he cu en s and dc-link ol age dynamics, which esul s in a speci ic con olle o he ol age balancing ha does no a ec he p e ious dynamics. A key poin o he p oposed app oach is ha pa o he modula ion s age is implici in he o mula ion. Two pa icula con olle s a e compa ed in his pape . The i s one ul ills he di e en con ol objec i es a he expense o a la ge numbe o commu a ions. This p oblem is o e come in a new p oposed con olle , which p esen s simila pe o mance and a sa is ac o y numbe o commu a ions. Expe imen al esul s a e pe o med showing he e ec i eness compa ed wi h a modi ied i ual space ec o modula ion wi h capaci o ol age balance capabili ies. Keywo ds: synch onous ec i ie applica ion; neu al-poin -clamped (NPC) con e e ; ol age balancing; in eg a ed con ol and modula ion (ICM); G id-connec ed powe con e e ; mul ile el con e e s 1. In oduc ion In he ield o ene gy con e sion sys ems, he ad an ages ha mul ile el powe elec onic con e e s o e a e well known. Bidi ec ional powe low, inc ease o he ou pu ol age magni ude, obus ness, e c. a e only some o he ad an ages ha ha e made mul ile el con e e s popula in medium and high powe applica ions in he indus y [ 1 – 3 ]. Some o he di e en ypes o opologies o hese con e e s a e neu al-poin -clamped, cascaded H-b idge o lying-capaci o [ 1 , 2 , 4 ]. A Neu al-poin -clamped (NPC) con e e , which was p oposed o he i s ime in [ 5 ], is one o he mos used mul ile el con e e opologies. Du ing no mal ope a ion, he ol age ha d ops ac oss each capaci o mus be balanced o he wise, i can esul in poo ou pu ol age quali y, a ec ing he pe o mance o he con ol o e en damaging he semiconduc o de ices. The e o e, NPC con e e s p esen an addi ional objec i e apa om he usual con ol objec i es in powe con e e s [ 6 ] ha is he ol age balance be ween capaci o s, which is he main ocus in his pape . O e he las ew yea s, nume ous echniques ha e been de eloped o co ec he ol age unbalance. Some o hem use addi ional ci cui y [ 7 – 9 ], bu his would lead o an inc ease in cos , losses and complexi y in ha dwa e. O he au ho s use di e en con ol echniques wi h algo i hms o a ying di icul y [ 10 – 15 ]. One o hese con ol echniques [ 13 ] uses a modi ica ion o i ual space Ene gies 2019,12, 1641; doi:10.3390/en12091641 www.mdpi.com/jou nal/ene gies Ene gies 2019,12, 1641 2 o 15 ec o [ 16 ] (deno ed as mVSVPWM in his pape ). In [ 13 ], he ol age e e ence ec o is ob ained by c ea ing i ual ec o s whe e he possible swi ching s a es a e weigh ed o gene a e cu en s ha bene i he ol age balance. Rega ding he modelling, i is usual o wo k wi h a e aged models whe e he disc e e alues o he ga ing elemen s a e conside ed as con inuous signals [ 1 , 17 ]. In o de o implemen he con ol laws ob ained wi h such models, a disc e iza ion s age, usually called modula ion, needs o be accomplished [ 18 ]. Modula ion plays an impo an ole in he o e all sys em pe o mance since p ope ies such as numbe o commu a ions and ha monic dis o ion o cu en s and ol ages a e a ec ed by he way modula ion is ca ied ou . Modula ion me hods can be classi ied in o h ee main g oups [ 19 ]: pulse wid h modula ion (PWM) [ 20 , 21 ], pseudo-modula ion [ 22 ] and closed-loop con ol me hods wi h implici modula o [23–25]. This pape p esen s a new app oach o deal wi h he con ol o h ee-le el NPC con e e s, which is based on [ 26 ]. In his pape , he ci cui model is o mula ed in e ms o he du y a ios o each phase a each le el. In his way, he e a e nine du y cycle a iables ( h ee du y cycles pe phase) ins ead o jus h ee (one du y cycle pe phase). This o mula ion is no new, e.g., a simila model, based on d–q ans o ma ion, was p esen ed in [ 27 ] o design an LQR con olle o an NPC in e e by linea iza ion using a small-signal model. In [ 26 ], i is shown ha his o mula ion allows o explici ly conside , in he con ol design s age, he ex a deg ee o eedom associa ed wi h he injec ion o homopola componen . The inc ease in he numbe o a iables does no make he design signi ican ly mo e di icul since, wi h an app op ia e change o a iables, he dc-link ol age and ac i e and eac i e powe con ol p oblems can be o mula ed in a simila way o o he usual app oaches. As a esul , he ol age balance con olle can be easily designed a he same ime ha an impo an pa o modula ion is no needed. Fo his, he p oposed app oach can be conside ed as a con ol me hod wi h pa o he modula ion s age included in he con ol o mula ion, he e o e, in wha ollows i would be called “In eg a ed Con ol and Modula ion” (ICM). The main ad an age o he p oposed con ol law is i s simplici y in implemen a ion compa ed o modi ied e sions o space ec o modula ion (SVM) [ 16 ] ha also ackle he capaci o ol age unbalance bu i s ill p esen s some ad an ages wi h espec o CB-PWM app oaches. This is due o he ac ha he modula ion s age is simpli ied wi hou losing pa o he lexibili y o SVM [ 28 , 29 ]. Once he nine du ies a e compu ed, he way hey a e sequenced can be chosen eely, spli ing hem up o shi ing hem among he di e en phases. This eedom allows he use o achie e seconda y con ol objec i es simila ly o SVM such as common-mode ol age educ ion o a oidance o ex a swi ching losses. Fo he sake o a ai compa ison, his a icle will use a simple iangula -shaped sequence simila o ha o CB-PWM app oaches. Fu he mo e, he p oblem o mula ion o powe , cu en and dc-link ol age con ol is he same as when using o he con en ional app oaches, such as model-based di ec powe con ol (DPC) [30–33] o p opo ional- esonan con olle (PR) o cu en s [34,35]. The d awback o he p oposed app oach in [ 26 ] is ha i may lead o an unnecessa y inc ease in he numbe o commu a ions. This is due o he ac ha , unless some o he du y cycle a iables u n ou o be ze o, he esul an swi ching signals will commu e among all he le els o he h ee phases e e y sampling pe iod. This is he case o he i s con ol law conside ed in his pape (ICM1), whe eas wi h he dc-link ol age, cu en o powe con ol can be accomplished by swi ching each phase be ween wo le els [ 36 ] when he ol age balance p oblem is no conside ed. In his pape , by using a emaining deg ee o eedom associa ed wi h he injec ion o homopola componen , some commu a ions a e a oided compa ed o ICM1 p esen ing a second con olle (ICM2) who explodes his capabili y. The ou pu wa e o m o ICM2, as i will be shown la e on in expe imen s, is simila o hose app oaches which injec a hi d ha monic signal in o he ou pu ol age o inc ease he undamen al signal ange wi hou o e modula ing [ 37 ]. The e o e, ICM2 also has his p ope y inhe en wi hou he need o ex a compu a ions. Bo h con ol laws a e alida ed by means o expe imen s and compa ed wi h a con ol echnique based on a modi ied i ual space ec o modula ion (mVSVPWM) [ 13 ], which includes capaci o Ene gies 2019,12, 1641 3 o 15 ol age balance capabili ies. The e o e, he main con ibu ions o his pape in compa ison wi h [ 26 ] a e he inclusion o a new app oach (ICM2) along wi h expe imen al e i ica ion o bo h app oaches. The nex sec ion is de o ed o p esen ing he con e e conside ed in his pape , as well as i s dynamic model. Sec ion 3p esen s he con olle design o ul illing he h ee con ol objec i es: egula ion o he cu en s and he dc-link ol age and capaci o ol age balance. Sec ion 4p oposes wo a ian s o he selec ion o he emaining deg ees o eedom esul ing in wo di e en con olle s, ICM1 and ICM2. Sec ion 5p esen s expe imen al esul s. The pape closes wi h a sec ion o conclusions. 2. Dynamic Model o he Sys em The con igu a ion o he con e e used in his pape is a h ee-phase h ee-le el NPC con e e in ec i ie mode wi h a esis i e load, as shown in he scheme o Figu e 1. Figu e 1. Schema ic diag am o he h ee-phase h ee-le el neu al-poin -clamped (NPC) ec i ie . The elec ical powe g id is conside ed as a h ee-phase ol age sou ce, whe e he phase ol ages a e ep esen ed by sa , sb and sc . The con e e is connec ed o he g id h ough an induc i e il e whe e induc ances ha e he same alue L. On he dc-link side, capaci o s ha e he same alue C and hei ol ages a e deno ed by c1 and c2 . Connec ed o he con e e e minals he e is a esis i e load R . The o al dc-link ol age is de ined as dc = c1+ c2. Conside ed Sys em Model The model conside ed in his pape is desc ibed in [ 26 ], which is based on a model p esen ed in [ 27 ]. This model uses he equa ions in αβγ coo dina es by in oducing he powe -in a ian o m o he Cla ke T ans o m. Fu he mo e, he swi ching signals ha e been eplaced by hei espec i e du y a ios in each le el [ 27 , 38 ], dij wi h i=α , β , γ and j=p , o , n , whe e p is he posi i e le el when swi ches Si 1 and Si 2 a e closed, o is he ze o le el when swi ches Si 1 and Si 2 a e closed and n is he nega i e le el when swi ches Si 1and Si 2a e closed. This o mulism yields Ldiα d = sα−(dαp−dαn) dc 2−(dαp+dαn) d 2(1) Ldiβ d = sβ−(dβp−dβn) dc 2−(dβp+dβn) d 2(2) Cd dc d = (dαp−dαn)iα+ (dβp−dβn)iβ−2 dc R(3) Cd d d = (dαp+dαn)iα+ (dβp+dβn)iβ, (4) Ene gies 2019,12, 1641 4 o 15 whe e d is he dc-link capaci o ol age di e ence de ined by d= c1− c2 . The con ol inpu s dαp , dαn , dβp and dβn a e he du y a ios in αβγ coo dina es. Con ol inpu s dγp and dγn do no appea in he model, as men ioned in [ 26 ], because hey a e mul iplied by iγ , whose alue is ze o o a balanced sys em. Simila ly, a iables dio do no appea in his model bu hei alues can be e ie ed a he inal s age o he con olle using he ollowing cons ain s: dap +dao +dan =1 (5) dbp +dbo +dbn =1 (6) dcp +dco +dcn =1 (7) dij ∈[0, 1], o i=a,b,cand j=p,o,n. Phase cu en s iαand iβcan be exp essed in e ms o powe s as iα=1 2 sα+ 2 sβ sαp− sβq(8) iβ=1 2 sα+ 2 sβ sβp+ sαq, (9) whe e p and q a e he ins an aneous ac i e and eac i e powe s o he sys em, espec i ely. In his way, (3) and (4) could be exp essed as Cd dc d =1 2 sα+ 2 sβdαp−dαn sαp− sβq+1 2 sα+ 2 sβdβp−dβn sβp+ sαq−2 dc R(10) Cd d d =1 2 sα+ 2 sβdαp+dαn sαp− sβq+1 2 sα+ 2 sβdβp+dβn sβp+ sαq, (11) whe e a iables dc and da e exp essed in e ms o he ins an aneous powe pand q. 3. Con olle Design Wi h he pu pose o dealing wi h he h ee con ol objec i es (cu en s, dc-link ol age and capaci o ol age balance con ol), he sys em dynamic model (1) – (4) p esen ed p e iously is conside ed o design he con olle s in his sec ion. I can be seen ha he p oposed modeling allows o cope wi h he capaci o ol age balance p oblem while i does no a ec he cu en and dc-link ol age con olle s. 3.1. To al DC-Link Vol age Con olle In o de o main ain cons an he dc-link ol age and close o i s e e ence ( dc ), as usual, a PI con olle is used [26,30,39], p =kdc p dc 2− 2 dc+kdc iZ 0 dc 2− 2 dcdτ, (12) whe e cons an s kdc pand kdc ia e con olle uning pa ame e s. Ene gies 2019,12, 1641 5 o 15 3.2. Cu en Con olle Obse ing Equa ions (1) and (2), wo i ual con ol a iables can be de ined as u1˙=dαp−dαn(13) u2˙=dβp−dβn. (14) In oducing hese a iables in o he cu en s dynamic model and assuming ha he alue o a iable dis small enough o be neglec ed, Ldiα d = sα−u1 dc 2(15) Ldiβ d = sβ−u2 dc 2. (16) These exp essions a e equi alen o hose cu en dynamics o he con en ional wo-le el con e e [ 26 ]. The e o e, by he use o he change o a iables (13) and (14), he added di icul y inhe en o he adop ed o mula ion disappea s, a leas a his s age. Addi ionally, in oducing he e e ences o he ac i e ( p ) and eac i e powe ( q ) in o (8) and (9) , he cu en e e ences can be e ie ed ( i α , i β ). Once hese alues a e known, a non-ideal p opo ional- esonan con olle [ 35 ], uned a he g id equency, is used o make he phase cu en s o ack hei e e ences. In his way, he acking e o (i α−iα , i β−iβ) inpu s he esonan con olle , p o iding he alue o con ol a iables (u1,u2). GPRω(s) = kp+2k ωcs s2+2ωcs+ω2 u1=2 dc (−GPRωg(i α−iα) + sα)(17) u2=2 dc (−GPRωg(i β−iβ) + sβ)(18) whe e kp and k a e he p opo ional and esonan con ol pa ame e s; ωc is he cu -o equency o he low-pass il e implemen ed in o he esonan pa ; and ω is he esonan equency— uned a he g id equency ωg. 3.3. Vol age Balance Con olle The objec i e o he ol age balance con olle is o keep he s a e a iable d close o ze o, a oiding he unbalance o he dc-link capaci o ol ages, and i is based on he de ini ion o wo new i ual con ol a iables u3˙=dαp+dαn(19) u4˙=dβp+dβn. (20) In oduc ion o (19) and (20) in o (11) yields Cd d d = sαp− sβq 2 sα+ 2 sβ u3+ sβp+ sαq 2 sα+ 2 sβ u4. (21) I is impo an o highligh ha he de ini ion o u3 and u4 causes a decoupling o he i ual con ol a iables o con ol pu poses. No e ha u1 and u2 a e designed o egula e he s a e a iables p and q , whe eas u3 and u4 can be used o egula e d(21) . This is an impo an bene i o he p oposed con ol app oach. Ene gies 2019,12, 1641 6 o 15 Taking in o accoun (21), he p oposed con ol laws [26] a e de ined as ollows u3=kd sαp− sβq p2+q2( d− d)+kdi sαp− sβq p2+q2Z 0 ( d− d)dτ(22) u4=kd sβp+ sαq p2+q2( d− d)+kdi sβp+ sαq p2+q2Z 0 ( d− d)dτ, (23) whe e posi i e cons an s kd and kdi a e cus oma y uning pa ame e s. The e e ence o d is deno ed by d , which is se o ze o o ensu e a balanced dis ibu ion o he dc-link ol age ac oss capaci o s C1 and C2. By in oducing (22) and (23) in o (21), he ol age balance dynamics become linea : Cd2( d− d) d 2+kd d( d− d) d +kdi( d− d) = 0, (24) whose s abili y is assu ed p o ided ha pa ame e s kd and kdi a e posi i e. A comple e schema ic block diag am o all con olle s is illus a ed in Figu e 2. Figu e 2. Comple e schema ic block diag am o he con olle s. Ene gies 2019,12, 1641 7 o 15 4. Modula ion S a egy The con olle p esen ed in he p e ious sec ion p o ides, a each sampling ime, he alues o u1 , u2 , u3 and u4 . The co esponding alues o dαp , dαn , dβp and dβn can be ob ained sol ing he sys em o Equa ions (13), (14), (19) and (20) yielding dαp=1 2(u1+u3)(25) dαn=1 2(−u1+u3)(26) dβp=1 2(u2+u4)(27) dβn=1 2(−u2+u4)(28) In o de o compu e he ac ual du y a ios dij , i=a , b , c ; j=p , n , he Cla ke ans o ma ion can be used    daj dbj dcj   = 2 3         1 0 1 √2 −1 2 √3 2 1 √2 −1 2−√3 2 1 √2            dαj dβj dγj   ,j=p,n. (29) whe e dγp and dγn a e emaining deg ees o eedom. Ob iously, he emaining du y a ios, dao , dbo and dco can be compu ed using (5)–(7). 4.1. Fi s P oposal, ICM1 The i s a ian , ICM1, was p oposed in [ 26 ]. In his a ian , dγp and dγn a e chosen o be cons an and can be conside ed as uning pa ame e s. In [ 26 ], guidelines a e gi en in o de o a oid sa u a ion p oblems. This app oach is a simple way o accomplish he modula ion bu i p esen s an impo an d awback: excep by chance, none o he dij will esul in ze o. This implies ha , in each sampling ime, each phase commu es among he h ee le els, which can be conside ed oo many commu a ions compa ed wi h o he con olle s, e en hose including ol age balancing. This ac could yield la ge swi ching losses. 4.2. Second P oposal, ICM2 In o de o a oid he la ge numbe o commu a ions, a new a ian is p oposed. This a ian akes ad an age o he wo deg ees o eedom associa ed wi h dγp and dγn by imposing, a each sampling pe iod, one dip and one din o be ze o. The wo phases o which one o he du y a ios has o be ze o need o be chosen ca e ully, as i is explained below, aking in o accoun ha he du y cycles in abc coo dina es ha e o be in he in e al [ 0, 1 ] . The esul is ha hese wo phases only swi ch be ween wo le els while he emaining phase commu es among he h ee le els. In o de o selec hese phases, he p ocedu e checks i se ing one o he dij,i=a,b,c;j=p,n o ze o yields o he ul illmen o he cons ain s 0 ≤dij ≤ 1 o he es o du y cycle a iables. S a ing wi h he case dap = 0, Equa ion (29) o j=p can be conside ed as a se o equa ions, whe e dαp , dβp a e known, dap = 0 and dbp , dcp and dγp a e he unknowns. The esul an sys em o equa ions can be sol ed in o de o check i his case is easible, ha is, i dbp and dcp a e in he in e al [ 0, 1 ] . Repea ing o he o he phases, h ee di e en cases ha e o be analyzed o j=p and o he h ee cases o j=n . The associa ed equa ions a e Ene gies 2019,12, 1641 8 o 15 Case 1: daj =0 dbj =−√6 2dαj+√2 2dβj(30) dcj =−√6 2dαj−√2 2dβj(31) Case 2: dbj =0 daj =√6 2dαj−√2 2dβj(32) dcj =−√2dβj(33) Case 3: dcj =0 daj =√6 2dαj+√2 2dβj(34) dbj =√2dβj, (35) whe e j = p , n . The esul an du y a ios o he conside ed sampling ins an a e he co esponding ones o he easible cases, ha is, cases whe e all he du y cycles a e in he in e al [ 0, 1 ] . I will be shown below ha , a e e y ins an , he e exis s a leas one (and apa om some bo de cases, only one) easible case. Rega ding he compu a ion bu den o his app oach, a e e ie ing (25)–(28), he calcula ion o he du y a ios dij can be achie ed by checking he cons ain s dij ∈[ 0, 1 ] o he 3 × 2 = 6 cases. Fo his, (30)–(35) ha e o be used wice ( o le els p and n ). This las s age implies he compu a ion o 20 mul iplica ions and 8 sums as well as 24 compa isons. Al e na i ely, a di e en app oach can be used o choose he co ec case a e e y sampling ins an wi h he help o Figu e 3. This igu e depic s he s aigh lines ha a e he bounda ies o cons ain s 0 ≤dij ≤ 1 using exp essions (30) – (35) . Consequen ly, he shaded egions ep esen he ul illmen o hese cons ain s o each one o he h ee cases abo e. As o each le el j=p , n he e is one dij ha i is equal o ze o, he e a e ou o such lines o each case, ins ead o six. I can be seen ha he h ee shadowed a eas do no o e lap (excep a hei bo de s) and ha hey co e a whole hexagon. An in e es ing ac is ha his hexagon is ela ed o he well-known hexagon o Space Vec o Modula ion (SVM), bu in his case, he e a e wo such hexagons (one o j=p and one o j=n ). The in e es o Figu e 3is wo old: (1) i shows ha , p o ided he con e e does no wo k in o e modula ion egion, one o he h ee cases is always easible, and (2) Figu e 3can be used as an al e na i e me hod o choose he co ec case: he wo king sec o can be compu ed as usual in SVM and once he app op ia e case is selec ed, he co esponding o mulae can be applied. In summa y, he ICM2 algo i hm, whose da a inpu and ou pu a e depic ed in Figu e 4, can be implemen ed in wo al e na i e and equi alen ways ha only di e in he compu a ional bu den, which in al e na i e B depends on he me hod used o he compu a ion o he sex an : Al e na i e A: •Compu a ion o u3and u4using (22) and (23). •Compu a ion o dαp,dαn,dβp,dβnusing (25)–(28). •Fo le els pand n: – Compu a ion o Equa ions (30)–(35) and selec ion o he case ha ul ills he cons ain s 0≤dij ≤1. This p ocedu e gi es di ec ly he esul an du y cycles. Ene gies 2019,12, 1641 9 o 15 Al e na i e B: •Compu a ion o u3and u4using (22) and (23). •Compu a ion o dαp,dαn,dβp,dβn(25)–(28). •Fo le els pand n: – Compu a ion o he sex an inside he hexagon o Figu e 3. This s ep is simila o he co esponding one in SVM (bu i mus be compu ed wice, one o le el p and one o le el n). The sex an gi es he co esponding case. –Compu a ion o he du ies using he co esponding Equa ions o (30)–(35). I can be no ed ha he compu a ional complexi y is lowe in bo h ICM1 and ICM2 wi h espec o SVM s a egies ha include con ol o he ol age balance, such as [13]. Figu e 3. G aphical ep esen a ion o he limi s o each case. Figu e 4. Da a inpu and ou pu o ICM1 and ICM2. 5. Expe imen al Ve i ica ion This sec ion p esen s he expe imen al esul s ob ained in he labo a o y o he wo app oaches p esen ed in his pape and he one used as a compa a i e (mVSVPWM) [ 13 ]. To his end, he h ee-le el NPC con e e shown in Figu e 5has been used. I has been con igu ed as ec i ie and i has he ci cui and con ol pa ame e s p o ided in Table 1. A i s -o de low-pass il e has also been added o he p opo ional pa o he o al dc-link ol age con olle , as i is ecommended in [ 33 ], uned a 5 KHz in o de o educe he ha monics p esence.