ONLINE MONITORING OF CAPACITORS IN POWER CONVERTERS

Transcription

1 / (%) ES ( ) ONINE MONITOING OF APAITOS IN POWE ONVETES Gustao M. Buatt (1), Juan A. Martínamos (), Juan A. Martínez (), A. M.. Amaral (3), A. J. Marques ardoso (3) (1) Alstom Transport, Tarbes, France. He s currently wth Mtsubsh Electrc, France, g.buatt@fr.merce.mee.com () Unersdad de Oedo, D.I.E.E..S., ampus de Gón, Span, amartn@uno.es (3) Unersty of ombra / Insttuto de Telecomuncações, ombra, Portugal, amcardoso@eee.org Abstract In power conerters, the flterng or D lnk capactors are usually the component wth shortest lfetme. Wth the agng, both electrolytc and flm capactors do often burst out, causng serous damages or een the destructon of the whole power stage. In ths paper, a technque for dagnostc purposes s presented to estmate onlne, and een n real tme, the agng of capactors n dfferent conerters. The technque s based on the estmaton of the capactors equalent seres resstance (ES) and capactance (). In ths way, predcte mantenance can be carred out and t s possble to alarm for the capactor replacement before falure. In order to demonstrate and aldate the effecteness and accuracy of the proposed technque, seeral power topologes are dscussed, showng expermental results. An embedded prototype wth the realtme estmaton through a Dgtal Sgnal Processor (DSP) s constructed and presented. 1 Keywords Power onerters, Dagnoss, Electrolytc apactors, Equalent Seres esstance, apactance, Predcte Mantenance. lfe of a capactor fnshes when ts ES becomes at least two tmes hgher than ts ntal alue at the same temperature condtons [6]. Ths large araton makes that most of the technques proposed n the lterature are based only n the ES estmaton [7] Howeer, also the alue of the capactance s affected when the electrolytc capactor s degraded. Its capactance decreases wth the olume of the remanng electrolyte. As result of ths process, the typcal eolutons of capactance and ES ersus tme, are shown n Fgure 1. Accordng to ths, t s possble to take nto account the capactance estmaton, whch may hae a decrease of % wth respect to ts ntal alue [17]. Ths can be done wth the goal of renforcng and mprong the dagnostc conclusons n electrolytc capactors. 1 1 I. INTODUTION 1 (a) elablty and the consequent aalablty of power systems s closely related to the performance of the power conerters and/or the electrcal machnes noled n them. Early stage dagnostc tools are ery useful to aod serous drawbacks n the whole system by means of predcte mantenance. These tools lead to lower expenses caused by breakdowns, downtmes, redundant equpments and so on. Besdes the great nterest on dagnostc technques for electrcal machnes, n the last years an ncreasng nterest on dagnoss of power conerters has been notced as well. The greater nterest has been focused on systems nolng nerters, where t s usual to deal wth semconductors falures, also callng for the need of faulttolerant structures [1]. egardng D/D conerters, electrolytc capactors are the usual choce for smoothng ther output oltage. Ths s due to ther cost, sze and performance []. Howeer, these components appear as the most lfelmtng component, beng responsble for more than 5% of ther falures [3]. Therefore, sutable dagnostc technques are needed to preent the falure of electrolytc capactors, especally n crtcal hgh performance applcatons. The deteroraton of electrolytc capactors s manly caused by the electrolyte eaporaton as a result of temperature effects durng ther serce lfe. Ths eaporaton s reflected n some electrcal parameters [4, 5]. The most affected one s the ES, whch ncreases largely wth respect to ts ntal alue. It has been stated by manufacturers that the Manuscrpt receed 14/1/1; resed 15/3/11. Accepted for publcaton 11/4// 11 recommended for the specal secton by the edtor n charge Antono J. Marques ardoso. Eletrôn. Potên., ampo Grande,. 16, n., p , mar./ma t (h) (b) t (h) Fg. 1. Δ/ and ES characterstcs under hghtemperature load test: (a) ES ersus workng hours characterstcs; (b) Δ/ ersus workng hours characterstcs [7]. Thus, n ths paper a noel technque appled for Buck and Boost based topologes of D/D conerters s presented. The greatest adantages concernng preously publshed works are the estmaton of both and ES parameters, mprong the dagnostc conclusons, and also the low computaton and samplng efforts lnked to the adopted approach n real tme, not callng for the need of powerful mcroprocessors and consequently resultng n a lower cost system. II. THE EETOYTI APAITO MODE IN A BUK ONVETE In the context of D/D conerson, the Buck conerter, Fgure, s ery mportant and wdely used. Thus, the 177

2 deelopment of dagnostc tools may enhance ts hgh performance. It s well known that eery swtchng D/D conerter presents seeral dscrete states. For the Buck conerter, n the case of ontnuous onducton Mode (M), only the dscrete states of Fgures 3a and 3b occur. Howeer, n Dscontnuous onducton Mode (DM) also the dscrete state of Fgure 3c s present. V G 178 M urrent measurement DS A1 M G D 1 S On test Voltage measurement Fg.. D/D Buck conerter connected to a resste load and usng an deal capactor. On the other hand, n Fgure 4a an equalent crcut for the electrolytc capactor s shown. Due to ts physcal desgn and constructon, a capactor does not only has capactance,, but t also has a seres resstance ( S ), an nductance (ES) and a parallel resstance ( EAKAGE ). V G M a) = b) c) Fg. 3. D/D Buck onerter: (a) MOSFET conducton state, (b) non conducton state and (c) dscontnuous state. S ES ES EAKAGE a) b) c) Fg. 4. Electrolytc capactor equalent crcut: (a) SeresParallel confguraton (b) Seres confguraton; (c) Smplfed Seres confguraton where ES s neglected. V1 Howeer, ths model can be smplfed by onng both resstances as n Fgure 4b. Ths s the so called Equalent Seres esstance (ES). Ths model s wdely used, and t s a ery good approxmaton when dealng wth only one sngle frequency alue, snce the ES s frequency dependent. Although a rule of thumb for ealuatng the ES alue s proposed n [1], n Buck conerters the model of Fgure 4c can be used. Ths smplfcaton can be nferred f (1), whch descrbes the equalent model represented n Fgure 4b, s analyzed n the enronment of a Buck conerter. d (t) d (t) d (t) 1 ES ES (1) In (1), and represent, respectely, the current and oltage n the dece (real capactor). From the analyss of Fgure 3a, and 3b through Krchhoff s law, t s possble to conclude that the current through the capactor s almost lnear. In fact, t s the result of the lnear nductor current less the quasconstant load current. Thus, ts second derate can be neglected n (1). In ths way, () s obtaned. d (t) d (t) 1 ES () In DM, when the nductor current s null, Fgure 3c, the current through the capactor s the same current flowng through the load, and so, t s approxmately constant. onsequently, ts frst and second derates can be neglected n (1). Thus, (3) s obtaned. d (t) 1 (3) In fact, the only effect of ES on the capactor oltage s only notced n the boundares between the states of Fgure 3, when the derate of, and therefore of present a large change. If necessary, ES can be estmated by measurng the capactor oltage step n these boundares. In summary, t can be concluded that the ES has almost no effect n the Buck conerter output oltage waeform durng the dfferent states of Fgure 3, and t can be neglected n the capactor model as shown n Fgure 4c durng the ust mentoned states. In order to estmate the passe components n Buck conerters, the nductor current and the capactor oltage wll be sampled smultaneously throughout the swtchng perod, Fgure. egardng (), samplng the capactor current would be smpler for the method. Howeer, the nductor current s usually sampled for control purposes and a current sensor n seres wth the flterng capactor s normally a strong layout problem, snce n large power applcatons the capactor s drectly screwed n the busbar. The sampled nductor current, whose shape s known, wll determne the begnnng and the end of eery perod. Wth the data of eery perod, aerage alues for nductor current and the output oltage are computed. Ths way, (4) can be used to estmate the resste load. (4) Once s known, capactor current can be computed as n (5), aodng ts samplng. Eletrôn. Potên., ampo Grande,. 16, n., p , mar./ma. 11.

3 t t t t (5) III. A NOVE TEHNIQUE FO ONINE ESTIMATION OF THE APAITANE AND THE ES For estmatng the capactor parameters, the startng pont s (), whch can be appled f the current and the oltage of the capactor are known. The capactance and the ES must be calculated smultaneously accordng the followng steps: 1. The nductor current and the capactor oltage are sampled smultaneously durng seeral swtchng perods, obtanng and (t). Analyzng these data t s possble to dstngush where the nductor current s ncreasng (Fgure 3a), decreasng (Fgure 3b) or beng null (Fgure 3c). Once the nductor current data hae been dded n dfferent sets (data related to states of Fgure 3a, Fg 3b or Fgure 3c.), only the data from two swtchng perods are selected for further use.. It s well known that the nductor current s lnear durng all the three states. Therefore, t s possble to approxmate each stretch of current by the equaton of a straght lne. On the other hand, to approxmate the capactor oltage accurately, a second degree polynomal s used for eery state of the conerter. These approxmatons are done by means of the east Mean Squares (MS) algorthm [11]. In all cases, to calculate the polynomal coeffcents, the data sampled near the begnnng and the end of each stretch are not used before applyng MS n order to elmnate swtchng nose. 3. Once the current of two perods has been approxmated by lnes, t s easy to calculate the ntersecton of these lnes. Thus, the maxmum and mnmum alue of the current and the ntal and fnal nstant of one perod are obtaned. At ths moment, the data from the sampled seres are substtuted by calculated data, and. These new sets of data are free from swtchng nose and samplng error. 4. In (6) the load s estmated accordng to (4), where n s the number of sampled and recalculated data per perod: 1 n1 n1 n n1 n1 1 n (6) 5. Then, regardng (5), the alue of the capactor current for each samplng nstant s obtaned from (7) and a polynomal functon for (t) can be obtaned for eery state of the conerter. (7) 6. One of the greatest adantages of the proposed method s that the derates of the waeforms are obtaned wth a surprsng smplcty. Once the coeffcents of the ftted polynomal functons are aalable, the derates are drectly obtaned through ery smple deraton. alculatng Eletrôn. Potên., ampo Grande,. 16, n., p , mar./ma. 11. the derates from smple polynomal functons s ery useful from the pont of ew of mplementng the onlne technque wth smple hardware. It allows workng wth low number of samples per perod, reducng the requred samplng frequency and mantanng the accuracy of the results. Ths d d way,, can be obtaned for one perod. 7. Once an expresson for the derates s also aalable, ES and are the only not known parameters of (). They can be calculated by seeral means. If MS algorthm s used, (8) and (9) are encountered. Both of them are easy to compute, contrarly to other methods nolng more complex calculatons [1619]. n1 n1 n1 n1 d d ES n1 n1 n1 n1 d d (8) ES n1 d n1 1 n 1 n n1 n1 d n1 8. It s also possble to obtan an estmaton for the nductance of the buck conerter throughout the use of (1). In the state of Fgure 3b the current of the nductor s measured. If the oltage of the dode s measured the oltage of the nductor would be also known. Knowng the current and oltage of an nductor, ts nductance can be also calculated usng MS. n1 d n1 d dode (9) (1) The same behaor of the buck conerter s obsered n other D/D topologes such as the forward, pushpull, half brdge and full brdge conerters. Therefore, the here ntroduced approach can also be appled to them. On the other hand, the boost conerter operates dfferently and deseres ts own mathematcal approach as t wll be descrbed n the next secton. IV. THE BOOST ONVETE In the context of D/D conerson, the Boost conerter shown n Fgure 5 s also ery mportant and wdely used. For nstance, t s possble to fnd t n hgh performance applcatons such as photooltac systems [], or A/D conerters wth Power Factor orrecton (PF) [3]. Thus, t 179

4 s ery mportant the deelopment of dagnostc tools n order to keep ts hgh performance. For the Boost conerter, dependng on the operaton mode two (M) or three (DM) dfferent dscrete states are obsered. In the case of M, only the dscrete states of Fgs. 6a and 6b occur. Howeer, n DM also the dscrete state of Fgure 6c s present. V G 18 urrent measurement A1 G S M DS D 1 D On test Voltage measurement Fg. 5. D/D Boost conerter connected to a resste load and usng and deal capactor. V G V G A1 DS = A1 D= ES On test b) a) On test D = V1 ES c) Fg. 6. Boost onerter: (a) conducton state, (b) non conducton state and (c) dscontnuous state. onsderng the conducton state of Fgure 6a, the followng equatons can be wrtten: d (t) VG (11) d (t) 1 (1) ES Durng the non conducton state (Fgure 6b), the equatons are now as follows: d (t) ( VG ) (13) d (t) VG ES ( t) ES ES 1 ES ES ES (14) Fnally, f the conerter operates n dscontnuous mode, Fgure 6c, (1) s agan ald. In Fgure 6 the capactor has been modeled by ts capactance n seres wth ts ES, as shown n Fgure 4c. Ths smplfcaton can be nferred agan because the current through the capactor can be consdered lnear, beng null ts V1 V1 V1 second derate n (1), as n the case of Buck conerters. In fact, durng the conducton and the dscontnuous states the current through the capactor s the same that flows through the load, beng also approxmately constant. onsequently, ts frst and second derates can be neglected n (1), whch can be smplfed to (15). The same concluson s true for the dscontnuous state, Fgure 6c. d (t) 1 (15) Then, from the analyss of Fgure 6b through Krchhoff s law, t s possble to conclude that durng the nonconducton state, the current through the capactor s almost lnear. In fact, t s the result of the lnear nductor current less the quasconstant load current. In ths way, (16) s obtaned. d (t) d (t) 1 ES (16) omng back to (15), t can be deduced that the effect of ES durng the conducton perod can be defntely neglected n the capactor oltage derate. Ths result s rather logc because the load s hdng the effect of the ES, snce they are connected n seres and the later s small enough. Therefore, dfferently from the Buck conerter, wth the Boost topology t s possble to ealuate the capactor parameters separately, each one n a dfferent dscrete state. It means that the capactance can be ealuated durng the conducton state of the swtch, when the ES effect s hdden by the load, and afterwards, knowng, the ES can be estmated durng the nonconducton state of the swtch. Agan, as done for the Buck conerters: 1. the nductor current and the capactor oltage are sampled: and.. Now, from the shape of the current t s possble to dentfy the acte swtch gate sgnal by a ery smple analyss, aodng any extra sensor. 3. The sampled current and oltage sgnals are ftted by a polynomal functon of sutable order: second order for the oltage n Fgure 6b condtons and frst order otherwse. 4. The sampled data are substtuted by ther ftted polynomal functons, 5. The capactor current s obtaned. Wth ths purpose, durng the conducton state and the dscontnuous state (17) s used, and for the nonconducton state (18) s used nstead. In both cases represents the output current. (17) ( t) ( t) ( t). (18) In order to obtan an estmaton of and to compute (17) and (18), (19) and () must be preously used. In fact, t s well known that t s possble to obtan for each dfferent operatng mode (DM or M), an aerage state model where the dfferent dscrete states are combned n one sngle system of dfferental equatons representng the aerage behaor of the consdered conerter. From these models t s possble to obtan, under steady state condton and consderng a resste load, some relatonshps between the statearables aerage alues. For the case of the Boost Eletrôn. Potên., ampo Grande,. 16, n., p , mar./ma. 11.

5 conerter the followng relatonshps are respectely found, for M and DM operatng modes: (19) D D D off ( on off ) () Doff where D on, called duty cycle, s lnked to the perod when the nductor s beng charged (Fgure 6a) and D off s lnked to the perod when the nductor s beng dscharged (Fgure 6b). In DM, D off lasts untl the moment when there s no more current flowng through the nductor. 6. Once the approprate current and oltage derates are computed, MS algorthm s used to obtan an estmaton of from (15). Ths estmaton s used n (16) to compute ES always through MS. Ths analyss can be straghtforward extended to Buck Boost and Flyback topologes. Also D lnk capactors can beneft from t, f the capactor s dscharged through a rheostat when swtchng off the conerter [4]. In ths stuaton ts capactance can be ealuated from (15). Moreoer, as the capactance s ealuated separately from ES ths method can be extended to flm capactors, where ES s extremely low and dffcult to measure een wth accurate nstruments. V. EATIME IMPEMENTATION OF THE DIAGNOSTI TEHNIQUE USING A DSP Ths method s smple enough to be mplemented n a low cost system based on a DSP. The current s acqured wth a Hall Effect sensor and ts leel s adapted wth one operatonal amplfer (OA) to the range of the DSP analogcal nputs. The output oltage s treated as well by means of OAs to extract the rpple and amplfy t to sutable leels. In both cases, the bandwh of the OAs seres as low pass flter. The samplng of the waeforms s done by means of the nternal A/D conerter of a dspi from Mcrochp (dspi33f famly). Ths dece was selected due to ts good compromse between performance and cost. Two nternal sample and hold acqure both waeforms smultaneously, beng conerted to dgtal sequentally. For ths confguraton, the maxmum samplng frequency per channel s 5 khz. Wth ths lmt, the whole system has been tested up to a swtchng frequency of khz, meanng 5 samples per perod, always prodng a good performance. As the nductor alue s not needed for dagnostc purposes n electrolytc capactors, n Buck conerters the dode oltage s not sampled, but reasonably approxmated by a constant. The alue obtaned wth (1) s n ths case only an approxmaton. Once the sgnals hae been acqured the capactance and the ES of the component are ealuated by the algorthm descrbed before. The process can be repeated almost nstantaneously for dfferent swtchng cycles, hang seeral measurements of and ES per second. The D dsplay of the deelopment system has been used to show the mong aerage of eery 64 measurements. For an ndustral product the dspi deelopment board s not necessary, but only the dspi tself, dramatcally reducng the sze of Fgure 7. In a fnal assemble, the resultng nstrumentaton, four OAs and a DSP, does not sgnfcantly affect the cost and the sze of a large power conerter. Fg. 7. Expermental prototype. Detal of the nstrumentaton and the DSP deelopment board wth measured data on the screen. The nstrumentaton and the DSP tself can be ntegrated n the power stage as small sze addtonal hardware. VI. EXPEIMENTA ESUTS FO A BUK ONVETE Tables IIII show some expermental results from the embedded system of Fgure 7 for two dfferent capactors and loads n a Buck conerter. The results obtaned are ery stable and repette n good agreement wth measurements performed by a hgh accuracy mpedance analyzer (model Aglent 494A), also gen n the tables. As can be obsered, they reproduce well the temperature nfluence n the capactor parameters. In fact, lower duty cycle mples hgher current rpple n the capactor and therefore an ncrement of ts temperature. Ths fact leads to a better conductty of the electrolyte and hence lower ES and larger capactance. TABE I Estmated Parameters Usng a DSPc apactor 1. Nomnal apacty µf 63V. load=.33ω Impedance analyzer estmaton. ES=9.mΩ, =19. µf Duty cycle load ES (mω) (µf) TABE II Estmated Parameters Usng a DSPc apactor. Nomnal apacty µf 35V. load=.33ω Impedance analyzer estmaton. ES=111.4mΩ, =188 µf Duty cycle load ES (mω) (µf) TABE III Estmated Parameters Usng a DSPc apactor. Nomnal apacty µf 35V. load=4.73ω Impedance analyzer estmaton. ES=111.4mΩ, =188 µf Duty cycle load ES (mω) (µf) Eletrôn. Potên., ampo Grande,. 16, n., p , mar./ma

6 Tables II and III hae been conceed for the same capactor under dfferent loads. The measured current n the nductor s dfferent. Howeer, the current n the capactor should be the same n both cases f the duty cycle s the same. So, ES and should match for the same duty cycle. The alues of both tables are qute smlar (error < 1.5%) and n ery good agreement wth the mpedance analyzer estmaton. As shown n Fgure 1, at the capactor end of lfe, a decrement of at least % n the capactance alue s expected. At the same tme the ES of the capactor should ncrease to een ten tmes ts ntal alue. egardng the accuracy of these expermental results t s possble to affrm that ths end of lfe could be foreseen n adance, and the replacement of the component can be alerted. In Fgure 8, t can be obsered the dfference among 3 dfferent measurements of both parameters, and ES, as proded by the embedded system. Eery sngle measurement and the mong aerage of the last 64 measurements are represented for each parameter. The aerage s consdered as the fnal estmaton gen by the system, and t s an almost constant and stable alue. Ths relablty s possble because the dstrbuton of the sngle measurements s ery narrow around the fnal alue, as shown n Fgure 9 for other measurements set. For all the measurements done, the deaton s so small that more than 9% of the real tme estmatons always dffer of less than 3% from the fnal aerage alue. apacty (µf) ES (mω) 8, 6, 4,,, 198, 196, 194, 19, 17, 16,5 16, 15,5 15, 14,5 14, 13,5 13, Inddual measurement Aerage of the last 64 measurements Number of the measurement (a) Inddual measurement Aerage of the last 64 measurements 1, Number of the measurement (b) Fg. 8. (a) apactance and (b) ES real tme estmatons for the condtons at Table II when the duty cycle n the conerter s.4. The hgher dfference between any measurement and the mong aerage s below 3%. VII. EXPEIMENTA ESUTS FO A BOOST ONVETE The method deduced for boost conerters has also been expermentally tested. The dagnostc tool has been mplemented n a DSP embedded system. Agan the results obtaned are ery stable and repette. They reproduce well the temperature changes n the capactor parameters, and are n good agreement wth the mpedance analyzer results as seen n Tables IV and V. The measurements wth the mpedance analyzer are done wth no losses n the capactor, meanng lower due to temperature effects μf s obtaned n for the 4% of the measurements Aerage=198.1μF 114.5mΩ s obtaned n the 7% of the measurements Aerage=114.4mΩ for 1.5% apactance (μf) (a) 1% >9% of the measurements are nsde ths area 9% of the measurements are nsde ths area ES (mω) (b) Fg. 9. (a) apactance and (b) ES real tme estmatons for the condtons at Table III when the duty cycle n the conerter s.6. More than 9% of the nddual measurements dffer less than 1.5% from the aerage alue of them all. The dfference of eery sngle measurement from the fnal (aerage) alue s always ery small: n more than 9% of the measurements, the dfference between nddual real tme measurements and fnal aerage alue s lower than %. A typcal dstrbuton of measurements s shown n Fgure 1. TABE IV Estmated Parameters Usng a DSPc apactor 1. Nomnal apacty 15µF 35V Impedance analyzer estmaton. ES=119.9mΩ, =137.6 µf, ES=1.9nH Duty cycle ES (mω) (µf) Eletrôn. Potên., ampo Grande,. 16, n., p , mar./ma. 11.

7 TABE V Estmated Parameters Usng a DSPc apactor. Nomnal apacty 47µF 5V Impedance analyzer estmaton. ES=9mΩ, =383.9 µf, ES=31.5nH Duty cycle ES (mω) (µf) Aerage=15.5μF ~1% of the measurements are nsde ths area.8% μf s obtaned n for the 51% of the measurements Aerage=14μF 1.7% apactance (μf) (a) mΩ s obtaned n for the 3% of the measurements Aerage=11.1mΩ 1.8% ~1% of the measurements are nsde ths area ~1% of the measurements are nsde ths area ES (mω) (b) Fg. 1. (a) apactance and (b) ES dstrbuton of the real estmatons for the 15 µf capactor (D on =.4). The same method has been also used to measure flm capactors. The dece under test has a nomnal capactance alue of 15 µf, and the alue of ts ES proded by the manufacturer s 65 µω. When measured wth an mpedance analyzer Aglent 494A the alue obtaned for ts capactance s 14.4 F. Howeer, t was mpossble to measure accurately such a low ES. Values below 5 mω hae been obtaned. The capactance and ES measurements wth the embedded system when workng as flterng capactor of a boost conerter are shown n Fgure 11. The capactance measurement has been also ery repette and stable, and the error s around 1% (Fgure 11a). On the other hand the results for the ES are qute smlar to the ones obtaned wth an mpedance analyzer apactance (μf) (a) ~7% of the measurements are nsde ths area ES s below 1.5mΩ ES (mω) (b) Fg. 11. (a) apactance and (b) ES dstrbuton of the real estmatons for a flm capactor (15µF). VIII. ONUSIONS In ths paper, a noel technque for montorng the condton of electrolytc capactors n power conerters was presented. Ths proposed technque presents seeral adantages regardng other alternates: 1. Both parameters used n the dagnoss of electrolytc capactors, ES and capactance, are estmated onlne, and een n real tme, whle the component s workng n the power stage. Snce the two parameters of the capactor are ealuated, the relablty of the predcte mantenance system s enhanced, what s ery nterestng n hgh performance applcatons.. For applyng t, t s only necessary to know the nductor current and the capactor oltage, beng these two waeforms, n most of the cases, already sensed for control purposes. The technque does not use a current sensor n seres wth the capactor. Such a current sensor would need space,.e. addtonal wrng from the capactor to the bus bar. As result, a parastc nductance would be added n seres wth the capactance due to the current sensor tself and the addtonal wrng. In a hgh power applcaton, mnmzng nductances n the power stage s a maor concern. 3. Due to the smplcty of the mathematcal treatment the hardware requred to run the algorthm s smple, nexpense and compact. In ths paper a low cost DSP s proposed to mplement the software and obtan both ES and capactance. 4. The only requrement for the hardware would be to sample the current and oltage at least tmes per swtchng Eletrôn. Potên., ampo Grande,. 16, n., p , mar./ma

8 perod. Wth a swtchng frequency of khz a samplng frequency of 5 khz per channel was used. The relablty and the accuracy of the resultng embedded system are proed to be excellent. Although n ths paper specal emphass was gen to D/D conerters, the proposed technque can be extended to any other confguraton and also to D lnk capactors. It s especally ndcated for those cases where t s not possble to place a current sensor n seres wth the capactor. All these propertes make the proposed technque a good soluton to perform electrolytc or Metallzed PolyPropylene Flm (MPPF) capactor predcte mantenance n hgh power conerters. 184 NOMENATUE apactance of the capactor. ES Equalent Seres esstance of the capactor. ES Equalent Seres Inductance of the capactor. oad resstance. apactor current. apactor oltage. Inductor current. oad current. Aerage alue of the capactor oltage. Aerage alue of the nductance current. th sampled alue of the nductor current. th sampled alue of the capactor oltage. Estmated th alue for the nductor current when ftted by a polynomal approach. Estmated th alue for the capactor oltage when ftted by a polynomal approach. Estmated th alue of the capactor current when ftted by a polynomal approach. d Estmated th alue of the derate of the capactor current when ftted by a polynomal approach. d Estmated th alue of the derate of the capactor oltage when ftted by a polynomal approach. d Estmated th alue of the derate of the nductor current when ftted by a polynomal approach. th sampled alue of the dode oltage. dode D ON Duty cycle. Fracton of the perod wth the transstor swtched on. D off Fracton of the perod wth the transstor swtched off and current n the nductor. n Number of tmes that the nductor current and the capactor oltage are sampled per perod. EFEENES [1] A. M. S Mendes, A. J. Marques ardoso, Fault Tolerant Operatng Strateges Appled to ThreePhase InductonMotor Dres, IEEE Transactons on Industral Electroncs, ol. 53, no. 6, pp , December 6. [] J. Steens, J. Shaffer, J. Vandenham, The Serce fe of arge Alumnum Electrolytc apactors: Effects of onstructon and Applcaton, IEEE Transactons on Industry Applcatons, Vol. 38, No. 5, pp , September/October. [3] Mltary Handbook 17 F, elablty predcton of Electronc Equpment, eson F, December 1991, Notce 1, 1 July 199, Notce, 8 February [4]. J. Hart, D. Scoggn, Predctng Electrolytc apactor fetme, Powertechncs Magazne, pp. 4 9, Anahem, A, Oct [5] J. A. auber, Alumnum Electrolytc apactors elablty, Expected fe and Shelf apablty, Sprague Techncal Paper TP839, pp. 4, [6] K. Harada, A. Katsuk, M. Fuwara, Use of ES for Deteroraton Dagnoss of Electrolytc apactor, IEEE Transacton on Power Electroncs, ol. 8, no. 4, pp , October [7] M.. Gasper, fe Predcton Modelng for Alumnum Electrolytc apactors, IEEE 31 st Industry Applcatons onference (IAS 96), ol. 4, pp [8] M.. Gasper, fe Predcton Modelng of Bus apactors n A VarableFrequency Dres, IEEE Transactons on Industry Applcatons, ol. 41, no. 6, pp , Noember/December 5. [9] V. A. Sankaran, F.. ees,. S. Aant, Electrolytc apactor fe Testng and Predcton, IEEE 3 nd Industry Applcatons Socety onference, (IAS 1997), San Dego, A, Vol., pp [1] G. M. Buatt, A. M.. Amaral, A. J. M. ardoso, Parameter Estmaton of a D/D Buck onerter Usng a ontnuous Tme Model, n Proceedngs of EPE 7, Aalborg, Denmark, 5 September 7. [11] G. M. Buatt, A. M.. Amaral, A. J. M. ardoso, An Onlne Technque for Estmatng the Parameters of Passe omponents n NonIsolated D/D onerters, IEEE Internatonal Symposum on Industral Electroncs, Vgo (Span), pp. 6661, 47 June 7. [1] A. ahyan, P. Venet, G. Grellet, P. Verge, Falure Predcton of Electrolytc apactors Durng Operaton of a Swtch Mode Power Supply, IEEE Transacton on Power Electroncs, ol. 13, no. 6, pp , Noember [13] G. E. hoades, A. W. H. Smth, Expected fe of apactors Wth NonSold Electrolyte, n Proceedngs of 34 th IEEE omponent onference, 1984, New Orleans, A, USA, pp [14] P. Venet, F. Persse, M. Hussen, G. oat, ealzaton of Smart Electrolytc apactor rcut, IEEE Industry Applcatons Magazne, nº 1, pp. 16, January/February. [15] E. Aeloza, J. H. Km, P. umnot, P. N. Enet, A ealtme Method to Estmate Electrolytc apactor ondton n PWM Adustable Speed Dres and Unnterruptble Power Supples, n Proceedngs of the IEEE Power Electroncs Specalst onference, pp , 116 June 5, ecfe (Brazl). Eletrôn. Potên., ampo Grande,. 16, n., p , mar./ma. 11.

9 [16] A. M. Imam, T.G. Hableter,.G. Harley, D.M. Dan, ondton Montorng of Electrolytc apactor n Power Electroncs rcuts usng Adapte Flter Modellng, Proceedngs of the IEEE Power Electroncs Specalsts onference, pp. 6167, ecfe (Brazl), 1 16 June 5. [17] H. Ma, X. Mao, N. Zhang, D. Xu, Parameter Identfcaton of Power Electroncs rcuts Based on Hybrd Models, n Proceedngs of the IEEE Power Electroncs Specalsts onference, pp , 1 16 June 5, ecfe (Brazl). [18] H. Ma,. Wang, Fault Dagnoss and Falure Predcton of Alumnum Electrolytc apactors n Power Electroncs onerters, 3 nd Annual onference of IEEE Industral Electroncs Socety, pp , IEON 5, 61 No. 5. [19] G. M. Buatt, A. M.. Amaral, A. J. M. ardoso, ES Estmaton through Smplfed egresson Models, IEEE Industry Applcatons Socety 4nd Annual Meetng, pp. 8994, New Orleans (USA), 37 September 7. [] A. Imam, D. Dan,. Harley, T. Habetler, ealtme ondton Montorng of Electrolytc apactors by Parameter Estmaton IEEE Appled Power Electroncs onference and Exposton, pp , 68 Feb and 1 March 7, Anahem (USA). [1] S. G. Parler, Improed Spce Model of Alumnum Electrolytc apactors for Inerter Applcatons, IEEE Transactons on Industry Applcatons, ol. 39, no 4, pp , July 3. [] E. oman,. Alonso, P. Ibanez, S. Elorduzapataretxe, D. Gota, Intellgent PV Module for Grdonnected PV Systems, IEEE Transactons on Industral Electroncs, ol. 53, no. 4, pp , June 6. [3] Koen De Gusseme, Dad M. Van de Sype, Alex P. M. Van den Bossche, Jan A. Melkebeek, "Inputurrent Dstorton of M Boost PF onerters Operated n DM," IEEE Transactons on Industral Electroncs, ol. 54, no., pp , Aprl 7. [4] G.M. Buatt, J. A. Martnamos, A. M.. Amaral, P. Dworakowsk, A. J. M. ardoso ondton Montorng of Metallzed Polyproplene Flm apactors n alway Power Trans, IEEE Transactons on Instrumentaton and Measurement, ol. 58, no.1, pp , October 9. BIOGAPHIES Gustao M. Buatt was born n Uberlânda, Brazl, n He receed the B.Sc. degree n electrcal engneerng from the Federal Unersty of Uberlânda n, and the Ph.D. degree from the Poltecnco d Torno, Turn, Italy, n 6. In 6, he was wth Internatonal ectfer, Borgaro, Italy, and from 6 to 8 he was wth Alstom Transport, Séméac, France. He s currently wth Mtsubsh Electrc, ennes, France, where he s Proect eader of research acttes on Sustanable Energy Systems. Hs research nterests are focused on modelng of power semconductor deces, desgn of D/D power conerters and condton montorng of passe components used n power applcatons. Juan A. Martínamos was born n Oedo, Span, n He receed the M.Sc. and Ph.D. degrees n Industral Engneerng from the Unersdad de Oedo (Span) n 1996 and 1 respectely. In 1997, he oned the Electronc Technology Area of the Unersdad de Oedo where he s currently an Assstant Professor. Durng ths perod, he has been noled n seeral ndustral proects regardng mcroprocessor based systems and power electroncs. Hs research actty has been manly related to the AD power conerson where he has worked n the ntegraton of magnetc power deces, the ncluson of pezoelectrc transformers and the concepton of new topologes and transformers for hgh oltage applcatons. Juan A. Martínez was born n Gón, Span, n 196. He receed the M.Sc. degree and the Ph.D. degree n electrcal engneerng, from the Unersty of Oedo, (Span) n 1987 and 1991 respectely. In 1987, he oned the Unersty of Oedo where he s currently Head of the Electrcal Engneerng Department. He has been noled n more than 5 academandustry research proects and s the author of seeral techncal papers and patents. Hs man research nterests nclude swtchmode power supples and mcrocontrollerbased ndustral systems. Acáco M.. Amaral was born n uso, Angola, n He receed the centate, M.S. and PhD degrees n electrcal engneerng from the Unersty of ombra, ombra, Portugal, n 1998, 5 and 11 respectely. Snce 1998, he has been wth the Polytechnc Insttute of ombra, where he s currently an Assstant eader n the Department of Informatcs and Systems. He s also n the Department of Electrcal and omputer Engneerng, Unersty of ombra. He s a esearcher wth the Portuguese Telecommuncaton Insttute. Hs research acttes nclude fault dagnoss and desgn of swtchmode power supples, wth emphass on the consequences of agng of electrolytc capactors, as well as the deelopment of solutons to ths problem. Hs teachng nterests coer dgtal systems, programmng, sgnal processng, nstrumentaton, and electroncs. He has authored more than papers publshed n conference proceedngs. Prof. Amaral s a Student Member of the IEEE Instrumentaton and Measurement Socety, the IEEE Industral Electroncs Socety, the IEEE Industry Applcatons Socety, and the IEEE Aerospace and Electronc Systems Socety. He has been lsted n Who s Who n the World. A. J. Marques ardoso was born n ombra, Portugal, n 196. He receed the E.E. and Dr.Eng. degrees from the Unersty of ombra, ombra, n 1985 and 1995, respectely. Snce 1985, he has been wth the Unersty of ombra, where he s currently an Assocate Professor wth the Department of Electrcal and omputer Engneerng, Insttuto de Telecomuncações, and the Drector of the Electrcal Machnes aboratory. He s currently a member of the Edtoral Board of the Internatonal Journal of ondton Montorng and Dagnostc Engneerng Management. He s Eletrôn. Potên., ampo Grande,. 16, n., p , mar./ma

10 the author of a book enttled Fault Dagnoss n ThreePhase Inducton Motors (ombra, Portugal: ombra Edtora, 1991) and more than 5 publshed papers. Hs teachng nterests coer electrcal rotatng machnes, transformers, and mantenance of electromechatronc systems. Hs research nterests are condton montorng and dagnostcs of electrcal machnes and dres. Dr. ardoso s a member of the New York Academy of Scences, the European Power Electroncs and Dres Assocaton, the Electrcal Machnes and the Industral Dres ommttees of the IEEE Industry Applcatons Socety, the Electrcal Machnes and the Power Electroncs ommttees of the IEEE Industral Electroncs Socety, the Techncal ommttee on Dagnostcs of the IEEE Power Electroncs Socety, and the Portuguese Federaton of Industral Mantenance. He s also a Senor Member of the Portuguese Engneers Assocaton. He was a member of the Oerseas Adsory Panel of.m.d. Technology. He s currently an Honorary Member of the Internatonal Bographcal entre Adsory ouncl, ambrdge, U.K., and an Honorary Professor of the Albert Schwetzer Internatonal Unersty, Genea, Swtzerland. He s actely noled n the feld of standardzaton on condton montorng and dagnostcs (.M.D.), where he has been actng as a onenor of the Internatonal Organzaton for Standardzaton (ISO)/T 18/S 5 Adsory Group D (.M.D. of power transformers) and the ISO/T 18/S 5 Workng Group 1 (.M.D. of electrcal equpment), and s also a member of seeral workng groups/ballotng commttees of the ISO, the IEEE, and the omté Européen de Normalsaton. He has been lsted n Who s Who n the World, Who s Who n Scence and Engneerng and BEST Europe, among others. 186 Eletrôn. Potên., ampo Grande,. 16, n., p , mar./ma. 11.