Internatonal Journal of Power Electroncs and Drve System (IJPEDS) Vol. 3, No. 2, June 2013, pp. 209~217 ISSN: 2088-8694 209 A Shunt Actve Power Flter wth Enhanced Dynamc Performance usng Dual-Repettve Controller and Predctve Compensaton Zhenfeng Xao, Ylong Chen, Xangtan Deng School of Power and Mechancal Engneerng, Wuhan Unversty e-mal: coolpatrce@qq.com, sandywac@sohu.com, mandxt@126.com Artcle Info Artcle hstory: Receved Jan 6, 2013 Revsed Apr 20, 2013 Accepted May 14, 2013 Keyword: Shunt Actve Power Flter Mult-Repettve Errors Total Harmonc Dstorton Dual-Repettve Controller Predctve Compensaton ABSTRACT In ths paper, the confguraton characterstc of shunt actve power flter (APF) wth splt capactor s analyzed, as well as ts prncple dagram and control module. In order to mprove the dynamc performance of a control system and to elmnate mult-repettve errors (MRE), a combnaton strategy based on dual-repettve controller (DRC) and PI controller s presented. One repettve controller s for ensurng the current trackng accuracy and the other one s for enhancng dynamc response. And for purpose of elmnatng the system delay brought by the nverter and specal control, an mproved predctve compensaton method s proposed by usng the pre-compensated angel. Usng ths composte control strategy to carry on ndustral prototype smulaton and feld test, the expermental result shows that system compensaton could effectvely reduce the total harmonc dstorton (THD) values from 26.02%, 26.94% and 26.27% to 4.20%, 4.59% and 4.35% for each phase of the current. And the full response tme are all less than 10ms, fully meetng the standard of IEEE-519. Copyrght 2013 Insttute of Advanced Engneerng and Scence. All rghts reserved. Correspondng Author: Zhenfeng Xao, School of Power and Mechancal Engneerng, Wuhan Unversty Luo-ja-shan, Wuchang, Wuhan, Hube Provnce, P.R.Chna 430072 E-mal: coolpatrce@qq.com 1. INTRODUCTION Wth the wdespread applcaton of power electronc technology for ndustral equpment, these not only provde a hgh speed, hgh effcency and energy savng methods of control, but also brng nto harmonc pollutons. Actve power flter (APF) s a knd of stable, hgh effcent, flexble optmzaton power qualty machne, whch plays an mportant role n mprovng the power qualty. APF s one of most mportant harmonc suppresson and reactve Power compensaton of Power electronc devces. At present, the common harmonc detecton methods are mostly based on the nstantaneous power theory [1], and the core of the detecton method s to subtract the fundamental current from the load current, whch ams to get all the harmonc current for compensaton. Ths tradtonal detecton method has been wdely used. But consderng the nherent delay of current control, voltage control and PI control etc., the tradtonal control method s nsuffcent, whose compensaton effects s not so well and the harmonc current compensaton s only a part of the whole. And n the control method of APF, mostly concernng about the current control, the easy approach s to use hysteress control method. But the dynamc response speed, swtchng frequency and current trackng accuracy wll be nfluenced by hysteress bandwdth, whch wll cause large current rpples and swtchng noses. Whle the tradtonal P control and PI control based on PWM technology s wdely appled to the APF system, ts closed-loop gan system s restraned by the stablty condtons, whch wll Journal homepage: http://aesjournal.com/onlne/ndex.php/ijpeds
210 ISSN: 2088-8694 lead to nadequate compensaton for the man harmonc compensaton and may not acheve a better harmonc compensaton effect [2]. Usng repettve control method s an effectve way of APF control, whch s mostly based on SRC (sngle repettve control). Such as Ref. [3]-[4] by usng ths method, the harmonc currents can be suppressed well. However, SRC has a response tme of one repettve perod delay [5]. The longer repettve perod, the slower dynamc response, but wder compensated bandwdth. And Ref. [6] mentoned that MRE (multrepettve errors), caused by SRC, wll affect the speed of harmonc compensaton performance and compensaton speed. Thus, usng the SRC to compensate, t s not easy to acheve the good harmonc compensaton performance and the faster response speed. On the pont of these, the paper, wth an eye to mprove the real-tme, accuracy and dynamc response speed of harmonc compensaton, proposes a predctve compensaton method and employs the "combnaton strategy of PI control and DRC (dual repettve control)" of the APF, t can not only acheve a complete soluton for system nherent delay, but also can detect the harmonc current n the real-tme wth an accurate ablty, and at the same tme has a favorable compensaton effect and a fast dynamc response speed. Ths control strategy has been successfully appled n a 100kVA actve power flter. Its valdty s verfed through the specfc smulaton and the feld test wth the harmonc resources. 2. Overall Desgn The overall structure of the shunt three-phase four-wre APF, as shown n Fgure 1(a), drectly connect the AC neutral lne wth the neutral pont, ths dvson capactor type nverter topology has better controllablty. Among them, s s for power supply current, L s for load current, c s for compensatng current, and U dc s for the DC-sde voltage. APF control functon module s shown n Fgure 1(b), whch adopts the random harmonc current detecton of predctve compensaton control, combnaton of PI control and DRC, and closed-loop voltage control of DC-sde. VSI L1 L2 L 3 Ln U dc S1 S 2 n S 3 Sn n Fgure 1(a). Prncple dagram of shunt APF dc U dc IJPEDS Vol. 3, No. 2, June 2013 : 209 217 Fgure 1(b). Control module dagram of Shunt APF
IJPEDS ISSN: 2088-8694 211 The harmonc detecton part can drectly detect the specfed order harmonc and make a specfc order compensaton and full compensaton. The combnaton part enhances the track precson of both even and odd harmoncs. Closed-loop voltage control of DC-sde ensures the DC-sde capactor voltage n the specfed voltage range [7]. 2.1. Current Detecton based on Predctve compensaton APF, wth predctve control, adopts the selectve harmonc detecton method based on predctve compensaton method and current/voltage closed-loop control based on ths strategy. Predctve compensaton of harmonc detecton can detect any specfed harmoncs. The theoretcal bass of ts detecton method s consstent wth the tradtonal method. That s the frequency of fundamental current and each harmonc bascally remans unchanged. Prncple of predctve compensaton control method s shown n Fgure 2. U a PLL Sn /cos a b c C32 Cn dn qn dn qn n n 1 C n n C23 n an* bn* cn* Fgure 2. Predctve compensaton method As shown n Fgure 2, accordng to ths assumpton and Fourer seres, t adopts PLL (Phase Lock Loop, PLL) and dgtal functon generator to produce the sne/cosne sgnal wth the same phase of power supply voltage for elmnatng the dstorton effect caused by power supply voltage. It means takng one phase voltage through PLL and frequency doublng process to get sn nwt and cos nwt. Then t can get the transformaton matrx: C n sn( nwt) cos( nwt) cos( nwt) sn( nwt) (1) C 32 1 1/2 1/2 2/3 0 2 /2 3/2 (2) Takng the three-phase current through formula (2), C 32 wll transform nto two phase current components and. And takng these two varables through formula (1), t wll get the actve component dn and the reactve component qn. Then, takng these current components pass through the low pass flter, t wll get the DC components dn and qn. In the conventonal algorthm, t usually takes no account of system delay and full delay tme. However, through the above dervatons, t can assume the total delay tme as T and assume the angular frequency of fundamental as to get the turnng angle of system durng tme of T : n n T 2 n f T( rad) (3) That s to say, the harmonc current value durng the tme of n s the harmonc current value durng the tme of n n, whch causes the erroneous compensaton of the system [8]. To elmnate the system delay and lag caused by the error compensaton, the paper n the nverse transformaton matrx has modfed electrcal angle to change the compensaton tme. That s addng pre-compensaton angle n n the orgnal electrcal Angle to compensate system tme delay. In ths way, DC component, through the formula (4), nverts the matrx and C32 to get Nth-degree harmonc current ah* 1 C n n bh * ch*, and C T 32 C C 23, A Shunt Actve Power Flter wth Enhanced Dynamc performance usng Dual-Repettve Controller
212 ISSN: 2088-8694 1 sn( nwtn) cos( nwtn) C n n cos( nwt n) sn( nwt n) (4) It s necessary to menton that to get the full compensaton just need to use every harmonc parallel computng way respectvely and get every harmonc desgnated, then adds every harmonc together [9]. 2.2. Combnaton Strategy of PI control and DRC PI controller can mmedately adjust the trackng error and has a good ablty to solve the DC-errors. However the trackng control accuracy for elmnatng DC-errors s not well. Especally when hgher harmonc s put nto the current regulator, the PI controller s dffcult to solve the problem of harmonc. Although the repettve control has advantages of zero steady-state error trackng and low output dstorton, the dynamc response speed s slow. And the conventonal SRC method has always faled to reach a good harmonc compensaton performance and fast dynamc response speed. Therefore, ths paper adopts the combnaton strategy of PI control and DRC to acheve a better compensaton effect and a fast speed of dynamc response. PI controller could use the ntegral acton to elmnate the DC-statc error. The nternal N N model of repettve controller can be descrbed as[1 / (1 z )] z,whch can understand for the "ntegraton" and "delay" two parts, and N presents for the samplng number n one cycle. Among them, N Z k presents for delay lnk, K r Z S( z ) s for compensator, S( z) k s for hgh-frequency attenuaton, and Z s for phase lag of compensator. The output of repettve controller s based on the error sgnal of the prevous cycle for reducng the system error and ensurng the steady accuracy. The compound strategy, proposed by ths artcle, s puttng these two optmzaton repettve controllers nto the feedback loop n parallel. The overall control prncple of combnaton of PI control and DRC can be shown n Fgure 3. 1 dabc (,, ) * habc (,, ) K 11 1 QzZ ( ) N N Z 1 KZ k r Sz ( ) Pz ( ) f ( abc,, ) K 22 2 QzZ ( ) N N Z 2 KZ k r Sz ( ) Fgure 3. Implementaton of compound current controller From the above fgure, the error transfer functon for over-all closed-loops * h( a, b, c) 1 N1 N2 dabc (,, ) K11z K22z 1 [1 ( ) G ( )] ( ) N1 N2 C z P z 1 Q( z) z 1 Q( z) z (5) Where P z 1 ( ) PI( z) z G( z) (6) IJPEDS Vol. 3, No. 2, June 2013 : 209 217
IJPEDS ISSN: 2088-8694 2 N1 1T 2 N1 T 2 (7) (8) 213 The characterstc equaton (6) can be wrtten as [10]: / (,, ) 0 (,, ) d a b c (9) * habc From (9), t can be concluded that [6] Qz ( ) K 1 (10) 11 Qz ( ) K 1 22 (11) K K ([1/ Qz ( ) (1 K )] [1/ Qz ( ) (1 K )]) 2 ( 11 22) 2 2 2 2 11 22 1 (12) It can be shown that repettve control system can elmnate any order harmoncs, and system wll reach excellent control accuracy whenqz 1. Artcle puts forward that nsertng K 11 and K 22 n front of the compound controller. Adjustng dstrbuton of K11 and K 22, PI controller could determne the proporton of DRC to ensure the dynamc response. On the other sde, wth the guarantee of stablzaton, usng the parameters of K11 and K 22, Q Z could be closer to 1. Ths result enables the compound system acheve rapd and stable current control. The PI control prncple s shown n Fgure 4. From ths, the man crcut of nverter s equvalent to Tf s a delay lnk: K f e and ts system response s delay tme st f. In Fgure.4, L s the equvalent nductance of transmsson lne and R s the equvalent resstance. When swtchng frequency s relatvely hgh, t can just use PI controller to make regulator pole zero cancellaton for controllng nerta lnk, so t only need to control P parameter to make current loop realze zero steady-state of step, whch wll acheve a hgh accuracy of harmonc compensaton [11]. * h ( abc,, ) s 1 KPI s T f K s f e 1 LS R Fgure 4. Implementaton of PI controller Therefore, the compound current control strategy based on PI control and DRC could not only track system current fast, but also can realze real-tme detecton and accurate compensaton. 3. Smulaton Results To verfy feasblty of combnaton of PI control and DRC for shunt APF as Fgure 1(b), smulaton s carred out under MATLAB/Smulnk envronment. Parameters of man crcut are: A Shunt Actve Power Flter wth Enhanced Dynamc performance usng Dual-Repettve Controller
214 ISSN: 2088-8694 U dc Table 1. Smulaton Parameters swtchng frequency f pwm Cdc N QZ K11 K22 680V 10kHz 9600F 144 1 0.5 0.5 Analyzed from the spectrums of Fgure 5 and Fgure 6, before the compensaton, the RMS of sngphase load s 32.87A, THD s 24.70%. And ncludng them, 5th THD s 21.74%, 7th THD s 8.82% and 11th THD s 6.09%.When APF s put nto operaton, by usng the conventonal control method the THD s 6.03%, mprovng a lot. 5th THD s 2.16%, 7th THD s 1.33% and 11th THD s 1.70%. However, wth the compound control strategy, the THD s 2.42%, mprovng greatly. 5th THD s 0.82%, 7th THD s 0.61% and 11th THD s 0.47%. Fgure 5. Current smulaton waveform wth conventonal control strategy Fgure 6. Current smulaton waveform wth compound control strategy It can be seen from the above spectrum, the compensaton performance of these two control methods are relatvely well, but the compound control strategy could deal wth current peak (current spkes) better, bascally elmnatng the sharp and rregular current peak wth a more stable and better compensaton effect. 4. Expermental and Industral Applcaton Results 4.1. Experment Desgn The tral of shunt APF has been developed, accordng to Fgure 1(a) and Fgure 1(b). Test-rg photographs of shunt APF can be shown n Fgure 7. Expermental facltes consst of 100 kva three-phase four-wre shunt APF, varable harmonc sources and statc harmoncs source. Analyzed from the smulaton model, specfc APF system hardware desgn has a relatvely clear theoretcal framework. Control system adopts the core processor TMS320F28335 for the functon of exchangng data and real-tme performance. And APF adopts the dual-dsp parallel processng system based on data exchange mode. One pece of TMS320F28335 s responsble for sgnal acquston, data communcaton and error detecton. The other pece of TMS32028335 s used for control algorthm. In order to prevent conflct for accessng to the same address, APF system flexbly uses dual-ram port as medum of data exchange between the two DSPs. Hardware crcut s consstng of power supply and protecton crcut, clock crcut, external crcut, D/A crcut, etc. Man power crcut adopts Neutral-Pont Clamped topology, usng dscrete component desgn for reducng costs. IGBT part uses FF300R12KT4 of In fne on company and drver module uses 2QD15A17K-C. IJPEDS Vol. 3, No. 2, June 2013 : 209 217
IJPEDS ISSN: 2088-8694 215 Fgure 7. Test-rg photographs 4.2. Experment Results APF system adopts the Code Composer Studo to carry on programmng and debuggng. Acquston and analyss of expermental data adopts HIOKI PQA-HVew PRO 9624-50 power qualty analyzer and AGILENT Osclloscopes. In consderaton of uncertanty factors durng the experment, load parameters have been lmted. Expermental spectrum can as shown n Fgure 8 Fgure 11. Result parameters can be shown n Table 2 Table 4. Table 2. Load Currents Phase A Phase B Phase C RMS THD RMS THD RMS THD 27.7A 26% 28.5A 26.94% 28.1A 26.27% Table 3. Grd-sde Currents wth Compound strategy Phase A Phase B Phase C RMS THD RMS THD RMS THD 26.8A 4.20% 27.7A 4.59% 27.6A 4.35% Table 4. Full Response tme APF Swtch-On wth Compound strategy APF Swtch-Off wth Compound strategy 1 Bx Ax 9.818138ms 2 Bx Ax 6.18182ms Fgure 8. Spectrum of current before compensaton A Shunt Actve Power Flter wth Enhanced Dynamc performance usng Dual-Repettve Controller
216 ISSN: 2088-8694 Fgure 9. Spectrum of current after compensaton Fgure 10. Dynamc response of swtch-on Fgure 11. Dynamc response of swtch-off Table 2 Table 3 gves the parameters of Load and compensated Grd-sde current respectvely, and Table 3 Table 4 gves the parameters of full response tme of APF. Detaled parameters are shown n table 2 to table 3, from whch we can fnd that wth the applcaton of actve power flter (APF), system compensaton could effectvely reduce the total harmonc dstorton (THD) values from 26.02%, 26.94% and 26.27% to 4.20%, 4.59% and 4.35% for each phase of the current. In the feld test, the precson of the response tme and speed of the Aglent osclloscope was also taken nto consderaton. Further measurement was carred out on the total full response tme wth the bult-n functon of the APF equpment desgned n ths paper. Full response tme, whch s dfferent from transent response tme, s defned as the tme t takes to flter the harmoncs out of the power grd after the applcaton of harmonc load. The shorter the full response tme s, the hgher the trackng speed wll be and there wll be more dffcultes n the desgn of APF. As shown n Fgure 10 and Fgure 11, the full response tme s the measurement length between Ax and Bx. Based on the parameters presented below the tables, t took 9.81838ms and 6.18182ms for the APF system to swtch on and off, respectvely. Response tme requred for both are less than 10ms, whch means excellent trackng speed was acheved and the system could meet the requrement of system wth rapdly fluctuatng load. IJPEDS Vol. 3, No. 2, June 2013 : 209 217
IJPEDS ISSN: 2088-8694 217 Thus, adoptng the mproved predctve harmonc current closed-loop control strategy and the combnaton of PI control and DRC, compensaton precson of APF mproved a certan extent, and mproved the dynamc response speed. Expermental results also fulflled the requrement of system stablty and totally met the standards of IEC-61000-3 or IEEE -519 [12]. 5. CONCLUSION The artcle proposes a three-phase four-wre shunt APF, whch adopts the combnaton strategy of PI control and DRC and also uses the mproved predctve compensaton method. The Predctve compensaton strategy, on the condton of meetng the system stablty, could mprove current trackng accuracy especally for man harmoncs, and t wll elmnate errors brought by harmonc detecton, nverter and current control. Especally, ths strategy could detect any specfed harmoncs. Adoptng the combnaton strategy of PI control and DRC, APF system has mproved ts dynamc response speed and enhanced the ablty for elmnatng the transent current. Usng ths composte control strategy to carry on ndustral prototype smulaton and feld test, the expermental result shows that system compensaton could effectvely reduce the total harmonc dstorton (THD) values from 26.02%, 26.94% and 26.27% to 4.20%, 4.59% and 4.35% for each phase of the current. And the full response tme are all less than 10ms, fully meetng the standard of IEEE-519. Ths scheme has a certan economc benefts and popularzaton sgnfcance. REFERENCES [1] H Akag, Y Kanazawa, etc. Instantaneous reactve power compensators comprsng swtchng devces wthout storage components. IEEE Trans Ind. Appl. 1984; 20(3): 625-630 [2] H Akag, Y Kanazawa, etc. Generalzed theory of the nstantaneous reactve power n three-phase crcuts. IEEE&JIEE. Proceedngs IPEC. Tokyo: IEEE. 1983, 1375~1386. [3] LR Lmong, R Bojo, G Grva and A Tencon. New control scheme for sngle-phase actve power flters. Power electroncs specalsts conference, 2008. PESC 2008. IEEE, pp. 2894-2900. [4] Jaume Mret, Josep M Guerrero. Selectve harmonc-compensaton control for sngle-phase actve power flter wth hgh harmonc rejecton. IEEE transactons on ndustral electroncs. 2009, vol. 56, no. 8, pp. 3117-3127. [5] Xechuan, he chao, chenmng, etc. Dual-repettve control algorthm for three-phase shunt actve power flters. IECON 2010 on IEEE ndustral Electroncs socety. 2010, pp. 2038-2042. [6] WOO SOK CHANG, IL HONG SUH. Analyss and desgn of dgtal dual-repettve controllers. Proceedng of the 35 th conference on decson and control. 1996, pp. 2495-2500. [7] Y fewang, Yun wel. Three-phase cascaded delayed sgnal cancellaton PLL for fast selectve harmonc detecton. IEEE transatons on ndustral electroncs. 2013, vol. 60, no 4, pp.1453-1463. [8] Xzhengguo, Qang Han, etc. Closed-loop control method and mplementaton of selectve harmonc type actve power flters. Transactons of chna electrotechncal socety. 2006, vol. 21, no. 9, pp. 51-56 [9] Avk Bhattacharya, Chandan Chakraborty. A shunt actve power flter wth enhanced performance usng ANNbased predctve and adapatve controllers. IEEE transactons on ndustral electroncs. 2011, vol. 58, no. 2, pp.421-428. [10] Qunha Huo, Dong L, etc. IPI control strategy for APF. Electrc power automaton equpment. 2012, vol. 32, no. 12, pp. 43-47. [11] An Luo, Xangyong Xu, etc. Feedback-Feedforward PI-Type teratve learnng control strategy for hybrd actve power flter wth njecton crcut. IEEE Transactons on ndustral electroncs. vol. 57. No. 11. pp. 3767-3779. [12] IEEE recommended practces and requrements for harmonc control n electrcal power systems. IEEE Std 519-1992, p. 0_1, 1993. A Shunt Actve Power Flter wth Enhanced Dynamc performance usng Dual-Repettve Controller