Voltage Analysis of Distribution Systems with DFIG Wind Turbines

Transcription

1 1 Voltage Analyi of Ditribution Sytem with DFIG Wind Turbine Baohua Dong, Sohrab Agarpoor, and Wei Qiao Department of Electrical Engineering Univerity of Nebraka Lincoln Lincoln, Nebraka , USA Abtract Wind energy i becoming the mot viable renewable energy ource mainly becaue of the growing concern over carbon emiion and uncertaintie in foil fuel upplie and the government policy impetu. The increaing penetration of wind power in ditribution ytem may ignificantly affect voltage tability of the ytem, particularly during wind turbine cut-in and cut-off diturbance. Currently, doubly fed induction generator wind turbine (DFIG-WT) i the mot popular wind turbine. Thi paper invetigate the iue of voltage tability improvement and grid-lo reduction of ditribution ytem which include DFIG-WT under unbalanced heavy loading condition. Simulation tudie are carried out in IEEE 34-bu tet ytem with DFIG uing DIgSILENT oftware to examine thee iue during teady-tate and tranient operation of the ytem. Optimal location of the WT are determined baed on thi analyi. A new index (ytem unbalanced voltage variance) i propoed to evaluate ytem unbalanced voltage. The dynamic impact between WT and motor load are alo examined. Index Term Voltage tability, doubly-fed induction generator (DFIG), ditribution ytem, optimal location, wind turbine, grid lo, unbalanced power flow. I. INTRODUCTION ind energy i a clean, renewable and indigenou energy W reource. Wind power capacity i annually added over 20% in recent five year becaue of it relatively low MWh cot [1]. Wind power could be combined with appropriate reactive power compenation to improve the operation of the electricity ditribution ytem. Some of the often-quoted benefit include the following [2]: 1) Emergency backup during utained utility outage. 2) Voltage upport. 3) Lo reduction. 4) Improved utility ytem reliability. 5) Ditribution capacity releae. 6) Potential utility capacity addition deferral. On the other hand, the integration of wind power may caue ome additional problem in voltage and frequency regulation, reactive power compenation, power quality (e.g., flicker and harmonic) and protection of the ditribution ytem [3]-[8]. In general, reactive power compenation are very important to WT. Lack of ufficient reactive power may reult in voltage typically lower ( p.u.) than normal voltage [3]. The mot eriou diturbance in ytem operation occur during the cut-in and cut-off of the WT from the grid, grid fault, etc. During the tranient tate after a diturbance, the flicker level, harmonic, frequency and voltage fluctuation harply increae; in ome autonomou power ytem, the frequency and voltage level would decreae more than 1Hz and more than p.u [6], repectively. The WT with large capacitie could poibly change the original power flow direction and the protection ytem need to be et new pick-up value to make the breaker work properly. Therefore, not only the protection ytem, but alo the interconnection requirement and ilanding operation hould be conidered more for ytem tability [9]. With the recent development in power electronic, the DFIG-WT become very popular. Comparing to traditional quirrel-cage induction generator wind turbine, the major advantage of DFIG are that they can operate in a higher wind peed range and produce or conume reactive power through the magnetization provided by the rotor-ide converter. DFIG- WT alo offer other advantage, uch a [10]: 1) Low converter and inverter cot, becaue the converter rating i typically 25% of the total power rating of the DFIG. 2) Capability for power-factor control and reactive power compenation. 3) Reduced cot of the inverter filter. 4) Improved WT efficiency. Mot work in thi area mainly focued on the reearch and imulation of a imple tet ytem with a DFIG-WT. Thi article i concerned with voltage analyi of a moderate ize ytem, i.e., the IEEE 34-bu tet ytem [11], which include DFIG-WT. Simulation tudie are carried out in DIgSILENT oftware [12] to invetigate effect of WT on the teady-tate and tranient behavior of the ditribution ytem. Thi article i organized a follow. Section II-A preent everal cae tudie during teady tate uing the tet ytem which ha a ingle DFIG-WT and provide a new index to evaluate ytem unbalanced voltage. Section II-B compare the teady-tate impact between a ingle large-dfig ytem and a multiple mall-dfig ytem. Section II-C compare the tranient impact between the two ytem. Section III analyze the imulation reult and provide ome dicuion. II. SYSTEM SIMULATION The IEEE 34-bu tet ytem hown in Fig.1 i a typical radial-ditribution ytem and the imulation ytem i contructed with DIgSILENT uing detailed data from [11].

2 2 The ytem ha two auto-regulator which regulate the voltage in ± 0.05 p. u. range and one auto-tranformer which control the voltage in ± 0.05 pu.. range. They are et to automatic-tap adjutment for the baic cae poition becaue their initial tap poition ha a ignificant role on the ytem voltage regulation. The ditributed load are handled a plit load with half on each bu. cae for the ret imulation tudie. The ingle 1.5-MW DFIG-WT (DIgSILENT provided) i connected at variou bu location to evaluate which location provide the mot ytem-voltage improvement and the mot reduction in grid lo. In order to eaily analyze the difference between the average bu voltage and the ytem unbalanced voltage variance, an index (1.05-) i elected. The reult i hown in Figure 2 and Fig. 1. IEEE 34-bu tet ytem. In the tet ytem, bu 800 i connected to the grid and bue 890 and 844 are the load center. The voltage at bu 890 i uually low becaue the line between bu 888 and 890 i relatively long. Several hunt capacitor are added at bue 844 and 848 to totally compenate for reactive power 0.75 MVar. The two regulator work together to control bu 852 voltage and it play an eential role in ytem voltage control. In light loading condition given by [11], the total unbalanced load i MW and MVar; the grid lo i 0.27 MW; the voltage at bue 852 and 890 are p.u. and p.u., repectively. Other bu voltage are in the range of p.u. Bu 890 i apparently the weaket bu. A. Baic cae tudie with a ingle DFIG-WT connected to the tet ytem The average bu voltage i uually ued to evaluate the ytem voltage profile [13]. However, it doe not reaonably reflect the ytem voltage unbalanced factor. A new index (ytem unbalanced voltage variance) i propoed to evaluate ytem voltage a follow: N ViA + ViB + ViC V = (1) i= 1 3N N ( ViA 1.05) + ( ViB 1.05) + ( ViC 1.05) = (2) i= 1 3N where ViA, ViB, V are the 3-phae voltage at bu i; N i the ic ytem bu number; V i the ytem average voltage; 2 i the ytem unbalanced voltage variance (uing 1.05 to replace V for a higher accuracy). If a contant load (1.0 MW, 0.1 MVar) i added at the terminal bu 840 and reactive power compenation (0.30 MVar) i added at bu 890, the regulator 2 and tranformer reach the maximum tap poition. Bu 852 voltage i p.u. and the grid lo i 0.71 MW. If the tap of the regulator 1 i manually increaed, the three-phae voltage at bu 890 change and become more unbalanced. Therefore, thi unbalanced heavy loading condition i elected a the bae Fig. 2. Average bu voltage and value when a 1.5-MW DFIG-WT i placed at different location. Fig. 3. Grid lo when a 1.5-MW DFIG-WT i placed at different location. With the DFIG location cloer to the load center (bue 840, 844, and 890), the voltage profile i improved more and the grid loe greatly reduced. The average bu voltage and value baically have the ame trend. However, the latter ha a higher accuracy to evaluate the unbalanced ytem voltage improvement, epecially where DFIG i connected at bue 840 and 888. The average bu voltage index could not reflect the degree of voltage unbalance. At bu 840, value reache the maximum value while at bue 834, 842, 844, 860, 836, 840, 862, and 838, the grid loe i minimum. Therefore, the optimal location i determined to be bu 840. The tranformer adjut voltage range to increae from ± 0.05 pu.. to ± 0.1 p. u. when DFIG i connected at bu 890 becaue it i a low-voltage bu (4.16 KV), otherwie more voltage (0.13 p.u.) drop in the long low-voltage line between

3 3 bue 890 and 888. If more power tranfer through lowvoltage line to the grid, the grid lo increae to 0.30 MW. If the WT i large, it i better to intall it at a high voltage tranmiion line and cloer to the load center. However, if the load center are far away from each other or the load are more ditributed, the optimal location hould be carefully calculated by a formulating multi-objective optimization: min : pki + (1 p ) Ploi,( i different cae) (3) Where: i the tandard deviation correponding to ytem i unbalanced voltage variance defined in Equation (2). k i a converion factor in (MW/Volt) to adjut them on the ame number cla. P i the grid power lo aociated with the cae loi. i p i the weight aociated with the cae i ; 1 p i the weight of P loi. Equation (3) i ued to minimize the ytem unbalanced voltage variation and the grid power lo. B. Comparion of teady-tate impact between a ingle large-dfig ytem and a multiple mall-dfig ytem Two maller DFIG-WT (0.75 MW) are added at bue 832 and 836 to compare the teady-tate impact with a ingle DFIG-WT (1.5 MW) at different bue in the ytem. The reult are hown in Fig. 4. are determined when ubjected to thee diturbance. However, the ize of ome dynamic model i adjuted to provide a table convergence. 1) Cae C1: the ingle DFIG (0.75 MW) i placed at bu 856; reactive power compenation i adjuted to 0.60 MVar at bu 890 and the load at bu 890 i adjuted to 0.25 MW motor. Thee following diturbance occur in equence: At t=0.18, a 3-phae hort circuit (uing fault impedance Z=15+j40 Ohm to control the voltage of bu 840 at 0.5 p.u. in the fault) at bu 862; At t=0.28, clear hort circuit; At t=0.35, cut the 1-MW load (at bu 840); At t=0.45, recover the load; At t=0.65, cut the motor; At t=0.8, cut-in the motor; At t=1, cut-off DFIG; At t=1.18, cut-in DFIG. The reult are hown in Figure 5 and 6. Fig. 5. The voltage of bue 840 and 890 for cae C1 Fig. 4. Comparion of grid lo and value between a ingle large-dfig ytem and a multiple mall-dfig ytem. A hown, the multiple mall-dfig ytem could provide higher value and better ytem voltage improvement. It grid lo i almot the ame a the optimal grid lo of the ingle large-dfig ytem. Moreover, the optimal location of the multiple mall-dfig ytem could be calculated and it grid lo would be le. Therefore, the multiple mall-dfig ytem i more uitable than the ingle large-dfig ytem. C. Comparion of tranient impact between a ingle large- DFIG ytem and a multiple mall-dfig ytem Short circuit, cutting load (including contant Z load and motor), and DFIG cut-off or cut-in are the major diturbance in ditribution ytem. The load-center voltage waveform Fig. 6. The voltage of bue 844 and 900 (connected DFIG and bu 856) for cae C1 From Figure 5 and 6 which are hown, the load-center voltage fluctuate about 0.1 p.u. becaue of the cut-off of the large load which affect almot all bu voltage. The dynamic impact of cut-in DFIG and the motor are larger than cut-off

4 4 and the motor ha a ignificant role on the local bu voltage. Therefore, the low voltage bu 890 i the mot fluctuant, and conidered a a weak bu. 2) Cae C2: Baed on the condition of cae C1, thee following diturbance occur imultaneouly: At t=0.18, 3-phae hort circuit at bu 862; At t=0.2, cut the motor; At t=0.25, cut-off DFIG; At t=0.38, clear hort circuit; At t=0.43, cut-in DFIG; At t=0.5, cut-in the motor; At t=1.4, cut the large load; At t=1.6, recover the load. The reult are hown in Figure7 and 8. in Figure 9 and 10. Fig. 9. The voltage of bue 840 and 890 for cae C3 Fig. 7. The voltage of bue 840 and 890 for cae C2 Fig. 10. The voltage of bue 844 and 900 for cae C3 By comparing Figure 9 and 10 with Figure 5 and 6, the multiple mall-dfig ytem bore a le dynamic impact and it take le time to return to it initial tate. Furthermore, the ytem voltage maintain in the normal range from 0.95p.u. to 1.05p.u. for longer period of time. 4) Cae C4: Baed on the condition of cae C2, two DFIG (0.35 MW) are added at bue 856 and 840 and thee diturbance imultaneouly occur and are the ame a the diturbance in cae C2. The reult are hown in Figure11 and 12. Fig. 8. The voltage of bue 844 and 900 for cae C2 From the reult hown in Figure7 and 8, the voltage of bu 890 connected to the motor i more affected and it take more time (0.5) to return to it initial tate when thee diturbance imultaneouly occur. 3) Cae C3: Baed on the condition of cae C1, two DFIG (0.35MW) are added at bue 856 and 840; the load at 840 i adjuted to 0.7 MW; the diturbance eparately occur and are the ame a the diturbance in cae C1. The reult are hown

5 5 Fig. 11. The voltage of bue 840 and 890 for cae C4 In dynamic imulation, different diturbance, hort circuit, cutting load (including contant Z load and motor), cut-off and cut-in DFIG, have been applied eparately or imultaneouly. The load-center voltage fluctuate about 0.1 p.u. in thi ditribution ytem becaue of cutting the large load or cut-in DFIG and they affect almot all bu voltage. The dynamic impact of cut-in DFIG i apparently more ignificant than cut-off DFIG. The motor ha a ignificant role on the local bu voltage and i very enitive to diturbance. The oft-tart motor and DFIG technology hould be paid more attention in ditribution ytem. The multiple mall- DFIG ytem bore a le dynamic impact and need le time to return it initial tate than the ingle large-dfig ytem. Moreover, the former voltage fluctuation i in a maller range and the boundary value are more uitable and the ytem voltage maintain in the normal range from 0.95p.u. to 1.05p.u. for longer time. Therefore, the multiple mall-dfig ytem ha a better tability than the ingle large-dfig ytem. IV. ACKNOWLEDGMENT The author are grateful to the Nebraka Center for Energy Science Reearch for their upport of thi project. Fig. 12. The voltage of bue 844 and 900 for cae C4 Conidering the difference between Figure 11 and 12 and Figure 7 and 8, the multiple mall-dfig ytem ha a better tability. The voltage fluctuation i in a maller range and the boundary value are more uitable. III. CONCLUSION Thi paper ha invetigated the effect of DFIG-WT on voltage tability and grid lo of ditribution ytem under unbalanced load condition. A new index (ytem unbalanced voltage variance) ha been propoed to evaluate ytem unbalanced voltage, which i more reaonable and more accurate than that uing ytem average voltage. A new multiobjective optimization function ha been provided to calculate the optimal location imultaneouly conidering the voltage profile and grid lo. Simulation tudie have been carried out in DIgSILENT oftware to examine the impact of DFIG- WT on the teady-tate and dynamic behavior of the IEEE 34-bu tet ytem under unbalanced heavy loading condition. The optimal location of the ingle large DFIG-WT in the tet ytem ha been determined. Through the comparion between the ingle large-dfig ytem and the multiple mall- DFIG ytem, the latter could provide higher ytem voltage advancement and almot the ame grid lo reduction a the former. V. REFERENCES [1] Ryan Wier, Mark Bolinger Annual Report on U.S. Wind Power Intallation, Cot, and Performance Trend: 2006, [2] Philip P. Barker, Robert W. de Mello Determining the impact of ditributed generation on power ytem. I. Radial ditribution ytem, Power Engineering Society Summer Meeting, Vol 3, pp , July, [3] Muljadi E., Butterfield C.P., Chacon J., Romanowitz H. Power quality apect in a wind power plant, Power Engineering Society General Meeting, pp.8-16, June, [4] Hanen A.D., Sorenen P., Janoi L., Bech J. Wind farm modelling for power quality, Indutrial Electronic Society, IECON'01. The 27th Annual Conference of the IEEE Volume 3, pp , 29 Nov.-2 Dec, [5] Gardner P. "Flicker from wind farm", BWEA/SERC RAL Workhop on wind energy penetration into weak electricity network, Rutherford, U.K., pp.27-37, [6] Ladakako P.D., Ioannide M.G., Koulouvari M.I. "Aement of wind turbine impact on the power quality of autonomou weak grid", Harmonic and Quality of Power, Proceeding. 8th International Conference on Volume 2, pp , Oct, [7] Seyoum D., Grantham C., Rahman F. "The dynamic of an iolated elfexcited induction generator driven by a wind turbine", Indutrial Electronic Society, IECON'2001. The 27th Annual Conference of the IEEE Volume 2, pp , 29 Nov.-2 Dec [8] Doyle M.T. "Reviewing the impact of ditributed generation on ditribution ytem protection", Power Engineering Society Summer Meeting, 2002 IEEE, Volume 1, pp , July, [9] Aktarujjaman M., Kahem M.A., Negnevitky M., Ledwich G. "Control Stabiliation of an Ilanded Sytem with DFIG Wind Turbine", Power and Energy Conference, PECon'06, IEEE International Nov pp [10] Muller S., Deicke M. "Doubly fed induction generator ytem for wind turbine", IEEE indutry application magazine, May /June, [11] [12] [13] Jone G.W., Chowdhury B.H. "Ditribution ytem operation and planning in the preence of ditributed generation technology", Tranmiion and Ditribution Conference and Expoition, T&D. IEEE/PES, pp.1-8, April 2008