Fault Behavior of Wind Turbines. Sulla, Francesco. Published: Link to publication

Transcription

1 Fault Behavior of Wind Turbine Sulla, Franceco Publihed: Link to publication Citation for publihed verion (APA): Sulla, F. (1). Fault Behavior of Wind Turbine General right Copyright and moral right for the publication made acceible in the public portal are retained by the author and/or other copyright owner and it i a condition of acceing publication that uer recognie and abide by the legal requirement aociated with thee right. Uer may download and print one copy of any publication from the public portal for the purpoe of private tudy or reearch. You may not further ditribute the material or ue it for any profit-making activity or commercial gain You may freely ditribute the URL identifying the publication in the public portal Take down policy If you believe that thi document breache copyright pleae contact u providing detail, and we will remove acce to the work immediately and invetigate your claim. L UNDUNI VERS I TY PO Box117 1L und +4646

2 Fault Behavior of Wind Turbine Franceco Sulla Doctoral Diertation Department of Meaurement Technology and Indutrial Electrical Engineering 1

3 Department of Meaurement Technology and Indutrial Electrical Engineering Faculty of Engineering Lund Univerity Box LUND SWEDEN ISBN: CODEN: LUTEDX/(TEIE-163)/1-15/(1) Franceco Sulla, 1 Printed in Sweden by Tryckeriet i E-huet, Lund Univerity Lund 1

4 To my family

5

6 Abtract Synchronou generator have alway been the dominant generation type in the grid. Thi fact affected both planning and operation of power ytem. With the fat increae of wind power hare in the grid in the lat decade, the ituation i changing. In ome countrie wind power repreent already a conitent amount of the total generation. Wind turbine can be claified a non-ynchronou generation and they behave differently than ynchronou generation under many circumtance. Fault behavior i an important example. Thi thei deal with the behavior of wind turbine during fault in the grid. The firt part focue on the fault current delivered by wind turbine with Doubly-Fed Induction Generator (DFIG). The econd part invetigate the impact of fault below the tranmiion level on wind turbine grid fault ride-through and the voltage upport that wind turbine can provide in weak grid during fault. A wide theoretical analyi of the fault current contribution of DFIG wind turbine with crowbar protection i carried out. A general analytical method for fault current calculation during ymmetrical and unymmetrical fault in the grid i propoed. The analytical method can be ued to find the maximum fault current and it AC or DC component without the need to actually perform detailed imulation, which i the method ued today. DFIG wind turbine may alo be protected uing a chopper reitance on the DC-link. A method to model the DC-link with chopper a an equivalent reitance connected to the generator rotor during ymmetrical grid fault i preented. Thi allow to calculate the hort-circuit current of a DFIG with chopper protection a an equivalent DFIG with crowbar protection. Thi i v

7 ueful ince fault current calculation method for DFIG with crowbar are available in the literature. Moreover, power ytem imulation tool include tandard model of DFIG wind turbine with crowbar protection, but often not with chopper protection. The ue of an aggregate model to repreent the fault current contribution of a wind farm ha been analyzed through imulation. It ha been found that the aggregate model i able to reproduce accurately the total fault current of the wind farm for ymmetrical and unymmetrical fault. The ue of aggregate model implifie imulation model and ave imulation time. The Swedih grid code require wind turbine at all voltage level to ride through fault at the tranmiion network. For fault at voltage level below tranmiion level fault clearing time are often longer and thi could impact on fault ride-through of wind turbine. Simulation of tudy cae with fault at ub-tranmiion level, performed uing the tandard Nordic 3 tet ytem, how that wind turbine hould till be able to ride through uch fault. Only in cae of high dynamic load cenario and failure of the protection ytem, wind turbine could diconnect from the grid. Load modelling i important when carrying out thi analyi. Fault on adacent MV feeder eriouly endanger grid fault ride-through (GFRT) of wind turbine. Finally, an invetigation on the voltage upport of wind turbine in weak network during fault ha been carried out. A implified model of the power ytem of the Danih iland of Bornholm ha been ued a a tet ytem. It ha been found that the minimum requirement for voltage upport et by grid code do not reult in atifactory voltage recovery in weak grid after fault clearing. However, if properly controlled, wind turbine are able to provide a voltage upport comparable to that upplied by power plant with ynchronou generation. vi

8 Acknowledgement Firt of all, I would like to thank my upervior Dr. Olof Samuelon for accepting me a a Ph.D. tudent, for all the help he provided during thi period and for hi trut and encouragement. Hi uggetion, guidance, advice and comment have been of maor importance for me. I would alo like to thank all the co-upervior during my Ph.D. tudie: Prof. Sture Lindahl, whoe deep knowledge and experience have been a valuable ource of inpiration, for the help during my Licentiate tudie; Dr. Magnu Akke for commenting on my Licenciate manucript; Dr. Jörgen Svenon for the comment, the careful and patient proofreading of the many verion of my Ph.D. manucript and other article. Special thank to all the other people who helped during thi work: the teering group member, Dr. Stefan Arnborg from Svenka Kraftnät, M.Sc. Ander Ekberg from Fortum, Dr. Jona Peron from Vattenfall and Tech. Lic. Nikla Stråth from E.ON, for providing idea, comment and inight. It ha been a pleaure to meet again Jona Peron, who ha alo been the main upervior of my mater thei at KTH in Stockholm. Dr. Qiuwei Wu from DTU, Copenhagen, i thanked for the help he provided with the model of the Bornholm power ytem ued in thi thei. I wih to expre my incere gratitude to my office mate Anna Guldbrand, Johan Börntedt and Ingmar Leie for the friendly atmophere they created, the intereting dicuion and the upport they alway provided. I would alo like to thank Johan for co-authoring an article and for all the help he gave me during my work in the lab, Anna and Ingmar for their company during variou conference travel. A incere thank goe to all the people at the department for the pleaant work environment. Special thank to Conny Högmark and Luyu Wang for the relaxing training moment, to Carina Lindtröm and Getachew Darge who have alway been very helpful. vii

9 I would like to thank all the people I met here in Lund and that contributed to a nice ocial life outide of the department. In particular Kaia, Kinga, Luca, Anna, Yuri, Johan for all the nice dinner, lunche, evening, concert, excurion. Thank for the nice time. Finally, I want to dedicate thi thei to my wife Ela, to my parent and to my iter. Queta tei e dedicata a voi. Elfork AB i acknowledged for the financial upport of thi work, Elektra proect Lund, April 1 Franceco Sulla viii

10 Preface When I tarted in 6, the ubect of my Ph.D. tudie wa iland operation with induction generator. That proect wa financed by Svenka Kraftnät and wa divided into two ub-proect. I wa reponible for the part on fault analyi and protection. That work wa documented in a Licentiate thei, publihed in 9. Since the proect wa not continued another proect wa formulated. At the end of 9, thi proect titled Power ytem integration of nonynchronou generation wa tarted, financed by Elfork AB. Thi proect i alo divided into two ub-proect. Thi thei deal with the part on fault current contribution and grid fault ride-through of wind turbine. The other part focue on the impact of non-ynchronou generation on power ytem frequency. ix

11

12 Content CHAPTER 1 INTRODUCTION MOTIVATION OBJECTIVES AND OUTLINE OF THE THESIS CONTRIBUTIONS PUBLICATIONS...8 CHAPTER WIND TURBINES DURING FAULTS WIND TURBINE CONCEPTS WIND TURBINE FAULT CURRENT GRID FAULT RIDE-THROUGH REQUIREMENTS GRID FAULT RIDE-THROUGH SOLUTIONS....5 SUMMARY...5 CHAPTER 3 FAULT CURRENTS OF SCIG AND DFIG WIND TURBINES INTRODUCTION INDUCTION GENERATOR MODELLING SCIG SHORT-CIRCUIT CURRENT SCIG SIMULATION RESULTS DFIG SHORT-CIRCUIT CURRENT DFIG SIMULATION RESULTS SUMMARY...53 CHAPTER 4 FAULT CURRENTS OF DFIG WIND TURBINES WITH CHOPPER INTRODUCTION DFIG WITH DC CHOPPER PROTECTION DIODE BRIDGE RECTIFIER WITH DC VOLTAGE SOURCE DFIG WITH DC CHOPPER PROTECTION SIMULATIONS SUMMARY...74 xi

13 CHAPTER 5 FAULT CURRENTS FROM WIND FARMS INTRODUCTION DETAILED WIND FARM MODEL AGGREGATE WIND FARM MODEL DESCRIPTION OF SIMULATIONS SIMULATION OF A SCIG WIND FARM SIMULATION OF A DFIG WIND FARM SUMMARY...87 CHAPTER 6 GFRT FOR FAULTS BELOW TRANSMISSION NETWORK INTRODUCTION NETWORK AND COMPONENTS MODELLING VOLTAGE DIP CALCULATION FAULTS AT SUB-TRANSMISSION NETWORK SIMULATION RESULTS FAULTS AT MEDIUM VOLTAGE NETWORK SUMMARY...13 CHAPTER 7 WIND TURBINES VOLTAGE SUPPORT IN WEAK NETWORKS INTRODUCTION WIND TURBINE VOLTAGE SUPPORT VOLTAGE SUPPORT LIMITS COORDINATED ACTIVE AND REACTIVE CURRENT INJECTION VOLTAGE SUPPORT STRATEGIES DURING FAULTS STUDY CASE THE BORNHOLM POWER SYSTEM SUMMARY...14 CHAPTER 8 CONCLUSIONS AND FUTURE WORK THESIS SUMMARY FUTURE WORK...19 REFERENCES xii

14 Chapter 1 Introduction The amount of intalled renewable energy in the electric power ytem ha been teadily increaing in the pat decade. Thi trend i due to different factor, ocial, environmental, political, economical and technological. Society increaingly demand for environmentally acceptable energy olution; government often promote renewable energy ource by economically ubidizing their intallation; technology and economic of ome type of renewable energy have evolved to a point that make them competitive with electricity generation baed on foil fuel or nuclear power. After hydropower, wind energy i the mot mature among the renewable energy technologie. According to (Pure Power 11) the global cumulative wind power capacity intalled in the whole world increaed from 6.1 GW in 1996 to 197 GW in 1 and the global annual intalled wind power capacity ha been lightly over 38 GW both in 9 and 1. Thi number repreent % of the total new electric power capacity intalled in the world in 1, equal to 194 GW (REN1 11). In the pat decade, from 1 to 1, the global wind power capacity increaed by 18 GW, while in the ame period the total new intalled nuclear capacity ha been 6 GW (Pure Power 11). In countrie like Denmark and Spain, the wind energy production in year 1 repreented a hare of 5 % and 15 % repectively of the annual electricity conumption in the country, referred to year 8. Thi trend can be found alo in Sweden; the total intalled wind power in 1 wa lightly over GW and the wind energy production grew from.63 TWh in 3 to 3.51 TWh in 1, correponding to.4 % of the total electricity production (Vindkrafttatitik 1). Preliminary data indicate that wind energy production in Sweden grew to 6.1 TWh in year 11 and i expected to reach 8 TWh 1. In (Pure Power 11) cenario are alo 1

15 Chapter 1. Introduction preented for wind power development in the EU up to. In the Baeline cenario, wind power i expected to increae from a hare of 5.3 % in 1 to roughly more than 15 % in of the total EU electricity demand. The growth in intalled wind power capacity ha been accompanied by advance in wind turbine technology. The average ize of intalled wind turbine ha continually increaed, paing, e.g. in Sweden, from.6 MW until 3 to 1.9 MW in 1 (Vindkrafttatitik 1). Wind turbine up to a 6 MW range are today available on the market and wind turbine in the 1 MW range are under deign (Manwell 9). The electrical technology in the wind turbine ha alo evolved. The firt wind turbine were equipped with quirrel-cage induction generator (SCIG) directly connected to the grid. Thee wind turbine operated at almot fixed peed and uffered from high flicker emiion. Partly-variable-peed wind turbine with induction generator equipped with a dynamic rotor reitance allow operation at up to 1 % uper-ynchronou peed and can be controlled to reduce flicker emiion and mechanical tree by aborbing the fluctuating wind power into the rotor kinetic energy and the external rotor reitance (Akhmatov 5). Modern wind turbine are of variable-peed type. The mot common olution have a Doubly-Fed Induction Generator (DFIG) or an induction or ynchronou generator interfaced to the grid through a Full-Scale Converter (FSC). They are able to optimize the power extraction at wind peed below rated peed, have low flicker emiion and feed a mooth power into the grid (Akhmatov 5). Both thee type of wind turbine are equipped with power electronic back-to-back converter. Up to ome year ago, wind turbine were required to diconnect from the grid during evere diturbance. Thi would avoid any poible negative effect of wind turbine on the grid voltage and the fault clearing ytem and it would eliminate the rik for non-intentional ilanding with it aociated afety rik (Piwko 1). With the high penetration reached by wind power today thi olution i no longer viable, ince it could reult in the diconnection of a large amount of generation, weakening the power ytem and poibly creating problem for the frequency control. A a conequence, the latet national grid code require wind turbine to remain connected and ride-through a fault in the grid providing active and reactive power upport. 1.1 Motivation The requirement on grid fault ride-through radically changed the role wind

16 1.1. Motivation 3 turbine play in the power ytem during and immediately after a diturbance. Wind turbine mut now remain connected through the whole duration of the diturbance under pecific circumtance. Remaining connected to the grid during a fault, wind turbine will alo feed hort-circuit current into the grid. The hort-circuit current delivered by wind turbine i different from the hort-circuit current fed into the grid by ynchronou generator. In area with a high concentration of wind power thi may impact on the operation of protection relay. Knowledge of the hort-circuit current delivered by wind turbine i alo of importance for power ytem component izing. The relevance of thi topic i underlined by the fact that a pecial Joint Working Group ha been et up in 8 by the IEEE Power Sytem Relay Committee in order to invetigate the hort-circuit current delivered by wind turbine. The aignment of the Working Group i (Webpage, To upport the activitie in the production of a report that characterize and quantifie the hort circuit current contribution to fault from wind plant for the purpoe of determining protective relay etting and fault interrupting equipment rating. The report will provide guideline on the modeling and calculation for that purpoe. Wind turbine of different type will deliver different hort-circuit current. For wind turbine with FSC, the hort-circuit current i mainly determined by the control cheme and therefore it i alo manufacturer-dependent. For wind turbine equipped with DFIG and fixed-peed wind turbine, it i intead the phyical ytem that determine the hort-circuit behavior immediately after evere voltage dip. For thee wind turbine it i therefore poible to lay down a general treatment of the hort-circuit behavior. The hort-circuit behavior of wind turbine with DFIG i of particular interet, ince thi i the mot commonly intalled type of wind turbine in the lat decade and will continue to repreent a ignificant hare of the total wind power intalled in the power ytem for many year to come. Some grid code alo require wind turbine to be able to upply voltage upport during depreed voltage condition and quickly return to normal operating condition once the grid diturbance ha ceaed. The limiting voltage dip profile condition above which wind turbine are required to ridethrough a fault are pecified in the national grid code and vary from country to country. Such a profile i a voltage-time curve repreenting a limiting condition within which voltage dip are expected to be contained, following fault in a point in the grid and the ubequent clearing by the protection ytem under normal circumtance.

17 4 Chapter 1. Introduction The reearch into grid fault ride-through of wind turbine ha flourihed in the lat year and thi fact underline by itelf the relevance of the topic. A multitude of paper produced by academic have been accompanied by advanced hardware and oftware olution adopted in the new wind turbine by manufacturer to aure grid fault ride-through capability. The competition in the field among manufacturer i fierce and thi make it very difficult to get detailed information on the particular grid fault ride-through technical olution adopted in commercial wind turbine. The vat maority of the literature publication in the field focue on technical olution for the individual wind turbine to aure ride-through capability when a pecified voltage dip profile i aumed. Le explored in the literature i the grid fault ride-through behavior of wind turbine for fault in the ub-tranmiion and medium voltage network. In thee cae, the voltage dip profile reulting from a fault depend on the protection ytem ued at thee voltage level and the type of load in the grid (Souza 1); the voltage dip may lat long enough to endanger the grid fault ride-through capability of wind turbine. On the other hand, the extenion of the area of the grid intereted by the voltage dip i more limited a compared to a fault in the tranmiion level (Souza 1, Dahlgren 6). It eem therefore intereting to further invetigate thee iue to better undertand in which cae wind turbine may be expected to diconnect from the grid and to get an inight onto the extenion of the area affected by the voltage dip. Another topic that ha not been extenively reearched i how wind turbine intalled in weak network perform during a diturbance in the grid. Modern wind turbine have voltage upport capability and may repreent an aet to improve voltage performance in thee ituation. In (Neumann 11, Shawarega 9, Ullah 7, Ahkhane 11), it i uggeted that a high proportional gain and continuou voltage regulation without deadband may reult in better voltage during a fault. Alo, active power inection from wind turbine i pointed out a a caue that deteriorate the voltage upport offered by reactive power inection. However, a general treatment of the voltage upport of wind turbine in weak network i eldom found in the literature and many open iue till remain. What i the bet voltage upport trategy for wind turbine in weak network, how doe thi trategy depend on the grid hort-circuit power and X/R ratio, how doe the active power inection affect the voltage recovery after fault clearing, are the baic requirement contained in the grid code enough to achieve a atifactory voltage recovery in weak network? An invetigation of thee apect i ueful and helpful in fully exploiting the voltage upport capabilitie of wind turbine in weak network.

18 1.. Obective and Outline of the Thei 5 1. Obective and Outline of the Thei Thi thei deal with the behavior of wind turbine during grid diturbance. The main obective of the thei can be ummarized a follow: - Analyze the hort-circuit current fed into the grid by DFIG wind turbine both with crowbar and DC-chopper protection. - Analyze the fault current contribution of wind farm. - Invetigate voltage profile reulting from fault at ub-tranmiion and medium voltage network to undertand if problem for grid fault ride-through of wind turbine may arie. - Invetigate and devie poible way to improve voltage upport by wind farm in weak network during a grid diturbance. A brief theoretical background about wind turbine, the way they are protected and controlled during a grid diturbance and the requirement et by grid code on fault ride-through i given in Chapter. Wind turbine with DFIG are equipped with a back-to-back converter connected between the grid and the generator rotor. A voltage dip may caue high rotor current that could damage the converter witche. The tandard olution ued to protect the converter witche i to divert the rotor current through a o-called crowbar reitance. During a fault, the DFIG i in effect an induction generator with hort-circuited rotor, but with high rotor reitance. Approximate method for etimating the hort-circuit current of uch wind turbine for bolted three-phae fault at the machine terminal have been previouly publihed in the literature (Morren 7), but a general analyi i lacking. In Chapter 3, a theoretical analyi on the ymmetrical and unymmetrical hort-circuit current delivered by wind turbine equipped with SCIG and DFIG with crowbar protection i preented. In the lat year, ome wind turbine manufacture have adopted a new olution to protect the rotor ide converter. Intead of uing the tandard crowbar olution, the high rotor current in the event of a fault flow through the anti-parallel diode of the converter into the DC-link. A chopper reitance i connected acro the DC-link capacitor and it i witched to keep the DC-link voltage within an acceptable range. The many non-linearitie in thi ytem make it difficult to analyze the hort-circuit current of thee wind turbine. A implified analyi, which i found to be accurate for bolted three-

19 6 Chapter 1. Introduction phae fault and when the wind turbine i at full power, i preented in (Martinez 11(a)). In Chapter 4, a general analyi of the hort-circuit current of uch wind turbine for ymmetrical fault i preented. The idea of modelling thee wind turbine a an equivalent wind turbine with crowbar reitance i ued. The analyi preented in Chapter 3 and 4 allow to predict the hort-circuit current of the individual wind turbine. It may be more intereting to know intead the total hort-circuit current delivered by a wind farm with many wind turbine of the ame type. The ame type of wind turbine in a wind farm may deliver different hort-circuit current mainly becaue of a different initial tate and of voltage drop within the farm. Uing an aggregate model of the wind farm i however beneficial and ave imulation time. The validity of aggregate model ha been invetigated mainly with regard to the active and reactive power contribution (Pöller 3, Garcia-Gracia 8) and only limited material i available in the literature on uing aggregate model for fault tudie (Aluko 1, Perdana 8). In Chapter 5, imulation reult are preented to ae the validity of an aggregate model for ymmetrical and unymmetrical hort-circuit current tudie. Wind farm with SCIG and DFIG are both conidered. In Sweden, the voltage profile for grid fault ride-through in the grid code are defined with reference to fault in the tranmiion network (SvKFS 5:). Fault at the ub-tranmiion and medium voltage network may caue voltage profile quite different from the one given in the grid code, mainly becaue of difference in the protection ytem, and till affect large area of the power ytem. Some cae tudie are preented in Chapter 6 to gain inight into whether thee fault may endanger the grid fault ridethrough capability of wind turbine, on the extenion of the power ytem area affected by the voltage dip and on the influence of the load mix compoition and load behavior during the diturbance on the voltage dip itelf. Modern wind turbine are able to and by ome grid code required to provide voltage upport during a grid diturbance. Thi capability can be very beneficial in weak network, where voltage retoration after fault clearing can be low and much influenced by the connected load. The influence of the proportional gain, of a regulation with or without deadband, on coordinating the active and reactive current inection and on the pot-fault active power ramp on the voltage upport i invetigated in Chapter 7. Different voltage upport cheme for wind turbine, including the one required by the Danih and German (E.ON) grid code, are compared and ome concluion

20 1.3. Contribution 7 are drawn on poible way to improve the voltage upport of wind turbine in weak grid. Finally, the performance of wind turbine in a weak network i compared to that of a power plant with a ynchronou generator in a tudy cae, uing a implified model of a real power ytem, the one of the Danih iland of Bornholm. Chapter 8 ummarize the main concluion of thi work, along with ome uggetion for future work. 1.3 Contribution The maor contribution of thi thei i a deep theoretical analyi of the hort-circuit current delivered by wind turbine with DFIG with both crowbar and chopper protection under ymmetrical and unymmetrical fault. A general analytical treatment of the hort-circuit behavior of uch wind turbine wa lacking in the literature. Moreover, an invetigation on the voltage dip profile reulting from fault below tranmiion level and on the voltage upport capability of wind turbine in weak network ha been performed. A method for calculating the ymmetrical and unymmetrical hort-circuit current of a wind turbine equipped with SCIG or DFIG with crowbar protection i developed. The method reult in an analytical expreion that allow the calculation of the wind turbine hort-circuit current without the need for dynamic imulation but with an accuracy compared to that obtained by uch imulation. The method i applicable during a fault, under the period the rotor ide converter witching i blocked and the crowbar reitance i inerted. The analytic expreion can alo be ued to calculate the maximum phae current for different fault, their DC and AC component, or it may be ued to get an envelope of the current if only their RMS value i needed. The preented method can be een a analogou to the well known three-phae hort-circuit formula of a ynchronou generator (Roeper 1985). Modern DFIG wind turbine may be protected with a chopper reitance on the DC-link, intead of the crowbar reitance. An analyi of the hort-circuit behavior of thi type of wind turbine i preented. The idea to model the DClink capacitor and chopper a an equivalent reitance een from the rotor circuit wa introduced in (Martinez 11(a)). Thi implie that method developed for hort-circuit current calculation for a DFIG with crowbar reitance are alo applicable to a wind turbine with DC chopper. In thi thei it i hown that the value of the equivalent crowbar reitance mut be properly choen, depending on the voltage dip and the initial loading of the

21 8 Chapter 1. Introduction wind turbine, for accurate hort-circuit current prediction. The propoed method i particularly ueful conidering that model of wind turbine with DC chopper are not commonly included in tandard imulation tool. The total hort-circuit current contribution of a wind farm ha been invetigated through imulation in order to ae the validity of an aggregate model for hort-circuit current tudie. It i found that aggregate model are ufficiently accurate both in the cae of wind farm with SCIG and with DFIG. Thi i valid for ymmetrical and unymmetrical fault. The invetigation on the voltage dip profile for fault at the ub-tranmiion network how that only grid bue electrically cloe to the fault may experience voltage dip that may compromie grid fault ride-through of wind turbine. It i hown how the modelling of the load in the ytem i eential for the evaluation of the reulting voltage dip and grid fault ride-through. The conidered cae tudie how that wind turbine grid fault ride-through i endangered mainly for fault cleared by the circuit breaker failure protection and under high motor load condition. The grid fault ride-through behavior of wind turbine connected to a weak network ha been invetigated. Wind turbine can actively contribute to improve the voltage during and immediately after a fault in the grid. However, the baic grid code requirement do not alway reult in atifactory wind turbine voltage upport in weak network. A high proportional gain for current inection depending on the voltage dip magnitude and continuou voltage regulation without deadband are two way to improve voltage upport. In particular, in weak network with low X/R ratio of the equivalent grid Thevenin impedance, a coordinated inection of active and reactive current along with fat active power pot-fault ramp help improving the voltage during the fault and the pot-fault voltage recovery. A tudy cae uing a implified model of the Bornholm power ytem how how the voltage upport provided by wind turbine i comparable to the one provided by ynchronou generator, if the proper voltage upport trategy i choen. The thei preent a wide analyi of the fault behavior of wind turbine with SCIG and DFIG. The expectation i that the analyi will help undertanding and predicting the hort-circuit behavior of thee type of wind turbine. 1.4 Publication Publication related to thi Ph.D. thei:

22 1.4. Publication 9 [1] Sulla F., Svenon J., Samuelon O., Symmetrical and Unymmetrical Short-Circuit Current of Squirrel-Cage and Doubly-Fed Induction Generator, Electric Power Sytem Reearch, vol. 81, No. 7, 11. [] Sulla F., Svenon J., Samuelon O., Fault Behavior of Wind Farm with Fixed-Speed and Doubly-Fed Induction Generator, PowerTech 11 conference, Trondheim, 19-3 uni 11. [3] Sulla F., Svenon J., Samuelon O., Short-Circuit Analyi of a Doubly Fed Induction Generator Wind Turbine with Direct Current Chopper Protection, Wind Energy, firt publihed online 9 December 11. Other publication: [a] Börntedt J., Sulla F., Samuelon O., Experimental Invetigation on Steady-State and Tranient Performance of a Self-Excited Induction Generator, IET Generation, Tranmiion and Ditribution, vol. 5, No. 1, 11. [b] Sulla F., Börntedt J., Samuelon O., Ditributed Generation with Voltage Control Capability in the Low Voltage Network, International Conference on Renewable Energie and Power Quality (ICREPQ 1), Granada, Spain, March 1. [c] Sulla F., Samuelon O., Short-Circuit Analyi and Protection of a Medium Voltage Network in Iland Operation with Induction Generation, Eight Nordic Ditribution and Aet Management Conference (NORDAC8), Bergen, Norway, September 8. [d] Sulla F., Samuelon O., Analyi of Iland-Operated Ditribution Network with Ditributed Induction Generation under Fault Condition, 43rd International Univeritie Power Engineering Conference (UPEC8), Padova, Italy, September 8. [e] Sulla F., Samuelon O., Etimation of the Zero Sequence Voltage on the D-ide of a Dy Tranformer by Uing One Voltage Tranformer on the D-ide, 9th International Conference on Development in Power Sytem Protection (DPSP8), Glagow, UK, March 8.

23

24 Chapter Wind Turbine during Fault The requirement on grid fault ride-through (GFRT) capability have puhed a rapid evolution of wind turbine technology. Remaining connected to the grid throughout the duration of the fault, wind turbine will feed hortcircuit current into the grid, affecting power ytem planning iue uch a component izing and relay protection etting that have traditionally been baed on the hort-circuit current contribution of ynchronou generator. Thi chapter tart with preenting the different wind turbine concept mot commonly intalled in the power ytem. A review of the literature on the fault current of wind turbine i then performed. Finally, requirement on GFRT from ome national grid code are introduced along with protection and control cheme devied to allow wind turbine GFRT capability..1 Wind Turbine Concept Commercial wind turbine can be claified into four maor type (Hanen 4), depending on their ability to operate at fixed- or variable-peed and on the ize, partial- or full-cale, of the power electronic converter, when ued. Fixed-peed wind turbine were the mot common type of wind turbine intalled until the ninetie and there are till many in operation. In year they till repreented 39 % of the total intalled wind turbine in the world (Hanen 4). Though being conceptually imple and cheap, they preent many diadvantage that led to a trong decreae of their market hare during the lat decade. Among other diadvantage, fixed-peed wind turbine deliver a power to the grid which follow wind peed fluctuation, with flicker emiion and high mechanical tree on the drive train a a conequence (Hanen 4). Thee drawback are eliminated in a limited variable-peed wind turbine. Such wind turbine are able of aborbing the wind power fluctuation and convert them to kinetic energy and heat into an external 11

25 1 Chapter. Wind Turbine during Fault rotor reitance, hence uppreing the grid power fluctuation (Akhmatov 5). However, urvey how that market hare of thi wind turbine concept ha teadily decreaed to a quite inignificant level after the ninetie (Li 7). The vat maority of wind turbine intalled nowaday i of variable-peed type. They are able to vary the haft peed over a wide range and perform optimal power extraction from the wind, at low wind peed, thu increaing the total yearly energy production (Peteron 5(a)). Power extraction optimization i done by keeping the tip peed ratio cloe to it optimal value (Ackermann 5). Moreover, they poe uperior control poibilitie, enabling grid ancillary ervice uch a reactive power control, frequency and voltage upport after grid diturbance. Modern variable-peed wind turbine alo have GFRT capability. Fixed-Speed Wind Turbine, Type A A fixed-peed wind turbine i equipped with a SCIG directly connected to the grid via a tep-up tranformer, a hown in Figure.1. Due to the teep torque-lip characteritic of a SCIG, it will operate at almot contant peed, with a lip in the order of % at rated power (Akhmatov 5). The peed of the SCIG i therefore et by the grid frequency. Figure.1 Fixed-peed wind turbine with SCIG. Fixed-peed wind turbine may be tall, pitch or active-tall controlled. To avoid large current tranient during grid connection, fixed-peed wind turbine may be connected to the grid through a oft-tarter (Hanen 4). Due to the tiff connection to the grid, a fixed-peed wind turbine cannot aborb the wind power fluctuation in it rotor kinetic energy. A a conequence, fluctuation in the torque and delivered power will reult. Torque fluctuation may caue mechanical tree to the turbine drive train, while power fluctuation may reult in voltage flicker. A SCIG need to be magnetized through the network and to decreae the reactive power drawn from the grid in normal operation, capacitor bank are intalled at the SCIG terminal. However in cae of a voltage dip, the wind

26 .1. Wind Turbine Concept 13 turbine accelerate and draw large amount of reactive power from the grid, well above it need in normal operating condition. Thi complicate voltage recovery and could lead to generator over-peeding and conequent diconnection (Ahkmatov 5) to avoid voltage tability iue. Limited Variable-Speed Wind Turbine, Type B Thi wind turbine concept ue a wound rotor induction generator (WRIG) and an external reitance that can be connected to the generator rotor through a converter. By controlling the converter, the value of the effective external reitance can be controlled. A cheme of thi wind turbine concept i hown in Figure.. When increaing the total rotor reitance, the maximum torque of the lip-torque curve of the induction generator i hifted toward higher lip and higher generator peed. Conequently, if the mechanical power i aumed contant, thi reult in operation at higher peed. Figure. Limited variable-peed wind turbine with wound-rotor induction generator and variable rotor reitance. Limited variable-peed wind turbine can be operated above rated peed, with a maximum lip range above 1 %. The peed range i limited by the high heat loe in the external reitance (Burnham 9). Thi control i mainly ued to aborb wind power fluctuation into kinetic energy of the wind turbine haft and then diipate them into heat in the external reitance (Akhmatov 5). A maor drawback of thi wind turbine concept i that active and reactive power cannot be independently controlled (Burnham 9). Jut like the fixed-peed concept, thi wind turbine draw reactive power from the grid and therefore capacitor bank are ued to improve the power factor (Akhmatov 5).

27 14 Chapter. Wind Turbine during Fault Variable-Speed Wind Turbine with DFIG, Type C Thi type of wind turbine ue a DFIG. The tator of the generator i directly connected to the grid. The rotor i alo connected to the grid, but through a back-to-back converter. A three winding tep-up tranformer may be ued, a hown in Figure.3. The converter on the rotor ide will be referred throughout thi thei a the rotor ide converter (RSC), while the converter on the grid ide a the grid ide converter (GSC). Figure.3 Variable-peed wind turbine with DFIG. The power flow through the rotor can be bidirectional (Akhmatov 5). The power flow into the rotor when the wind turbine operate at ubynchronou peed, with low mechanical input power. The rotor power flow revere at uper-ynchronou peed. Thu, with high mechanical input power, part of thi power i fed to the grid through the tator and part through the back-to-back converter. The RSC i normally et to control the active and reactive power inection into the grid through the tator. Active and reactive power can be controlled independently by aduting the external voltage applied to the rotor. By properly aduting the external rotor voltage, the tator current can be controlled alo to deliver a reactive power to the grid. Thi i a maor advantage of thi wind turbine type over type A and B. The fat control of the RSC make it poible to feed a mooth active power into the grid. The RSC can be controlled to optimize the wind power extraction at low wind peed (Akhmatov 5). During normal operation, the GSC control the DC-link capacitor voltage and uually doe not contribute to any reactive power exchange with the grid. A maor advantage of thi type of wind turbine i that the back-to-back

28 .1. Wind Turbine Concept 15 converter only need to be ized to handle approximately 3 % of the generator rated power (Peteron 5(a)). Thi i a direct conequence of the fact that the power flowing through the rotor i given by the product of the tator power and the lip and that thee wind turbine are operated within a lip range from -.3 to +.3. It i alo important that the rotor-to-tator turn ratio be choen properly to reduce the current rating of the converter. If thi turn ratio i choen around 3, the maximum rotor voltage in normal operating condition will reult cloe to 1 pu. Conequently the rotor current will not exceed.3 pu of the tator current. The maor drawback of thi wind turbine concept i that it i very enitive to grid diturbance. A dip in the voltage may in fact caue high current in the rotor that may damage the RSC. In thee ituation the witching of the RSC i therefore blocked. There are different cheme to protect a DFIG wind turbine during fault and at the ame time allow grid fault ride-through capability. Some of thee cheme are expoed in Section. and.4. The hort-circuit current delivered by DFIG wind turbine during a fault are widely dicued in Chapter 3 and 4. Variable-Speed Wind Turbine with FSC, Type D Thi wind turbine concept i connected to the grid through a full-cale backto-back converter. The generator may be either of ynchronou or induction type. In the firt cae, both option with a eparately excited and a permanent magnet ynchronou generator are available in commercial wind turbine (Manwell 9). Gearle olution uing a ynchronou generator with a high number of pole are alo preent in the market (Hanen 4). If an induction generator i ued, thi need to be magnetized by the generator ide converter. Therefore, thi ha to be rated to handle not only the rated active power of the generator but alo it reactive power. Reactive power can be provided alo by capacitor intalled at the generator terminal (Akhmatov 5). The generator ide converter control the peed of the generator to optimize power extraction from the wind. Thee wind turbine have a wider operating peed range than wind turbine with DFIG (Tili 8). The grid ide converter control the DC-link capacitor voltage feeding active power into the grid. It can alo independently control the reactive power inection.

29 16 Chapter. Wind Turbine during Fault Figure.4 Variable-peed wind turbine with full-cale converter. The generator may be either of ynchronou type, eparately excited or permanent-magnet, or an induction generator. Wind turbine with FSC are eaier to control during voltage dip in the grid, a compared to DFIG wind turbine. In thi cae, the voltage dip doe not directly caue any tranient in the generator. The main iue i the rie of the DC-link capacitor voltage when no active power can be delivered to the grid. A number of trategie to mitigate and olve thi problem are preented in Section.4. The grid ide converter can alo provide fat reactive power upport.. Wind Turbine Fault Current Different type of wind turbine will deliver different fault current into the grid in the event of a fault. A brief review on the hort-circuit current provided by different wind turbine type i preented below. Fixed-Speed Wind Turbine The hort-circuit behavior of fixed-peed wind turbine with a SCIG i determined by the dynamic of the generator itelf. High hort-circuit current will be delivered during the fault, decaying with the fluxe preent in the generator tator and rotor. The hort-circuit behavior of a SCIG ha been analyzed in variou text book and in many paper. Cloed formula for the hort-circuit current calculation are given in the cae of a bolted three-phae hort-circuit at the generator terminal in (Kovak 1984, Sarma 1986, Va 199, Jenkin ). The ymmetrical hort-circuit current i made up of two component, an AC and a DC component that decay repectively with the rotor and tator tranient time contant. Thee formula have been deduced baed on the aumption that the rotor and tator reitance are negligible. The effect of the reitance i then accounted for in the tranient time contant. The hort-circuit current i limited by the tranient reactance, and it maximum value uually varie between 5 and 9 time the generator rated current.

30 .. Wind Turbine Fault Current 17 The analyi of the hort-circuit behavior of a SCIG during unymmetrical fault i uually performed by mean of digital imulation (Chen 1991, Samaan 8). Limited Variable-Speed Wind Turbine The ymmetrical hort-circuit current of a limited-variable peed wind turbine with WRIG i hown in (Martinez 11(b)). The high external rotor reitance caue the AC component of the tator current to decay rapidly. In thi repect the hort-circuit current of thee wind turbine reemble the hort-circuit current of DFIG wind turbine. Variable-Speed Wind Turbine with DFIG At the occurrence of a fault in the network, high tator and rotor current are induced in a DFIG. Theoretically, the RSC could be over-ized to handle the hort-circuit current in the rotor but thi would increae the overall cot for DFIG wind turbine and reduce the main advantage for DFIG ytem compared to wind turbine connected to the network through a FSC. Therefore, the RSC witche mut be blocked to avoid their damage. After blocking the RSC, there are different method in which a DFIG wind turbine can be protected. A tandard olution i blocking the RSC and at the ame time hort-circuiting the rotor circuit through an external reitance, called crowbar reitance (Akhmatov 5, Morren 7), ee Figure.3. Modern DFIG wind turbine can alo be protected with only a DC chopper (Martinez 11(a)), hown in Figure.3. At the event of a hort-circuit, the RSC witche are blocked and the rotor current i led into the DC-link capacitor through the anti-parallel diode of the RSC. A DC chopper i inerted to regulate the DC capacitor voltage. In thi way the rotor circuit i rapidly demagnetized and the RSC can be re-tarted when the rotor current and DC-link voltage decreae below a certain value. For thi configuration of the DFIG wind turbine, the anti-parallel diode of the RSC mut be overized to handle the hort-circuit current. Both for a DFIG wind turbine with crowbar and with DC chopper protection, the fault current immediately after the fault i olely determined by the phyical ytem and not by a pecific control trategy. Thi allow carrying out analytical tudie and drawing general concluion on the delivered hort-circuit current. The hort-circuit behavior of a DFIG wind turbine with crowbar protection depend on whether or not the crowbar i activated. During the time of

31 18 Chapter. Wind Turbine during Fault crowbar activation the control of the generator current i lot and the DFIG may be regarded a a SCIG, but with high rotor reitance, up to time the value of the generator rotor reitance (Akhmatov 5), and poibly high lip. Many paper have been publihed on the hort-circuit current contribution of a DFIG but they confine their analyi to a olid ymmetrical three-phae hort-circuit at the generator terminal (Vicato 1991, Morren 7, Yang 9, Pannell 1). (Yang 8, Rahimi 1) analyze voltage dip of any magnitude but focuing on control iue and not on fault current contribution. The high crowbar reitance caue the AC component of the ymmetrical DFIG hort-circuit current to decay much more rapidly a compared to a SCIG. After ome period, the hort-circuit current of a DFIG i made up predominantly of a DC component. In (Morren 7) it ha been propoed to calculate the ymmetrical hort-circuit current of a DFIG in the ame manner a done for a SCIG, but incrementing the value of the rotor reitance in the expreion for the rotor tranient time contant by the value of the crowbar reitance. Moreover, it ha been uggeted that the hortcircuit current hould be limited not only by the tranient reactance, but by the um of thi with the crowbar reitance. A it will be hown in Chapter 3, thee aumption are till not enough for accurate DFIG hort-circuit current calculation. In (Vicato 1991, Pannell 1), the ymmetrical hortcircuit current of a DFIG i obtained with the help of Laplace tranformation. However, the analyi in (Vicato 1991) i valid only under the aumption of mall rotor reitance, which i not a valid aumption for a DFIG with crowbar. In (Pannell 1) it i pointed out how in a DFIG, the frequencie of the decay component of the hort-circuit current deviate from pure DC and rotor peed due to magnetic drag effect originating from the interaction between the tator and rotor fluxe. In thee reference the analyi i confined only to three-phae bolted hort-circuit and, moreover, phyical comprehenion of the hort-circuit proce i not alway traightforward from the pure mathematical treatment. When the RSC i reconnected it control the tator current and the DFIG may be looked at a a contant current ource. The DFIG hort-circuit behavior become then imilar to that of a FSC (Walling 9). The fault current delivered by a DFIG wind turbine with DC chopper protection i eldom analyzed in the literature. Meaurement under network fault are hown in (Engelhardt 9), but a method to calculate the fault current i not provided in that reference. In (Martinez 11(a)), it ha been propoed a theoretical analyi of the behavior of a DFIG with chopper

32 .3. Grid Fault Ride-Through Requirement 19 protection during ymmetrical fault. The analyi reult in the parameterization of a ynchronou generator to repreent a DFIG with chopper protection. The analyi propoed in (Martinez 11(a)) i baed on the aumption that during a ymmetrical fault the DC-link capacitor and chopper can be repreented imply a an equivalent reitance, whoe value i choen to be the reitance value of the chopper. In Chapter 4, it will be hown that for accurate hort-circuit current prediction the value of the equivalent reitance ha to be choen depending on the initial loading of the generator and on the voltage dip magnitude. An analyi of the DFIG behavior under unymmetrical voltage dip i performed in (Lopez 8(a)), though the author focu not on the hortcircuit current but intead on the rotor voltage of the DFIG under uch condition for control purpoe. Unymmetrical fault caue high rotor voltage in a DFIG, caued by the negative equence in the grid. Thee voltage may be well above the control range of the RSC and, a a conequence, controllability of rotor current i lot. Moreover, ince the negative equence voltage doe not decay, the crowbar mut remain inerted and the RSC blocked for the whole duration of the fault (Seman 6(b)). Variable-Speed Wind Turbine with FSC The hort-circuit behavior of FSC wind turbine i mainly determined by the way the grid ide converter i controlled and it i therefore pecific to each particular commercial wind turbine (Walling 9). However, in general, it can be aid that the FSC limit the current fed into the fault to the nominal current rating of the converter or lightly above it. For evere fault, the witching of the converter may alo be topped during the fault period (Martinez 11(b))..3 Grid Fault Ride-Through Requirement The baic way national grid code pecify the limiting condition for a wind turbine to remain connected to the grid during a voltage dip, i by providing a curve where a voltage dip i given a a function of time. Voltage dip following a fault in the ytem and cleared under normal circumtance are expected to be above thi curve. The duration of the voltage dip i dependent on the peed with which the protection ytem clear the fault. Since Tranmiion Sytem Operator (TSO) in different countrie may have different protection philoophie, it follow that GFRT requirement are pecific to each national grid code. The voltage-time curve from ome European national grid code are hown in Figure.5.

33 Chapter. Wind Turbine during Fault 1.8 Voltage (pu).6.4 Sweden. EON, Germany Danmark Spain time () Figure.5 Voltage-time curve from GFRT requirement of ome European national grid code. A proce toward the harmonization of the European grid code ha tarted when Regulation (EC) 714/9 aigned ENTSO-E, European Network of Tranmiion Sytem Operator for Electricity, the role of developing the Network Code for Requirement for Grid Connection applicable to all Generator. A wind turbine i required to remain connected to the grid if the voltage at the Point of Common Coupling (PCC) during a fault alway remain above the voltage-time curve given in the grid code. Note that in the cae of the Swedih grid code, the voltage-time curve i applicable in the tranmiion network and, in general, not at the PCC. For wind turbine connected directly to the tranmiion network the Swedih GFRT curve i applicable at the PCC. Wind turbine connected below the tranmiion network mut ride-through fault in the tranmiion network cauing voltage dip at the connecting tranmiion network bue above the Swedih GFRT curve. In ome cae, diconnection i allowed if followed by a fat re-ynchronization within a few econd, e.g. econd in E.ON grid code (E.ON 6). Otherwie, the wind turbine i allowed to diconnect from the grid. Some grid code may have an explicit requirement on recurring fault, i.e. a fault followed within a hort period of time by another fault of the ame type. In the cae of the Danih grid code (Technical Regulation ), uch a requirement exit for two ingle-phae-to-earth or two-phae fault following

34 .3. Grid Fault Ride-Through Requirement 1 each other within a time of.5 up to 3. Remaining connected to the grid under the pecified condition i the baic GFRT requirement. Beide remaining connected, a wind turbine may alo be required to provide reactive current upport to the grid during the depreed voltage condition. The amount of reactive current to be inected in the grid i pecified in the grid code a a function of the voltage. E.ON grid code pecifie that in cae of a voltage dip, the reactive current inection mut increae with at leat % per every 1 % voltage drop, i.e. rated reactive current mut be delivered when the voltage fall below 5 % of rated value. Thi i combined with a deadband of +/- 1 % in voltage where the inection i zero. Voltage upport mut be continued for a further 5 m after the voltage return in the deadband. A imilar reactive upport cheme i alo preent in the Danih grid code, with the exception of the extra 5 m voltage upport. Spanih grid code only et ome limit on the conumed reactive power. In the cae of a three-phae fault, thi i allowed in the firt 4 m after the fault and in the firt 8 m after fault clearing and it cannot exceed 6 % of the nominal wind plant power rating (Leon-Martinez 11). Swedih grid code doe not have any requirement on reactive current inection during grid fault ride-through. A requirement on active power ramp after fault clearing i alo common in grid code (Tili 9). E.ON grid code require that wind turbine remaining connected during a fault have to recover their active power with a minimum gradient of % rated power per econd after fault clearance. Great Britain and Ireland grid code tate that active power mut be retored to at leat 9 % of it pre-fault value within 1 econd after voltage retoration (Tili 9). GFRT part of grid code imply a requirement not only for the technology, protection and control of the wind turbine, generator and power electronic, but alo on the deign and izing of the wind plant auxiliary equipment. The Danih grid code require for example that the hydraulic, pneumatic and emergency upply equipment in a wind plant mut be deigned o that continuou operation i aured for at leat ix different fault event at five minute interval. The GFRT requirement alo imply that the wind plant protection relay, e.g. the undervoltage relay, mut be et o that they do not affect the GFRT capability of the wind plant. GRFT requirement till repreent a challenge for turbine manufacturer. In

35 Chapter. Wind Turbine during Fault the next ection, a brief overview on different olution adopted to achieve GFRT capability i preented..4 Grid Fault Ride-Through Solution In thi ection, a review i preented on the different trategie propoed in cientific literature to achieve grid fault ride-though capability of wind turbine, with particular emphai given to the DFIG wind turbine. Fixed-Speed Wind Turbine A wind turbine with a SCIG will react to a fault in the grid by accelerating and drawing a large amount of reactive power when the grid voltage return. Depending on the unbalance between input mechanical power and output electric power during the fault, the generator may accelerate beyond the critical peed and become untable. In thi cae, diconnection and ue of an emergency brake are neceary (Akhmatov 5). The large amount of reactive power drawn by the SCIG after fault clearing may caue prolonged low voltage in the grid. Thi alo contribute to generator over-peeding and may lead to intability of the SCIG and conequent diconnection (Nguyen 1). Generator over-peeding and voltage recovery are therefore the two main iue for GFRT of thi type of wind turbine. To minimize generator overpeeding, pitch control i ued to reduce the mechanical input power. Becaue of the limited peed of the pitch ervo, thi olution may however not be enough. Serie connected dynamic braking reitor are propoed in (Cauebrook 7) to reduce generator over-peed and improve it tability. SVC or STATCOM are ued in (Foter 6, Molina 8) and hown to improve the pot-fault voltage recovery of fixed-peed wind turbine. The performance of a STATCOM i hown to be uperior to that of an SVC for improving tability of fixed-peed wind turbine. Controlling nearby variable-peed wind turbine to help GFRT of fixed-peed wind turbine ha been analyzed in (Luna 8, Muyeen 1). Variablepeed wind turbine can control the inection of reactive power during a fault and therefore upport the voltage retoration after fault clearing. Variable-Speed Wind Turbine with DFIG The tandard way to protect a DFIG wind turbine during a fault in the ytem i by blocking the RSC witching and inerting a crowbar reitance in the rotor circuit. Until ome year ago, the crowbar reitance wa inerted

36 .4. Grid Fault Ride-Through Solution 3 through a thyritor witch, which made it impoible to diconnect the crowbar reitance before the rotor current had decayed to zero. Thi configuration of crowbar protection i named paive crowbar. The DFIG wind turbine with paive crowbar i diconnected from the network ome time after the hort-circuit (Seman 6(a), Bak-Jenen 9). Since it need to be diconnected from the grid in the event of a fault, thi DFIG configuration doe not comply with the mot recent GFRT requirement. By replacing the thyritor with a fully controllable witch, a for example an IGBT, the crowbar reitance can be diconnected before the rotor current ha decayed to zero and even when the fault i till preent. Thi configuration i called active crowbar (Seman 6(b)). When the rotor current ha decayed below a certain value, the active crowbar may be diconnected and the RSC may be re-tarted. In thi way, the DFIG doe not need to be diconnected from the network but can ride through the fault contributing to active and reactive power inection after the crowbar ha been diconnected (Kayikci 8, Salle 9). When the crowbar i connected, pitch control i actuated to avoid over-peeding of the generator (Akhmatov 5). During the period of crowbar activation, the DFIG will deliver active power if in a uper-ynchronou condition or will aborb active power if in a ub-ynchronou condition (Kayikci 8). In both cae it will aborb reactive power from the network and ince it pre-fault lip may be far from zero the aborbed reactive power may be higher than that of a correponding SCIG running almot at almot zero lip (Akhmatov 5). The amount of aborbed reactive power may be reduced by properly chooing the value of the crowbar reitance (Hanen 7(a)). However, the crowbar alone may not be ufficient to achieve GFRT. After the fault, the DC-link voltage tend to increae due to the energy contribution from the rotor. A chopper reitance on the DC-link may be neceary to keep the DC-link capacitor voltage within an acceptable range (Erlich 7). A imilar olution i teted in (Engelhardt 9), where it i found that even for the mot eriou fault in the network the DFIG i able to ride-through continuing it upply of reactive power, thank to an advanced control cheme not fully decribed in the reference. The RSC would block only for olid three-phae fault at the generator terminal. Some other approache for the grid-fault ride through of a DFIG wind turbine propoed in the literature are dicued in the following. Thee approache partly rely on extra hardware and partly on advanced control cheme.

37 4 Chapter. Wind Turbine during Fault (Peteron 5(b)) propoe to dimenion the RSC to handle the high tranient rotor current during the hort-circuit. However, to avoid even higher tranient rotor current at fault clearing, it i propoed to diconnect the DFIG from the grid through anti-parallel thyritor, magnetize it through the RSC and re-connect it to the grid. (Yang 8) propoe the connection of a erie dynamic reitor in erie with the rotor circuit in the event of a fault. Thi reitor limit the rotor current a the crowbar reitor, but it doe not require the RSC blocking, ince it i connected in erie with it. (Xiang 6, Lopez 8(b)) point out how the high tranient rotor current are due to the natural tator flux, i.e. the flux trapped into the tator after a voltage dip. The author of the reference propoe to ue the RSC to quickly inect a current in the rotor counteracting the inducing tator natural flux, thu obtaining a rotor current decay fater than with the generator tator tranient time contant. (Eandi 9) propoe the imultaneou ue of the demagnetizing current trategy and an extra reitance to be inerted in the tator circuit after fault occurrence. A non-linear control law for determining the reference rotor current, along with a current controller made up of a PI controller and a reonant controller, i propoed in (Rahimi 1). (Lima 1) propoe a new control cheme for the rotor current controller baed on feeding back a reference to the controller the meaured tator current. In thi way the controller yntheize rotor current equal in hape but oppoite in phae with the tator current, reducing both the tator and rotor current. A propoal for inerting a converter in erie with the DFIG tator circuit i found in (Abdel-Baqi 1, Flannery 9). The erie converter i ued to inect a voltage at the generator terminal during ymmetrical and unymmetrical fault, to reduce the impact of the tranient on the generator. A direct power control (DPC) method combined with a trategy for crowbar triggering i propoed in (Zhou 8) and allow loing the DFIG controllability only for a hort time. In (Martinez De Alegria 4) a new control method for a DFIG, named Power Error Vector Control (PEVC), i briefly introduced. The author how that the PEVC may alo be ued to improve the ride-through capability of the DFIG during ymmetrical voltage dip, reducing the need of a crowbar only for very evere voltage dip. Variable-Speed Wind Turbine with FSC Being connected to the grid through a FSC, a voltage dip doe not affect directly the generator dynamic in thi type of wind turbine. The grid ide

38 .5. Summary 5 converter i able to ride-through and upply reactive current. A certain overpeeding of the generator i allowed to tore part of the mechanical input power into kinetic energy. Pitch control i adopted to reduce the input mechanical power. During evere fault however, a chopper reitance i needed to diipate the input power and keep the DC-link voltage within acceptable level (Tili 9, Nguyen 1). A non-linear control technique i propoed in (Mullane 5) to enhance the dynamic performance of the grid ide converter and avoid temporary overcurrent..5 Summary The mot common commercial wind turbine technologie and the reearch done on the hort-circuit behavior of thee wind turbine have been introduced. Particular emphai ha been given to the DFIG wind turbine, partly becaue it ha been the mot common wind turbine type intalled in the lat decade. The DFIG wind turbine fault behavior i alo the mot apt to be tudied analytically ince, during the period the RSC i blocked, it i not influenced by a particular control law but merely by it phyical ytem. Thi allow to draw general concluion on the delivered hort-circuit current. A brief review of the grid code requirement for GFRT from different countrie ha been done and olution adopted for different wind turbine type to achieve GFRT have been preented.

39

40 Chapter 3 Fault Current of SCIG and DFIG Wind Turbine SCIG wind turbine are till common in the power ytem and DFIG wind turbine have been the mot commonly intalled wind turbine type during the lat decade. The tandard olution to protect a DFIG wind turbine during a fault i to block the RSC witching and inert a crowbar reitance in the rotor circuit. Knowledge of their fault current contribution i important when dealing with protection relay etting or power ytem component izing. A thorough theoretical analyi of the fault current delivered by SCIG and DFIG wind turbine with crowbar protection i performed in thi chapter. The mathematical derivation i accompanied by a imple phyical explanation of the proce going on inide the generator during a grid fault. The material preented in thi chapter i baed on Publication [1]. 3.1 Introduction Induction generator have a different hort-circuit behavior when compared to ynchronou generator and prediction of thi behavior i an important iue in power ytem planning, tranient tability analyi and protection etting tudie. In thi chapter a general method for calculating the hort-circuit current of a DFIG with crowbar reitance i propoed. For pedagogical purpoe, the hort-circuit current of a SCIG i calculated firt. The method deal with both ymmetrical and unymmetrical fault and voltage dip of any magnitude at the generator terminal. It will be hown that the 7

41 8 Chapter 3. Fault Current of SCIG and DFIG Wind Turbine implification done to derive the hort-circuit current of a SCIG neglect ome dynamic that become important when the rotor reitance i high, a i the cae for a DFIG with crowbar protection. Therefore applying a hortcircuit formula derived for a SCIG to predict the hort-circuit current of a DFIG lead to inaccurate reult. The propoed method permit to predict the hort-circuit current a a function of time in the three phae of a SCIG or a DFIG by uing an analytic expreion, eliminating the need for dynamic imulation. The analytic expreion can alo be ued to calculate the maximum phae current for different fault, their DC and AC component, or it may be ued to get an envelope of the current if only their RMS value i needed. The reult obtained with the propoed method are compared to thoe obtained through imulation uing the well-etablihed claic linear fifth order model of the induction machine (Va 199). A main aumption made in thi work i to neglect aturation of the induction machine reactance. Thi i certainly an approximation and will impact on the accuracy of the propoed method when compared to meaurement on a real induction machine. Saturation of leakage reactance may increae the hort-circuit current of an induction generator while mainflux aturation ha little impact on the hort-circuit current (Jabr 7). A imple method to take into account the aturation of the leakage reactance i alo expoed in (Jabr 7) and could poibly be applied to the method propoed in thi chapter. However, thi i not done here ince it i not within the primary cope of thi work to invetigate the effect of aturation on the hort-circuit current of an induction generator. 3. Induction Generator Modelling The induction generator dynamic equation are here preented in a tator reference frame uing pace-vector notation, a done in (Va 199). A threephae voltage ytem may be expreed, with obviou meaning of the notation, a in Equation 3.1. v v v a b c t V ˆ cot t V ˆ co t 3 t V ˆ co t 4 3 (3.1) The correponding pace-vector i calculated in Equation 3.. Notice that the

42 3.. Induction Generator Modelling 9 amplitude of the defined voltage pace-vector i equal to the peak amplitude of the intantaneou voltage and that V i a phaor: v t t Ve ˆ e Ve a b t v t av t a v t 3 c (3.) where e 3 a, a 3 e, V Ve ˆ. The pace-vector are here indicated by an overlined arrow. The phaor V i defined in uch a way that it magnitude i equal to the peak-value of the voltage. The firt part of Equation 3. i valid alo if the three-phae quantitie do not form a balanced ytem. In thi cae, under the aumption that no zero-equence i preent, the pace vector become (Va 199): v t t t t Vˆ 1 e e Vˆ 1 e e V1e V e t (3.3) Similar expreion can be obtained for current and fluxe. The zeroequence i not conidered here, ince commonly an induction generator i not grounded and therefore no zero-equence current can flow. If no zeroequence component i preent, the intantaneou value of the current in the three phae can be obtained from the correponding pace-vector a (Va 199): i i i a b c t Re i t Rea i t Reai (3.4) Uing the introduced pace-vector notation and uing a tationary reference frame, the equation decribing the electrical dynamic of a quirrel-cage induction machine are given by Equation 3.5 and Equation 3.6 (Va 199). d v Ri dt d r Rrir r r dt Li L i L i L i r m m r r r (3.5) (3.6)

43 3 Chapter 3. Fault Current of SCIG and DFIG Wind Turbine where L = L l + L m and L r = L rl + L m. 3.3 SCIG Short-Circuit Current The hort-circuit current will be calculated under the following aumption: a) before the occurrence of the fault balanced condition are preent in the network; b) the generator i aumed to run at no-load and to be lole, which reult in zero lip; c) the generator peed doe not change after the hort-circuit; d) the generator i connected to a trong network, o that the pot-fault voltage i not influenced by it hort-circuit current e) the generator i linear, i.e. it doe not aturate. Linearity of the induction generator allow to ue the uperpoition principle to find out the total hort-circuit current a um of different component. The above aumption mean that the hortcircuit current calculated with the method expoed in thi chapter i an approximation of the hort-circuit current delivered by a real machine. The hort-circuit current of a SCIG i made up of three component. The firt component i due to the pot-fault teady tate voltage at the generator terminal which may be ymmetrical a well a unymmetrical. We will refer to thi component a the forced hort-circuit current component. The econd component i due to the natural tator flux and the third component i due to the natural rotor flux. The natural tator and rotor fluxe arie ut after the fault to aure the continuity of the tator and rotor fluxe before and after fault inception, according to the contant flux linkage theorem (Kimbark 1968). In a SCIG, thee component decay exponentially with time contant that depend upon the generator parameter (Kovak 1984). The term forced and natural are ued, a done in (Lopez 8(a)). Once the pot-fault tranient tator and rotor fluxe are known, the hortcircuit current can be calculated by olving Equation 3.6 with repect to the tator current. Pot-fault Stator Flux The tator flux i made up of a forced component and of a natural component. In turn, the forced component i due to the contribution of the poitive and the negative equence of the pot-fault voltage. Therefore the tator flux after fault occurrence can be expreed a in Equation 3.7. t t t t f, 1 f, n (3.7)

44 3.3. SCIG Short-Circuit Current 31 The pre-fault flux and the pot-fault poitive and negative equence forced flux component in a tator reference frame can be derived from Equation 3.5 neglecting the tator reitance: t pre t pre pre pre pre e V e e V v ˆ, (3.8) t t f e V e V e v 1 1 1,1 1 ˆ (3.9) t t f e V e e V v, ˆ (3.1) Negative equence quantitie rotate with a frequency of - in a tator reference frame. Next, the natural component of the tator flux mut be found. In general, for any kind of fault, becaue of the contant flux linkage theorem, the natural flux ut after the fault i given by the difference between the tator forced flux component immediately before and after the fault, a in Equation The natural tator flux i actually not contant becaue of the preence of the tator reitance, but decay with a time contant given by Equation 3.1, (Kovak 1984, Va 199, Jenkin ), which i valid under the aumption that the rotor reitance i mall, a it will be hown later. By combining Equation and taking into account that the natural tator flux decay with time contant T, the total tator flux after fault occurrence i given by Equation pre f f pre n n V V V 1,,1, (3.11) r m R L L L R L T ' (3.1) T t pre t t e V V V e V e V t 1 1 (3.13)

45 3 Chapter 3. Fault Current of SCIG and DFIG Wind Turbine Pot-fault Rotor Flux The pot-fault rotor flux i the um of three component: r t t t t rf, 1 rf, rn (3.14) With the aumption of zero lip, the poitive equence rotor current i zero, ee Figure 3.1, and the pre- and pot-fault poitive equence rotor fluxe are given in Equation 3.15 (Morren 7), where k =L m /L. r, pre rf,1 Lm L Lm L, pre f,1 k k f,1, pre k Vpre k e V1 e t t (3.15) Figure 3.1 Poitive and negative equence equivalent circuit of a SCIG. Thee are alo applicable to a DFIG with rotor winding connected to a crowbar reitance, if R r i replaced by R r +R cr. The negative equence forced rotor flux can be found by firt expreing the negative equence tator and rotor current. With reference to Figure 3.1, the tator negative equence current i given by Equation 3.16, where both the tator and rotor reitance have been neglected. i, v Z v L ml Ll Lr rl (3.16)

46 3.3. SCIG Short-Circuit Current 33 The minu ign in front of the reactance i due to the fact that a negative equence voltage induce a flux whoe direction of rotation i oppoite to that of a poitive equence induced flux. Thi i alo the reaon why the rotor reitance i divided by the negative equence lip - in Figure 3.1 (Anderon 1995). The negative equence rotor current can be found by a imple current diviion between the rotor and the magnetizing branch: i L m m r, i, i, Lm Lrl Lr L (3.17) Finally, by inerting the negative equence tator and rotor current into Equation 3.6 lead to Equation L i L i (3.18) rf, m, r r, Thi reult indicate that for a SCIG the negative equence forced rotor flux can be neglected. The natural rotor flux i the flux trapped in the rotor circuit at fault occurrence. It magnitude and phae immediately after the fault are found a in Equation V pre rn r pre rf rf k V,,1, 1 rn (3.19) Thi flux i fixed with the rotor circuit, i.e. it rotate with the rotor peed in a tator reference frame (Morren 7). In a rotor reference frame, it i a DC component decaying exponentially with time contant T r, defined in Equation 3. (Morren 7). The expreion for the inductance in Equation 3.1 and 3. can eaily be derived by conidering the induction generator equivalent circuit in Figure 3.1 or uing Equation 3.5 and 3.6. See alo (Va 199, Morren 7). T Lm ' Lr Lr L r (3.) Rr Rr Summarizing what ha been found in thi ection, the tranient rotor flux for a SCIG in a tator reference frame i given a

47 34 Chapter 3. Fault Current of SCIG and DFIG Wind Turbine r r T t t pre t r e e V V k e V k t 1 1 (3.1) Short-Circuit Current of a SCIG The pot-fault tator and rotor fluxe are given by Equation 3.13 and 3.1. Solving Equation 3.6, the relation between fluxe and tator current i given by Equation 3., (Va 199, Morren 7), where k r =L m /L r. ' ' ' ' r r r r m L k L L L L L i (3.) Subtitution of Equation 3.13 and 3.1 into 3. give the final general approximate expreion for the hort-circuit current of a SCIG, reported in Equation 3.3. r r T t t pre r T t pre t t r e e L V V k k e L V V V e L V e k k L V i ' 1 ' 1 ' ' 1 1 (3.3) Thi equation i valid under any ymmetrical or unymmetrical fault in the network. If the network and tep-up tranformer inductance cannot be neglected they need to be added in erie with the generator tator leakage inductance in all the previou equation. To get the current in the three phae of the induction generator it uffice now to apply Equation SCIG Simulation Reult The network of Figure 3. ha been modelled in MATLAB SimPowerSytem (MATLAB R9b). The induction generator i conidered linear, without aturation, and it i connected to the network directly, without a tep-up tranformer. The network voltage i 575 V and the pu generator parameter are reported in Table 3.1. To comply with the aumption made above of contant rotor peed during the fault, a high inertia contant ha been defined for the generator. The generator initial lip i cloe to zero and the generator remain unloaded during the fault. A timetep of 5 microecond ha been ued in all imulation. The network reactance X th i aumed equal to one hundredth of the generator bae impedance and the ratio X th /R th i aumed equal to 1. Different kind of fault have been imulated with different value for the parameter p.

48 3.4. SCIG Simulation Reult 35 Figure 3. Network diagram ued for the imulation. By varying the parameter p, the voltage dip magnitude at the generator terminal during the fault can be changed. Table 3.1 Induction generator parameter SN (MVA) 1717 VN (V) 575 R (pu).73 Rr (pu).5 Ll (pu).1766 Lrl (pu).161 Lm (pu) 3 Simulation reult how that the propoed method for calculating the hortcircuit current give accurate prediction of all phae current under both ymmetrical and unymmetrical fault in the network. Figure 3.3 and Figure 3.4 how the calculated and imulated hort-circuit current in the cae of a three-phae and of a phae-b to phae-c hort-circuit with two different value of the parameter p. The calculated and imulated current are practically inditinguihable.

49 36 Chapter 3. Fault Current of SCIG and DFIG Wind Turbine time () Figure 3.3 Simulated (blue-olid) and calculated (red-dahed) SCIG hort-circuit current for a three-phae-phae fault, with parameter p=1. SCIG initially unloaded. Pre-fault voltage angle =9º. I a (pu) I b (pu) I c (pu) time () Figure 3.4 Simulated (blue-olid) and calculated (red-dahed) SCIG hort-circuit current for a phae-phae (b-c) fault, with parameter p=.85. SCIG initially unloaded. Pre-fault voltage angle =45º. I a (pu) I b (pu) I c (pu)

50 3.5. DFIG Short-Circuit Current DFIG Short-Circuit Current A DFIG uing crowbar protection i conidered in thi ection. The analyi aume that the crowbar remain connected during the whole duration of the fault, 5 m in thi tudy. Thi may not be the cae for three-phae fault ince the RSC would be re-tarted a oon a the rotor current decay below a certain predefined value. During ymmetrical fault, the analyi here preented i therefore applicable during the period between crowbar inertion and RSC re-tarting. However, for unymmetrical fault, the RSC will mot likely not be re-tarted during the fault ince the caue of high rotor current i the negative equence network voltage which doe not decay during the fault period (Semaan 6(b)). For the mot evere unymmetrical fault, the propoed analyi i therefore applicable during the whole duration of the fault. The method propoed above for calculating the hort-circuit current of a SCIG cannot be directly applied to a wind turbine driven DFIG, becaue of mainly two reaon. The firt reaon i that the value of the crowbar reitance may be up to time the value of the generator rotor reitance (Akhmatov 5) and the total reulting rotor reitance can no longer be neglected. In (Morren 7) it ha been propoed a method for calculating the maximum hort-circuit current of a DFIG with high crowbar reitance during a ymmetrical threephae fault at the generator terminal. The author of the mentioned reference propoed to include the effect of the crowbar reitance to calculate the maximum hort-circuit current of a DFIG in two tep. Firt, the rotor tranient time contant i modified according to T ' Lr r, DFIG Rr Rcr ' r L (3.4) R r, tot The econd tep to account for the preence of the high crowbar reitance propoed in (Morren 7) i to include it in the impedance limiting the hort-circuit current. Thu in Equation 3.3, one hould ue R r,tot + L intead of L, where R r,tot indicate the um of the rotor and crowbar reitance. However, thi proved to be till a too rough approximation when comparing with the imulation, leading to inaccurate calculation of the DFIG hort-circuit current a a function of time. The econd reaon, that make the SCIG hort-circuit current calculation

51 38 Chapter 3. Fault Current of SCIG and DFIG Wind Turbine method inaccurate when applied to a wind turbine driven DFIG, i that a wind turbine driven DFIG may operate in a lip range between -.3 and +.3. The forced tator and rotor fluxe, which for the SCIG have been calculated baed on the aumption of zero lip, for a DFIG hould be calculated baed on it initial rotor peed and delivered active and reactive power. Influence of high crowbar reitance on natural tator flux Let u tart with the tator tranient time contant. In paper dealing with the DFIG hort-circuit current (Morren 7, Yang 9), the DFIG tator tranient time contant i till aumed to be equal to the one in Equation 3.1. However, for a DFIG with high total rotor reitance, the tator tranient time contant need to be expreed in a lightly different way. The natural tator flux, which i fixed with repect to the tator, generate a voltage in the rotor whoe frequency and magnitude in a rotor reference frame are proportional to the rotor peed. A current will flow in the rotor, having the ame frequency of the induced voltage and oppoite to the rotor peed. The tator and rotor current due to the natural tator flux have the form Ie -rt e -t/t,dfig when expreed in a rotor reference frame. T,DFIG i the tator tranient time contant for a DFIG and i defined below. By deriving thi expreion and neglecting the term proportional to 1/T,DFIG, which for a typical induction machine i much maller than r, the voltage drop over an inductance L can be expreed a - r LIe -rt e -t/t,dfig. With reference to the equivalent circuit in Figure 3.1, a imple current diviion therefore till hold between the rotor and magnetizing branch and the rotor natural current in a rotor reference frame i: i rn r Lm i R L r, tot r r n (3.5) Subtituting in Equation 3.6, lead to 3.6. The term L n can be regarded a a complex operator that give the relation between the natural tator flux and current. A imilar concept, named operator inductivity, i introduced in (Kovak 1984) when dealing with the hort-circuit behavior of a ynchronou generator. rl m ' n L i n Lnin (3.6) Rr, tot rl r

52 3.5. DFIG Short-Circuit Current 39 L n i a modification of the inductance L in Equation 3.1 and coincide with it if the total rotor reitance i negligible. Therefore the tator tranient time contant of a DFIG with high rotor reitance i given in Equation 3.7. Thi mean that the natural tator flux i no longer fixed with repect to the tator, but it i actually lowly rotating becaue of the preence of a high rotor reitance. rl m L ' L R n r tot rl r T,, DFIG (3.7) R R Influence of high crowbar reitance on natural rotor flux Let u denote the natural rotor flux immediately after the fault a rn. In the SCIG cae, thi flux in a rotor reference frame i a DC component decaying with the rotor tranient time contant. Thi fact i no longer true for a DFIG with high rotor reitance. To explain why thi no longer hold, we may find it ueful to refer to a impler analogou ituation. Conider a hort-circuited coil with certain reitance R and inductance L. At a certain point, an alternating flux m (t) with frequency i induced in the coil. The coil may have an initial flux,. Thi ituation i analogou to that of a DFIG with a non negligible rotor reitance under a udden tranient. The coil correpond to the DFIG cloed rotor circuit, while the external flux m (t) would correpond to the natural tator flux of the DFIG. The external flux will induce a voltage in the coil and thi will reult in an alternating current component, oppoing the inducing flux: i ac t t t vm M me (3.8) R L R L M tell how much of the external flux i linked by the coil. A DC current component depending on the initial flux of the coil and decaying exponentially will alo flow in the coil: i dc L R t c coil e L t (3.9)

53 4 Chapter 3. Fault Current of SCIG and DFIG Wind Turbine The meaning of the contant c coil i that only a fraction of the initial flux i due to the DC current component i dc in the coil. The remaining part of the initial flux i due to the AC current component i ac and the external flux linkage with the coil. The total flux linkage with the coil will therefore be given by Equation 3.3. R L t L m ac dc m coil t t L R t M e Li t Li t 1 M e c e (3.3) Without the coil reitance, the AC current component in the coil would completely counteract the inducing flux, and only a DC flux would be preent in the coil. Thi ideal ituation i cloe to what happen in the hortcircuited rotor of a SCIG during a tranient, with the natural rotor flux being almot a pure DC component decaying with the rotor tranient time contant. However, when the coil reitance in not negligible, the induced AC current component doe not completely counteract the inducing flux. Thi i due to the fact that the AC current component magnitude decreae and that it acquire a phae difference with the inducing flux. A a reult the total coil flux will be compoed of an AC component and a DC component. The coefficient c coil tell how big the DC flux component in comparion to the total coil flux i. Thi ituation i analogou to what happen in the rotor circuit of a DFIG with high crowbar reitance. In thi cae the natural rotor flux can no longer be conidered a pure decaying DC component. To find out the value of the contant c coil, we can refer to the contant flux linkage theorem which ay that the coil flux cannot change intantaneouly. Therefore it mut hold that: L 1 M R L d c coil coil coil m c coil L M 1 R L 1 d m d coil c coil (3.31) Thee reult can now be tranlated to the cae of a DFIG with high crowbar reitance to find out the natural rotor flux a a function of time. In thi cae, the external inducing flux i the natural tator flux and the initial coil flux i the pot-fault natural rotor flux rn. Neglecting it low rotation a found in Equation 3.7, the natural tator flux induce in the rotor circuit a voltage

54 3.5. DFIG Short-Circuit Current 41 whoe frequency i proportional to the electrical rotor peed with the oppoite ign. The total equivalent inductance een from the rotor circuit i given by the erie connection of the rotor leakage inductance and the parallel connection between the magnetizing and the tator leakage inductance. By analogy with the coil example, we can conclude that the coefficient c and d for the DFIG are given a in Equation 3.3, where // denote the parallel operator. It ha been aumed that all the natural tator flux link the rotor, i.e. correponding to M in Equation 3.8 being equal to 1. L L L d r rl m // l 1 Rr tot r Lrl Lm // L c 1 d n, l rn (3.3) The coefficient c and d depend on the ratio between the initial natural tator and rotor fluxe and therefore vary for different fault condition. Similar reult can be derived by olving the DFIG differential equation in the cae of a three-phae hort-circuit at the generator terminal. The AC part of the natural rotor flux in a rotor reference frame, which depend on the inducing natural tator flux, will decay with the ame time contant T,DFIG a the natural tator flux. The DC natural rotor flux component, fixed with the rotor circuit, will decay with the rotor tranient time contant T r,dfig. We can now expre the natural rotor flux a a function of time in a tator reference frame: rn t de t T, DFIG ce t Tr, DFIG e rt rn (3.33) The initial natural rotor flux rn for a DFIG with high crowbar reitance will be calculated later in thi ection. The natural tator flux i not ignificantly affected by the rotor reitance value, therefore we can continue to ue the value calculated for a SCIG in Equation 3.11 alo in the cae of a DFIG with a high crowbar reitance. Influence of high crowbar reitance on negative equence fluxe The rotor negative equence current can be obtained with a imple current diviion between the magnetizing and the rotor circuit branche, a done for a SCIG:

55 4 Chapter 3. Fault Current of SCIG and DFIG Wind Turbine,,,, Z v L R L i L R L i r tot r m r tot r m r (3.34) where Z i the total negative equence impedance of the generator, which can be calculated from the equivalent circuit in Figure 3.1. By ubtituting into Equation 3.6, the rotor negative equence flux can be expreed a in 3.35.,,,,,, Z v L i L i L R L L L i L i L r r r tot r r m m r r m rf (3.35) L r give the relation between the rotor negative equence flux and the tator negative equence current and it i equal to zero, a expected, if the rotor reitance i zero. The negative equence tator flux i till given by Equation 3.1. However, proceeding a for the rotor flux, it can alo be expreed a in r tot r m f v i Z i L i L R L L,,,,, (3.36) Total rotor tranient flux The rotor forced flux mut now include the part due to the negative equence network voltage and it i given a: t r t rf e Z V L e V k 1 (3.37) Therefore the natural rotor flux, given by the difference of the pre- and potfault forced fluxe at t=, become 1, Z V L V k V k r pre rn rn (3.38)

56 3.5. DFIG Short-Circuit Current 43 Taking into account what ha been aid for the natural rotor flux with Equation 3.33, the total rotor tranient flux i finally calculated a in ,, Z V L V k V k e ce de e Z V L e V k t r pre t T t T t t r t r r DFIG r DFIG (3.39) Wind turbine driven DFIG The initial lip of a wind turbine driven DFIG may be ignificantly different from zero, thu the initial and pot-fault forced component of the rotor flux can no longer be calculated under the aumption of zero rotor current, a done for example in Equation 3.15, 3.37 and With reference to the poitive equence equivalent circuit of the DFIG in Figure 3.1, the poitive equence forced component of the pot-fault rotor flux can be calculated uing 3.6 a in 3.4, where Z 1 i the DFIG poitive equence impedance. t rf f rf t rf r tot r r m m rf r tot r m r e v k k e i L R L L L i L R L i Z v i,1 1,1,1,1,1,,1,1,,1 1 1,1 : :, (3.4) The pre-fault rotor flux mut alo be re-calculated taking into account the initial condition of the DFIG. If the initial apparent power, fed into the grid according to generator convention, and rotor lip of the DFIG are known, the pre-fault rotor flux i calculated, uing Equation 3.6, a: t pre r pre m r pre pre m m r pre r m pre pre pre r pre pre pre pre pre e v L L v S L L L L L i L i v v S i, * *,,,,, * *, 3,, 3 (3.41) where * denote the complex conugate.

57 44 Chapter 3. Fault Current of SCIG and DFIG Wind Turbine The negative equence forced rotor flux component i unchanged with repect to what we found previouly in Therefore, the total rotor flux equation for a wind-driven DFIG i finally given by Equation 3.4, which i a modification of 3.39, where rf,1 depend on the initial lip according to 3.4.,1,,1,, Z V L e ce de e Z V L e t r rf pre r t T t T t t r t rf r r DFIG r DFIG (3.4) No modification are needed for the tator flux, which i till given by 3.13 with the ubtitution of T with T,DFIG. The tator and rotor fluxe can now be ubtituted into Equation 3., leading to the final expreion, 3.43, for the hort-circuit current of a wind turbine driven DFIG uing crowbar reitance a a protection mean during network fault. Equation 3.43 can be eaily implemented in a programming language or a preadheet program and ued to get an approximate prediction for the DFIG hort-circuit current. t i t i t i Z V L e ce de L k e L V V V t i e Z V e L k L V e Z V L L k L L e L k L V t i n f r rf pre r t T t T t r T t pre n t t rf r t r r t rf r f r DFIG r DFIG DFIG,1, ' ' 1 ',1 ' 1 ' ' ',1 ' 1,,, (3.43) 3.6 DFIG Simulation Reult In thi ection, reult from imulation in SymPowerSytem (MATLAB R9b) are hown in order to validate the theoretical analyi preented in the previou paragraph. A wind turbine uing a DFIG model available in the tandard library of SimPowerSytem ha been ued. A detailed decription of the DFIG controller can be found in (Miller 3). The parameter for the DFIG are according to Table 3.1. Alo from the ame reference, the inertia contant for the wind turbine and the generator have now been aumed equal to 4.3 and.6 repectively. Thi model ha been modified by including a crowbar protection. The crowbar reitance i conidered to be time the DFIG rotor reitance and i connected to the rotor circuit through a ix-pule diode bridge and a witch. After crowbar inertion, the RSC i blocked. The crowbar remain inerted during the whole duration of the imulated period, i.e. 5 m. Typical time for RSC blocking may be lower than a few m (Akhmatov 5, Pannell 5). The crowbar inertion time

58 3.6. DFIG Simulation Reult 45 in the imulation ha been varied, to invetigate it influence on the hortcircuit current of the DFIG. For ake of clarity the current before fault occurrence are not hown, a wa intead done in Figure 3.3 and Figure 3.4. DFIG directly connected to the network The DFIG wind turbine ha been connected directly to the network a hown in Figure 3., without a tep-up tranformer. In Figure 3.5, Figure 3.6, Figure 3.7, the reult from different hort-circuit in the network are reported. The imulated hort-circuit current delivered by the DFIG i compared to the one predicted with Equation It can be noticed that the propoed method i capable of accurately reproducing the wind turbine driven DFIG hort-circuit current, if the crowbar i ideally inerted at the moment of fault inception. Error increae with increaing crowbar inertion delay time. Thee error were epecially appreciable for three-phae hortcircuit with crowbar inertion time delay higher than about 5 m. With lower delay or during unymmetrical fault, the crowbar inertion delay time had little ignificance on the fault current. I a (pu) I b (pu) time () Figure 3.5 Simulated with intantaneou (red +),.5 m delayed (green ), 5 m delayed (blue ) crowbar inertion and calculated (black dotted) DFIG hort-circuit current for a three-phae fault, with parameter p=1. The area encompaed in the rectangle in the firt m after the fault i magnified on the right part of the figure. Pre-fault voltage angle =9º. Initial apparent power S=.8-.5 pu, rotor peed r =1.5 pu. I c (pu)

59 46 Chapter 3. Fault Current of SCIG and DFIG Wind Turbine I a (pu) I b (pu) I c (pu) time () Figure 3.6 Simulated with intantaneou (red +),.5 m delayed (green ), 5 m delayed (blue ) crowbar inertion and calculated (black dotted) DFIG hort-circuit current for a phae-phae (b-c) fault, with parameter p=.8. Pre-fault voltage angle =º. Initial apparent power S=.8-.5 pu, rotor peed r =1.5 pu. The non-linearity due to the connection of the crowbar reitance through a diode bridge are alo noticeable in the imulation reult epecially in the cae of the phae-phae fault, ee Figure 3.6. However, they do not caue ignificant deviation from the prediction obtained with Equation Alo, notice that the realitic value of the wind turbine inertia contant caue a change in rotor peed after the fault, leading to a mall difference between the hort-circuit current imulated and calculated with 3.43, which aume a fixed rotor peed. Thi difference i mot viible in the imulated cae of a phae-phae fault, Figure 3.6. In Figure 3.7 it i reported a three-phae fault, ame a in Figure 3.5, but with the DFIG operating at ub-ynchronou peed, i.e. at lower initial loading. It i noted that the DFIG operating at ub-ynchronou peed deliver le peak current.

60 3.6. DFIG Simulation Reult 47 I a (pu) I b (pu) I c (pu) time () Figure 3.7 Simulated with intantaneou crowbar inertion (blue olid) and calculated (red dotted) DFIG hort-circuit current for a three-phae fault, with parameter p=1. Pre-fault voltage angle =9º. Initial apparent power S= pu, rotor peed r =.7 pu. Influence of MV line and tep-up tranformer Equation 3.43 ha been obtained under the aumption that the reitance of the generator tator winding i negligible. However, the reitance of the tep-up tranformer and MV line connecting the DFIG to the network may not be negligible. Their reitance and reactance hould be added in erie with the DFIG tator impedance when calculating the hort-circuit current. To invetigate how the reitive part of the tranformer and MV line impedance influence the accuracy of Equation 3.43, imulation have been performed by adding a tep-up tranformer and a MV line to the network in Figure 3.. A Dyn 33/.575 kv tep-up tranformer, rated 1.5 time the DFIG rating, with X=.6 pu and R=.1 pu ha been conidered. The voltage angle hift due to the tranformer Dy connection mut be conidered in the calculation. The line i a 5 mm cable with X=.14 Ω/km and R=.4 Ω/km. Data for the tranformer and the line parameter are from (Roeper 1985). Reult for three-phae fault at the end of the line for two different line length, 1 and 45 km, are reported in Figure 3.8. The reitive part of the MV line appreciably decreae the accuracy of Equation 3.43 only when the line length become higher than 45 km, correponding to a total

61 48 Chapter 3. Fault Current of SCIG and DFIG Wind Turbine reitance equal to 1 time the DFIG tator reitance. Therefore, in mot practical ituation, the reitive part of the MV line hould not decreae the accuracy of the method. The tranformer reitance doe not caue any appreciable lo of accuracy. The reitive character of the MV line impedance contribute to fater decay of the DC hort-circuit current component. 5 Line length = 1 km I (pu) Line length = 45 km I (pu) time () Figure 3.8 Simulated (olid) and calculated (dotted) DFIG phae current at a three-phae hort-circuit with parameter p=1 conidering the tep-up tranformer and a MV line with two different length value. Pre-fault voltage angle =9º. Initial apparent power S= pu, rotor peed r =1 pu. A cae of an earth-fault at the end of the MV line i alo reported in Figure 3.9, howing that Equation 3.43 provide accurate reult alo for unymmetrical fault on the D-ide of the tranformer. It hould be noted that the current hown in Figure 3.9 i the one at the DFIG terminal and not on the D-ide of the tranformer. To get the current on the D-ide of the tranformer, all the voltage in Equation 3.43 mut be expreed on that ide, taking into account the phae hift introduced by the Dy connection.

62 3.6. DFIG Simulation Reult 49 I a (pu) I b (pu) I c (pu) time () Figure 3.9 Simulated (olid blue) and calculated (dotted red) DFIG hort-circuit current for a phae-ground (a-g) fault at the end of a 15 km MV line, with parameter p=.99. Pre-fault voltage angle =9º. Initial apparent power S= pu, rotor peed r =.7 pu. Zero-equence current In the cae reported in Figure 3.9, the wind turbine doe not deliver any zeroequence current into the fault, becaue of the D connection of the tep-up tranformer. In the general cae, the main tranformer of a wind farm may be grounded on the high voltage ide. Thi may for example be the cae of a wind farm connected to the tranmiion ytem. In uch a cae, a zeroequence current will flow into earth fault on the high voltage ide. It i important to take into account thi contribution when calculating the total wind farm current delivered into the earth fault. Thi can eaily be achieved in two tep. Firtly, applying Equation 3.43, which i till valid apart the zero-equence current contribution, and calculating the phae current uing Equation 3.4. Since the interet i now on the current on the high voltage ide of the tranformer, the voltage ued in Equation 3.43 mut be the one on the high voltage ide. Secondly, the zero equence contribution mut be added to the phae current. If the total zero-equence impedance to the fault i Z =R +L, for a phae-ground (a-g) fault the zero equence contribution to be added to each phae current i:

63 5 Chapter 3. Fault Current of SCIG and DFIG Wind Turbine i f i t V Z e t Re f t L R t i t i e f (3.44) A phae-ground (a-g) fault with a direct grounding on the high voltage ide created by a 5 kva zig-zag tranformer, i hown in Figure 3.1. The zeroequence impedance of the zig-zag tranformer i Z =.5+.6 pu. The voltage on the high voltage ide i conidered to be 13 kv. The total fault current contribution from the wind farm into the fault in Figure 3.1 i calculated on the high voltage ide and include the DFIG contribution and the zero equence current. Even in thi cae the calculated fault current accurately reproduce the imulated one. I a (pu) I b (pu) time () Figure 3.1 Simulated (olid blue) and calculated (dotted red) high voltage ide hort-circuit current for a phae-ground (a-g), with parameter p=.99. Pre-fault voltage angle =9º. Initial apparent power S=.8-.5 pu, rotor peed r =1.5 pu. The voltage on the high voltage ide i 13 kv and a zig-zag tranformer i ued to create a direct grounding. I c (pu) Influence of GSC and crowbar reitance The total fault current delivered by a DFIG wind turbine i in reality the um

64 3.6. DFIG Simulation Reult 51 of the DFIG fault current, which can be calculated according to Equation 3.43, and the GSC fault current. The GSC maximum current may be around one third of the DFIG rated current. Depending on how fat the GSC control i, it may take ome cycle before it deliver maximum current. A three-phae fault i reported in Figure 3.11 including alo the contribution of the GSC. A een, the GSC contribute to a teady-tate current component, but doe not affect much the DFIG current delivered under the firt cycle after the hort-circuit. Finally, the impact of the crowbar reitance value i here invetigated by changing it value from, a in all previou cae, to 5 time the DFIG rotor reitance. A three-phae fault a in Figure 3.5 ha been conidered and the calculated and imulated reult are hown in Figure 3.1. When comparing with Figure 3.5, one can ee that a decreaed crowbar reitance reult in a higher peak current and a longer rotor tranient time contant, which implie a lower decay of the AC tator current component. Thi i in agreement with what ha been hown in (Pannell 1). I a (pu) I b (pu) I c (pu) time () Figure 3.11 Simulated DFIG and GSC (olid blue) and calculated DFIG (dotted red) phae current at a three-phae hort-circuit with parameter p=1 conidering the tep-up tranformer and a MV line of 1 km. Pre-fault voltage angle =9º. Initial apparent power S= pu, rotor peed r =1 pu.

65 5 Chapter 3. Fault Current of SCIG and DFIG Wind Turbine Alo, in the ame reference it ha been found that the tator natural flux decay at near-dc, meaning that it i actually lowly rotating while decaying. In (Pannell 1) the near-dc frequency of a 7.5 kw DFIG wa found to be.46 Hz without crowbar reitance and 1.76 Hz when a crowbar reitance i connected. Here, the angular rotation of the tator natural flux i given by the imaginary part of 1/T,DFIG. For the conidered machine, with a crowbar reitance of time the DFIG rotor reitance, the tator natural flux rotate with.3 Hz. Thi frequency drop to.1 Hz with a crowbar reitance of 5 time the DFIG rotor reitance. The difference in near-dc frequencie found here and in (Pannell 1) i due to different machine parameter. If the method propoed in thi chapter i applied to the machine conidered in (Pannell 1) without crowbar reitance, a frequency of.46 Hz i found. Chooing a crowbar reitance that reult in the ame rotor tranient time contant a in (Pannell 1), reult in a frequency of 1.67 Hz. Thee reult match very well with what ha been found in the mentioned reference time () Figure 3.1 Simulated (olid blue) and calculated DFIG (dotted red) phae current at a three-phae hort-circuit with parameter p=1. Pre-fault voltage angle =9º. Initial apparent power S=.8-.5 pu, rotor peed r =1.5 pu. The crowbar reitance i choen to be equal to 5 time the DFIG rotor reitance. I a (pu) I b (pu) I c (pu)

66 3.7. Summary Summary An approximate method for predicting the ymmetrical and unymmetrical hort-circuit current of a SCIG and a DFIG ha been propoed. The propoed method give a good prediction of the hort-circuit behavior of a wind farm uing a DFIG with crowbar protection, both for ymmetrical and unymmetrical fault in the network. A linear model of the induction machine ha been conidered and aturation ha been neglected. The accuracy of the reult obtained with the propoed method may be ufficient to replace the ue of imulation in many context, e.g. calculation of maximum current, calculation of it DC and AC component and hortcircuit calculation for protection relay etting. The impedance of tep-up tranformer and MV line hould be added in erie with the DFIG tator impedance. Even though thee impedance may have a non-negligible reitive part, it ha been found that in practical ituation thi fact doe not affect the accuracy of the method. Moreover, the method i capable of accurately reproducing the DFIG fault current even for unymmetrical fault on the MV ide of the tep-up tranformer. A factor that may limit the accuracy of the propoed method i the delay with which the crowbar reitance i inerted relative to the fault inception intant. Delay below 5 m reult in almot no lo of accuracy.

67

68 Chapter 4 Fault Current of DFIG Wind Turbine with Chopper Some of the commercial DFIG wind turbine intalled nowaday are protected by blocking the RSC and uing a chopper reitance on the DClink without inerting a crowbar reitance. The chopper reitance i witched to keep the DC-link capacitor voltage within an acceptable range. The analyi of the fault current contribution of thi type of DFIG wind turbine i rare and repreent a new iue in the literature. Thi chapter preent a method to model the DC link with the chopper a an equivalent reitance directly connected to the rotor during ymmetrical fault. Thi allow calculating the three-phae hort-circuit current of thee wind turbine a that of a wind turbine with an equivalent crowbar reitance. The method i epecially ueful conidering that accurate model of DFIG wind turbine with chopper protection are not commonly included in power ytem imulator. The material preented in thi chapter i baed on Publication [3]. 4.1 Introduction The fault current delivered by a DFIG wind turbine with DC chopper protection i eldom analyzed in the literature. In (Martinez 11(a)) it i propoed to model the DC-link capacitor and chopper imply a a reitance, with the ame value a the chopper reitance. Thi aumption mean that the whole rotor current rectified by the diode bridge of the RSC goe into the DC-chopper reitance, i.e. no current goe into the DC-link capacitor. Thi may be correct ut after a olid three-phae fault cloe to the DFIG terminal when the DFIG i at full power, depending on the chopper reitance value. But ome time after the fault, for le evere fault or if the DFIG i not at full power, the DC chopper will however not be continuouly witched on after 55

69 56 Chapter 4. Fault Current of DFIG Wind Turbine with Chopper the fault, but it will be witched according to a modulation index choen to control the DC-link voltage within a tolerance band. Under thee condition, ut uing the DC chopper nominal reitance value i not accurate. A cloer analyi of the ytem made up of the rotor circuit, the diode bridge of the RSC, the DC-link capacitor and the chopper i performed in thi chapter. The baic idea behind the analyi i the ame a in reference (Martinez 11(a)), i.e. to model the DC-link capacitor and chopper a an equivalent reitance een from the DFIG rotor circuit. However, it i hown that accurate modelling require a different value of the equivalent reitance for each different ymmetrical fault and initial loading of the DFIG. The analyi propoed here i valid under the period the RSC i blocked. Thi i alo the period when the highet fault current occur. Moreover, the fault current i related only to the phyical ytem compoed by the induction generator with it rotor circuit connected to the DC-link capacitor and DC chopper reitance through a diode bridge rectifier. Detailed model of a DFIG with chopper protection are not commonly available in power ytem imulation tool. Before imulating, modelling of the DFIG with chopper protection hould be performed, with explicit detailed repreentation of the witched chopper reitance with it control logic and of the free-wheeling diode of the RSC. Thi would alo reult in increaed imulation time. The approach propoed here allow diregarding all thee iue and allow uing the tandard model of a DFIG with crowbar protection alo to repreent a DFIG with chopper protection during hortcircuit. The propoed equivalence between DFIG with chopper and crowbar protection alo allow to ue hort-circuit calculation method developed for a DFIG with crowbar protection, a for example the method expoed in (Morren 7) and in Chapter 3. In particular the maximum hort-circuit current delivered to the grid can be calculated. The propoed theoretical analyi ha been validated through imulation. The reult for the hort-circuit current obtained through the propoed method are compared with imulation of a detailed model of a DFIG with chopper protection under different condition, howing good agreement. The DFIG model include a detailed repreentation of the RSC and GSC witching. It i alo hown that the DFIG with chopper protection deliver lower hort-circuit current than a DFIG with tandard crowbar protection, epecially for low initial loading.

70 4.. DFIG with DC Chopper Protection DFIG with DC Chopper Protection After the hort-circuit occurrence, the RSC i blocked and the rotor current flow into the DC-link. Under thi period, the rotor circuit of the DFIG i thu connected to the DC-link capacitor through the diode bridge made up by the anti-parallel diode of the RSC, ee Figure 4.1. Figure 4.1 DFIG with DC chopper protection. Rotor current flow through the antiparallel diode of the RSC during a fault with the RSC blocked. The chopper reitance will be witched to keep the DC capacitor voltage within acceptable value. The GSC will alo trive to control the DC-link capacitor voltage, but under eriou fault it capability for DC-link voltage control i highly reduced. The rotor circuit demagnetize feeding current into the DC-link capacitor, whoe voltage i controlled through the chopper. A explained in (Lopez 7), it i mainly the voltage caued in the rotor circuit by the natural tator flux that caue high rotor current. A already mentioned in Chapter 3, the natural tator flux arie to aure the continuity of the tator flux before and after the fault, according to the flux linkage theorem (Kimbark 1968). Thi voltage decay with the tator tranient time contant and ha a frequency equal to 1-lip time the network frequency. During thi period the rotor circuit of the DFIG can be conidered a a threephae decaying voltage ource, induced in the rotor by the tator natural flux, connected through a diode bridge rectifier, the anti-parallel diode of the blocked RSC, to a contant DC voltage ource, ee Figure 4.. When the rotor current ha decayed ufficiently, the RSC may be retarted while the fault i till preent. During thi tage the RSC i able to control the rotor current and hence the tator current. The problem addreed here i to etimate the fault current delivered during the period the RSC i blocked. The decribed ytem i highly non-linear due to the preence of the diode

71 58 Chapter 4. Fault Current of DFIG Wind Turbine with Chopper bridge and due to the action of the chopper. Some implification mut be done to allow for an analytical tudy to be carried out. Figure 4. Equivalent circuit of a DFIG rotor circuit connected to the DC-link through the anti-parallel diode of the RSC during a three-phae hortcircuit. The AC voltage i induced by the natural tator flux. L i the perphae inductance on the AC ide. The reitance on the AC ide i neglected. In thi chapter, the main idea i to look upon the DC-link ytem compoed of the DC-link capacitor and the chopper a if it were a fixed DC voltage ource. Thi i of coure an approximation ince the DC-link voltage will actually vary during the tranient but if the chopper i properly ized and controlled the DC-link voltage variation will not exceed a predefined value, e.g. ±1 % of it nominal value. In turn, the DC-link voltage ource i een from the rotor circuit ide of the diode bridge rectifier a a variable reitance, whoe value i equal to the ratio between the DC-link voltage and the rectified rotor current. The value of thi reitance will change with time, becaue of the decaying tranient rotor current. The equivalent reitance een from the rotor circuit i much higher than the rotor winding reitance and it increae with time, ince the rotor current decreae with time. In effect, it i therefore poible to look at thi configuration of the DFIG a a hortcircuited induction generator with an equivalent high rotor reitance, with the only difference being that the equivalent rotor reitance i now changing with time. However, to implify the analyi the value of the equivalent reitance een from the rotor circuit i aumed to be contant and equal to the ratio between the nominal DC-link voltage and the rectified rotor current at the intant of the hort-circuit. The equivalent reitance will act a a crowbar reitance implying a very fat demagnetization of the rotor natural flux, i.e. a fat decay of the AC tator current component induced by the rotor natural flux. Once the value of the equivalent reitance een by the rotor circuit i known, it can be ued to

72 4.3. Diode Bridge Rectifier with DC Voltage Source 59 calculate the maximum hort-circuit current of the DFIG a if thi had a crowbar protection. The key iue i therefore to etimate the equivalent reitance een by the rotor circuit during the period the RSC i blocked. To do thi the circuit hown in Figure 4. i firt analyzed. 4.3 Diode Bridge Rectifier with DC Voltage Source When a three-phae fixed magnitude voltage ource with a erie AC inductance i connected to a DC voltage ource through a diode bridge rectifier, a in Figure 4., it will upply a current whoe magnitude decreae with increaing DC voltage. If the ratio between the peak AC and DC voltage i ufficiently high, the AC ide current i cloe to a inuoidal curve and it flow continuouly. In the following it i aumed that thi condition hold, when not otherwie tated. Simulated voltage and current for uch a ytem are hown in Figure Voltage (V) x D1,D3 - D5 D1 - D6,D5 D1,D - D6 D - D4,D6 14 Current (A) time () Figure 4.3 Simulated phae a (olid), phae b (dahed) phae c (dahed-dotted) and DC (o) voltage (above) and current (below) for the ytem hown in Figure 4.. The peak AC voltage i higher than the DC voltage. The DC current will conit of ix pule during one period. To calculate the AC current a a function of the AC and DC voltage, it can be noticed that the diode D1 in Figure 4. mut tart conducting when the current in phae a become poitive, i.e. at t=θ, ee Figure 4.3. Meanwhile diode D3 will till

73 6 Chapter 4. Fault Current of DFIG Wind Turbine with Chopper conduct until the current in phae c become negative. Therefore under thi period the DC rectified current i the um of the current in phae a and phae c. Diode D5 i alo conducting during thi period. After the current in phae c ha become negative, diode D3 will ceae to conduct and diode D6 take over the current in phae c. At thi tage the rectified current i equal to phae a current, flowing through diode D1. The AC voltage and current can be expreed a: v v v i a an bn cn V in t V in t 3 V in t 4 3 I in t i b I in t 3 i I in t 4 c 3 (4.1) (4.) In the period in which diode D1, D5 and D6 are conducting the circuit hown in Figure 4. i equivalent to the one hown in Figure 4.4. Figure 4.4 Equivalent circuit for the ytem hown in Figure 4., when diode D1, D5 and D6 are conducting. During thi period one can write, with reference to Figure 4.4: v v bn an dib dic L L vcn dt dt dia dib vbn L L Vd dt dt (4.3) The above equation are valid in particular for t=θ+π/, i.e. when phae a current i at it maximum. By olving the above equation at t=θ+π/ for θ

74 4.3. Diode Bridge Rectifier with DC Voltage Source 61 and I uing the definition of voltage and current given in Equation 4.1 and 4., after ome manipulation one get: V co 1 d 3V V in I L (4.4) The above expreion are valid under the aumption that the AC current i inuoidal. Simulation of the ytem hown in Figure 4. how that thi i a good approximation when the ratio between V and V d i greater than 1. Seen from the AC ide, the DC voltage ource i equivalent in teady-tate to a reitance whoe value i the ratio between the DC voltage and the average rectified current. The term 3/π i ued to get the average value of the rectified current, made up of ix pule, given the peak value of the AC current. The value of the equivalent reitance on the DC ide i reported in Equation 4.5. R d Vd Vd (4.5) I 3 d I The equivalent DC reitance R d ha to be connected on the DC ide of a diode bridge rectifier. If intead it i deired to get directly an equivalent reitance a een from the AC ide and to kip the diode bridge, ome further conideration are neceary. For a DFIG, a realitic range for the product L i between. and.7 Ω, auming the um of the tator and rotor leakage reactance to vary between.15 pu (Anderon 1995) and.34 pu (a the conidered DFIG, whoe parameter are reported in Table 3.1), a tator voltage of 575 or 69 V and a generator rated power between 1.5 and MVA. Realitic value of the ratio V/V d are between 1 and, ee Equation Given thee aumption the ratio between the equivalent DC reitance R d and the value of L i in the range between.5 and 1.5. The iue conit now in replacing the diode bridge and the reitance R d with an equivalent reitance R eq, directly connected on the AC ide. The value of R eq mut be choen o that the ame peak current I on the AC ide i obtained a with the diode bridge and R d. It turn out that, under the made aumption for the value of the different parameter involved, the ame peak

75 6 Chapter 4. Fault Current of DFIG Wind Turbine with Chopper current I i obtained when the power P AC delivered to R eq i equal to the peak power delivered to R d, P DC,max. Note that the power delivered to R eq i contant, being the conidered AC circuit a three-phae and balanced one. Thi fact can be oberved in Figure 4.5 where two different cae are reported I d-peak /I (pu) P DC-max /P AC (pu) Figure 4.5 Ratio between peak current with diode bridge and R d and peak current with R eq veru ratio between peak power with diode bridge and R d and power delivered to R eq. V/V d = and L=. Ω (olid), V/V d =1.8 and L =.7 Ω (dahed). Therefore, the condition to be ued to find R eq i that the maximum intantaneou power on the DC ide hould be equal to the power delivered to the equivalent AC reitance R eq along with the condition of ame peak current in both cae, i.e. I d-peak =I. Thi lead to: / DC, max Rd I PAC 3Req I P (4.6) By olving Equation 4.6, one find that the equivalent reitance een from the AC ide i R eq R d (4.7) 3 Applying thi relation reult in the ame peak current in both cae. The reult in Equation 4.5 and 4.7 have been verified by comparing imulation of the ytem in Figure 4. with the ame ytem but with the diode bridge

76 4.3. Diode Bridge Rectifier with DC Voltage Source 63 and V d replaced by a reitance, whoe value i given by Equation 4.7. The above analyi i valid under the aumption that the peak AC voltage i higher than the DC voltage. The current in a cae when the peak AC voltage i almot equal to the DC voltage are hown in Figure Voltage (V) Current (A) time () Figure 4.6 Simulated phae a (olid), phae b (dahed) phae c (dahed- dotted) and DC (o) voltage (above) and current (below) for the ytem hown in Figure 4.. The peak AC voltage i almot equal to the DC voltage. It i oberved that they are no longer pure inuoid. Even in thee cae it i poible to replace the DC voltage with a reitance R d behind the diode bridge. However, the reitance R d increae a compared to the one calculated in Equation 4.5. The increae i non-linear and depend on the ratio between the AC RMS voltage and the DC voltage according to Figure 4.7, which ha been obtained by imulation. Simulation alo how that thee reult are independent of the value of the inductance on the AC ide, at leat for a wide range of inductance value. Even in thi cae, the equivalence between the ytem in Figure 4. and the ame ytem in which V d i replaced by a properly choen reitance ha been checked through imulation. When the current are no longer inuoidal, Equation 4.7 i no longer exact, ince replacing the diode rectifier and the

77 64 Chapter 4. Fault Current of DFIG Wind Turbine with Chopper DC voltage ource with an equivalent reitance on the AC ide, obviouly, cannot reproduce the non-inuoidal AC current. However, chooing R eq a in Equation 4.7 and according to Figure 4.7, till give a reaonable approximation of the AC current for the ytem hown in Figure 4.. A cae i reported in Figure 4.8, where the non-inuoidal current of the ytem with DC voltage ource and diode rectifier i hown along with the current in the ame ytem, but with an equivalent reitance on the AC ide. Figure 4.7 Ratio between the effective reitance R d,eff on the DC ide (found by imulation) and the reitance R d,calc calculated with Equation 4.5 a a function of the ratio between the RMS AC voltage V/ and the DC voltage V d.

78 4.4. DFIG with DC chopper protection 65 5 Current (A) time () Figure 4.8 Simulated AC phae current for the ytem hown in Figure 4. (olid) and for the ame ytem but with the DC voltage ource and diode rectifier replaced by an equivalent reitance on the AC ide (dahed). The ratio between the RMS AC voltage V/ and the DC voltage V d i choen equal to DFIG with DC chopper protection The reult obtained above for a diode bridge rectifier and a DC voltage ource can be directly tranlated to the rotor circuit of a DFIG with chopper protection during a three-phae fault when the RSC i blocked. All DFIG AC voltage and current in thi ection are expreed in a tator reference frame and are oberved from the tator. The DC-link voltage een from the tator i N ' tat V d Vd, nom (4.8) Nrot where N tat and N rot indicate the number of turn of tator and rotor winding. The apex i ued in Equation 4.8 to indicate that the DC quantity i referred to the tator ide. A typical value of the rotor to tator winding turn ratio N rot /N tat for a DFIG in wind power application may be around 3 (Peteron 5). Thi i to get low current in the rotor circuit during normal operation thu reducing the current rating for the RSC. A explained in (Lopez 7) it i the voltage caued in the rotor by the tator natural flux that i the highet component in the tranient rotor voltage during ymmetrical fault. It frequency i equal to the electrical rotor peed and it peak value een from the tator at the moment of fault occurrence i

79 66 Chapter 4. Fault Current of DFIG Wind Turbine with Chopper given a: RMS RMS L V V V (4.9) r, nat m pre, LL pot, LL 1 L 3 3 V r,nat i the peak value of the AC voltage in the rotor circuit, i.e. it mut RMS RMS ubtitute V in Equation 4.4. V pre, and V LL pot, are the line-line teady-tate LL voltage before and after the fault. To find the peak value of the rotor current after fault occurrence, one mut take into account that the angular frequency of the induced AC rotor voltage i equal to the rotor angular peed r. The equivalent inductance on the AC ide of the rectifier, i.e. the rotor circuit, i equal to the erie connection of the rotor leakage inductance and the parallel between the magnetizing and the tator leakage inductance, ee Figure 3.1. Thi i the tranient rotor inductance L r, defined, for example, in (Morren 7). The reulting equation for a DFIG, correponding to Equation 4.4, 4.5 and 4.7, are thu: co I r V 1 V 3V in r, nat ' r Lr ' d r, nat (4.1) R eq ' d ' d ' d V V (4.11) 3 I 3 3 I r For the DFIG model conidered in the imulation, the following value apply: V V ' d N N pre, LL 3 335V rot tat RMS V (4.1)

80 4.5. Simulation 67 The ratio between the induced peak AC rotor voltage and the DC-link voltage, both referred to the tator ide, depend on the initial lip and on the magnitude of the voltage dip during the fault. The following value are obtained in cae of a olid hort-circuit for three different lip: V V V r, nat ' Vd r, nat ' Vd r, nat ' Vd , for. 3, for (4.13), for. 3 A it can be een, for high poitive lip value the ratio between the induced rotor voltage and the DC-link voltage approache 1. Thi mean that the equivalent reitance calculated in Equation 4.5 mut be increaed a hown in Figure 4.7. For negative lip value, the analyi reulting in Equation 4.5 hould give acceptable reult. Negative lip value are alo thoe reulting in higher hort-circuit current. The ratio in Equation 4.13 are actually valid only at the moment of fault occurrence. They decreae with time a the induced rotor voltage decreae. A a conequence, a variable equivalent reitance hould be conidered, and it would increae with time. However, thi i not practical for ue in a imulation tool and the ue of a fixed reitance i propoed here intead, whoe value i obtained ut after fault occurrence a in Equation Simulation The aim of thi ection i to check how accurately the DC-link ytem of a DC chopper protected DFIG can be repreented a an equivalent reitance, according to the theoretical analyi performed above. The hort-circuit current delivered by a DFIG wind turbine with chopper protection ha been therefore compared to the one delivered by the ame DFIG wind turbine but equipped with a crowbar protection, whoe reitance ha been appropriately choen. Simulation have been performed in MATLAB SimPowerSytem (MATLAB R9b) and a DFIG model available in the tandard library ha been ued. Detailed repreentation of RSC and GSC witching i included in the model. The DFIG wind turbine i connected to an external power ytem a hown in Figure 4.9, without a tep-up tranformer. The voltage dip

81 68 Chapter 4. Fault Current of DFIG Wind Turbine with Chopper during the fault can be varied by changing the value of the parameter p, with p=1 correponding to a olid hort-circuit at the DFIG terminal. Figure 4.9 Single-line diagram of the power ytem ued in the imulation. To imulate a trong network, the parameter X th ha been fixed to one hundredth of the generator bae impedance. X th /R th =1. The DFIG conidered in thee imulation i the ame a decribed in Chapter 3 and it parameter can be found in Table 3.1. The model ha been modified to repreent a DFIG with crowbar protection or with chopper protection. The crowbar i inerted at the moment of fault occurrence. The chopper i connected through a controllable witch when the DC-link capacitor voltage exceed a predefined value and diconnected when it reache a econd lower value. The RSC i kept blocked in both cae for 1 m, i.e. for the total duration of the fault. Only one model at a time i connected. The crowbar reitance for the model with crowbar protection i choen according to Equation 4.11 and when neceary modified according to the reult in Figure 4.7. The value of the crowbar reitance depend therefore on the voltage dip during the fault and on the initial lip of the DFIG. Therefore, for each different imulation, the value of the crowbar reitance ha been aduted. In commercial wind turbine, the RSC i retarted when the rotor current decreae below a threhold value and therefore the time the RSC i blocked will depend upon the type of fault, voltage dip magnitude and the DFIG prefault lip. However, not knowing in detail how the condition to retart the RSC are implemented in a commercial wind turbine, it i here aumed that the RSC remain blocked for 1 m. Only the DFIG fault current during the time the RSC i blocked are reported here. Thi implie that the hortcircuit current hown in the next figure mut be interpreted a the one the DFIG would deliver if the RSC i not retarted.

82 4.5. Simulation 69 The current contribution of the GSC i not included either. In the wort cae, the GSC current may be up to about 3 % of the rated wind turbine power, depending on the GSC rating, and can be conidered to be known. Thi current contribution could be accounted for if needed, by adding it to the fault current delivered by the DFIG. A tudy on the influence of the GSC current contribution on the total wind turbine fault current ha alo been performed in Chapter 3. Different cae have been conidered with different initial loading for the DFIG and/or different voltage dip magnitude during the hort-circuit. In each cae, two imulation have been performed, one with the DFIG with crowbar and one with the DFIG with chopper protection. The two DFIG are at the ame operating point before the hort-circuit occurrence. The time tep ued in all the imulation i 1 µ. Simulation reult Three different initial loading for the DFIG have been conidered. For each of thee cae, three-phae hort-circuit reulting in a voltage dip of 1 % and 8 % at the generator terminal have been imulated. The initial DFIG lip, the active and reactive power correponding to thee initial condition and the equivalent reitance ued in the imulation are ummarized in the text of Figure 4.1 to Figure 4.15, where the intantaneou phae current for the chopper and crowbar protected DFIG are reported. It i een how the propoed method give good reult for all the conidered initial loading and fault cae. The accuracy of the reult would decreae for ub-ynchronou operation and for fault cauing low voltage dip if the correction reported in Figure 4.7 would not be applied to the equivalent crowbar reitance calculated with Equation 4.11.

83 7 Chapter 4. Fault Current of DFIG Wind Turbine with Chopper 4 I a (pu) I b (pu) I c (pu) time () Figure 4.1 Three-phae current of DFIG with DC chopper (olid) and equivalent crowbar (dotted) protection during a three-phae fault with parameter p=1. Crowbar equivalent reitance i.345 Ω. Initial apparent power S=.85+ pu, r =1. pu. 3 I a (pu) I b (pu) I c (pu) time () Figure 4.11 Three-phae current of DFIG with DC chopper (olid) and equivalent crowbar (dotted) protection during a three-phae fault with parameter p=.8. Crowbar equivalent reitance i.455 Ω. Initial apparent power S=.85+ pu, r =1. pu.

84 4.5. Simulation 71 3 I a (pu) I b (pu) I c (pu) time () Figure 4.1 Three-phae current of DFIG with DC chopper (olid) and equivalent crowbar (dotted) protection during a three-phae fault with parameter p=1. Crowbar equivalent reitance i.368 Ω. Initial apparent power S=.39+ pu, r =.99 pu. 3 I a (pu) I b (pu) I c (pu) time () Figure 4.13 Three-phae current of DFIG with DC chopper (olid) and equivalent crowbar (dotted) protection during a three-phae fault with parameter p=.8. Crowbar equivalent reitance i.59 Ω. Initial apparent power S=.39+ pu, r =.99 pu.

85 7 Chapter 4. Fault Current of DFIG Wind Turbine with Chopper 3 I a (pu) I b (pu) I c (pu) time () Figure 4.14 Three-phae current of DFIG with DC chopper (olid) and equivalent crowbar (dotted) protection during a three-phae fault with parameter p=1. Crowbar equivalent reitance i.43 Ω. Initial apparent power S=.19+ pu, r =.76 pu. I a (pu) I b (pu) I c (pu) time () Figure 4.15 Three-phae current of DFIG with DC chopper (olid) and equivalent crowbar (dotted) protection during a three-phae fault with parameter p=.8. Crowbar equivalent reitance i.679 Ω. Initial apparent power S=.19+ pu, r =.76 pu.

86 4.5. Simulation 73 In all cae the accuracy of the reult obtained with the equivalent crowbar reitance decreae with time. Thi i however expected, ince the actual equivalent reitance increae with time while a fixed reitance ha been ued in the imulation. The time divergence between the reult from the two model eem however not critical and the ue of a fixed equivalent reitance i therefore utified, epecially in the firt cycle after the fault when the RSC i blocked. The effect of the diode bridge rectifier are viible in the pule haped phae current for the DFIG with chopper protection. Thee pule are not preent in the DFIG with crowbar protection cae ince the crowbar reitance ha been directly connected to the rotor. It i intereting to note how the hort-circuit current magnitude of the DFIG with chopper protection decreae not only with decreaing voltage dip magnitude caued by the fault but alo with decreaing initial loading of the DFIG. In the cae with maximum poitive lip, correponding to a low initial loading, the DFIG with chopper protection deliver a current below pu even in the cae of a olid hort-circuit at it terminal, ee Figure With thi initial loading, the delivered hort-circuit current i below 1 pu if the three-phae hort-circuit caue a voltage dip of 8 % at the DFIG terminal, ee Figure Thi mean that in thee cae, it i poible that the RSC would block for a time horter than 1 m or not block at all. Therefore when the hort-circuit current obtained with the propoed method are low, a for example in Figure 4.13, Figure 4.14 and Figure 4.15, they are to be interpreted a the current that would be delivered by the DFIG if the RSC would be blocked. The propoed method i therefore alo helpful in invetigating what i the minimum time the RSC will remain blocked under different initial condition and fault everity and if the RSC i likely to block at all during the fault. Another intereting iue emerging from the analyi i that the DFIG with chopper protection preent an equivalent crowbar reitance increaing with decreaing everity of the fault and with lighter loading condition. Therefore, the DFIG with chopper protection deliver lower current the le evere the voltage dip caued by the fault and the lower it initial loading. A comparion between the hort-circuit current of a DFIG with chopper protection and a DFIG with crowbar protection i hown in Figure The crowbar reitance i choen to be equal to the cae preented in Figure 4.1, i.e. about 34 time the rotor reitance. The reaon why thi value ha been choen i that it reult in the ame hort-circuit current magnitude a the one delivered

87 74 Chapter 4. Fault Current of DFIG Wind Turbine with Chopper by a DFIG with chopper protection for a hort-circuit at the DFIG terminal, while thi i cloe to full load, a hown in Figure 4.1. Thi may actually be a too high value for a crowbar reitance. In (Akhmatov 5) a crowbar reitance equal to time the rotor reitance in uggeted. Depite thi fact, it i een in Figure 4.16, that that the DFIG with chopper protection would deliver a lower hort-circuit current compared to the crowbar protected DFIG, for a fault at the DFIG terminal with light initial loading correponding to the ame condition a in Figure I a (pu) I b (pu) I c (pu) time () Figure 4.16 Three-phae current of DFIG with DC chopper (olid) and crowbar (dotted) protection during a three-phae fault with parameter p=1. Crowbar reitance i.345 Ω, i.e. about 34 time the rotor reitance. Initial apparent power S=.19+ pu, r =.76 pu. 4.6 Summary A DFIG uing only chopper protection on the DC-link along with RSC blocking ha been analyzed under ymmetrical fault condition. It ha been hown that the DC-link ytem, made up of the DC-link capacitor and chopper protection, during three-phae fault condition i equivalent to an external reitance added to the rotor circuit. Therefore, the ymmetrical fault behavior of a DFIG with chopper protection i equivalent to the behavior of a DFIG with crowbar protection. A method for calculating the equivalent reitance ha been propoed. Even if

88 4.6. Summary 75 in theory, a time varying equivalent reitance hould be conidered, it ha been hown that the ue of a fixed equivalent reitance lead to good fault current prediction under different fault and initial loading condition. The equivalent reitance value increae with decreaing DFIG initial loading and decreaing fault everity. Therefore, accurate fault current prediction require the ue of a different equivalent reitance for each different fault and initial loading. An important conequence i that the DFIG with chopper protection deliver le fault current than a DFIG with crowbar protection. The lower the DFIG initial loading and fault everity, the lower the hort-circuit current delivered by the DFIG with chopper protection when compared to the one delivered by a DFIG with crowbar protection. The theoretical analyi ha been confirmed by imulation reult. Reult from different fault cae and operating condition are reported, howing that the DFIG with chopper protection deliver a hort-circuit current with a predominant DC component and with a very fat decaying tranient AC hort-circuit current component. The propoed method allow repreenting a DFIG with chopper protection with an equivalent model of a DFIG with crowbar protection both for hortcircuit current calculation and for hort-circuit imulation tudie. The advantage in uing a model of a DFIG with crowbar protection, when imulating, i that it i a tandard model uually available in power ytem imulation tool, unlike the DFIG with chopper protection. Moreover, hortcircuit current calculation method for the DFIG with crowbar protection are common in the literature.

89

90 Chapter 5 Fault Current from Wind Farm In the previou chapter, a detailed decription of the fault current of the individual SCIG or DFIG wind turbine ha been preented. The cope of thi chapter i to invetigate the validity of a ingle-machine approach to predict the fault contribution of the entire wind farm. Only wind farm with SCIG and DFIG wind turbine are conidered. 5.1 Introduction Wind farm may contain a large number of wind turbine. A an example, the Rödand offhore wind farm in Denmark comprie 9 wind turbine. The internal collector ytem of a wind farm can be tructured in variou configuration (Dutta 11). A imple wind farm configuration made up of three parallel row, ued in the imulation in thi chapter, i hown in Figure 5.1. Each ection contain a number of wind turbine, connected between them by the internal wind farm cable network. Each wind turbine i equipped with a tep-up tranformer from low, 575 V, to medium voltage, kv. At the Point of Common Coupling (PCC), the main wind farm tranformer connect the wind farm to the external grid. Becaue of hadowing effect between wind turbine in relation to the incoming wind, wind turbine in a wind farm are uually not all producing the ame output power. Moreover, the voltage may vary lightly throughout the internal wind farm cable network. Becaue of thee reaon at the event of a hort-circuit in the grid, the individual wind turbine will be in different initial tate. The quetion addreed here i how accurately the fault current contribution of the entire 77

91 78 Chapter 5. Fault Current from Wind Farm wind farm can be predicted uing an aggregate model of the whole wind farm, i.e. uing only one wind turbine with appropriate rated power and initial condition. The ue of an aggregate model intead of the detailed decription of the wind farm ave imulation time and reduce model complexity. Figure 5.1 Single-line diagram of detailed wind farm model with 9 wind turbine. 5. Detailed Wind Farm Model A detailed model of a wind farm, a hown in Figure 5.1, ha been implemented in DigSilent PowerFactory (DigSilent PowerFactory 1). The wind farm i compoed of three ection. Each ection contain three equal wind turbine with tep-up tranformer, connected to a kv internal

92 5.3. Aggregate Wind Farm Model 79 cable network. A realitic value for the cable ection for uch a wind farm configuration i 95 mm (Dahlgren 6, Garcia-Gracia 8). A 95 mm aluminum cable model in the PowerFactory tandard library ha been ued in the imulation and the parameter are reported in Table 5.1. Table mm cable parameter R (Ω/km).3 X (Ω/km).1 The wind turbine in the wind farm may be equipped with SCIG or with DFIG. Standard wind turbine model in the PowerFactory library have been ued, but the DFIG model ha been changed by adding a GSC, not implemented in the tandard model. The generator are rated 1.7 MVA and their parameter are a in Table 3.1. The tep-up tranformer are rated MVA. The main wind farm tranformer i rated MVA and connect the wind farm to a 13 kv external grid. The inertia contant for the wind turbine and the generator are reported in Table 5.. In the DFIG cae, the crowbar reitance i conidered to be time the DFIG rotor reitance and i inerted at the ame time a the RSC i blocked. Table 5. Inertia contant for the wind turbine and the generator H WT () 4.3 Hgen ().6 The wind turbine have different incoming wind and they are at a different operating point before fault occurrence. The initial condition for each wind turbine in the wind farm are reported in Table 5.4 and Table 5.5. The maximum difference between the highet and lowet active power production from the wind turbine i around 3 % of their rated power. A two-ma model i ued to repreent the haft ytem. In the cae of SCIG, a capacitor bank rated one third of total generator rating i alo connected at the kv common terminal. 5.3 Aggregate Wind Farm Model The aggregate model i repreented a a ingle wind turbine model with a rated power equal to the um of the rated power of the individual wind

93 8 Chapter 5. Fault Current from Wind Farm turbine in the wind farm. The ame hold for the unit tranformer of the aggregate model. Determining the equivalent of a wind farm internal cable network for an aggregate model i an iue dicued in many paper. A methodology for chooing the appropriate cable parameter in an aggregate model i laid down in (Muladi 6) and it i applied in thi chapter for determining the cable parameter for the aggregate model. Applying the method propoed in (Muladi 6) to the wind farm model in Figure 5.1 reult in the cable parameter reported in Table 5.3 for the aggregate model. Table 5.3 Cable parameter for the aggregate wind farm model R (Ω).13 X (Ω) Decription of Simulation Symmetrical and unymmetrical fault have been imulated on the 13 kv line in the network, cloe to the PCC. It i aumed that the wind turbine generator remain connected to the network during the whole fault duration. The total wind farm current a meaured on the 13 kv ide at the PCC i reported in the next figure during ymmetrical and unymmetrical fault. Thi current i compared to the current meaured at the ame point when the detailed model of the wind farm i replaced by an aggregate model. The aggregate model i initialized by uing the maximum, the minimum or the average wind peed, a experienced by the detailed wind farm model. Thi i done in order to check which of thee value lead to the bet fault current prediction accuracy. 5.5 Simulation of a SCIG Wind Farm The wind farm hown in Figure 5.1 ha been imulated auming all the wind turbine to be equipped with SCIG. The initial operating condition for the wind turbine are reported in Table 5.4. The reactive power in Table 5.4 i the total delivered to the network by the generator. The cable parameter in the aggregate model are choen a reported in Table 5.3. The total wind farm current a meaured on the 13 kv ide at the PCC i reported in the next figure during ymmetrical and unymmetrical fault. Fault current are reported in pu of the main 13/ kv tranformer current. The fault occur after 4 m and lat for m.

94 5.5. Simulation of a SCIG Wind Farm 81 Table 5.4 Initial condition for SCIG wind farm P (MW) Q (Mvar) v wind (m/) WT WT WT WT WT WT WT WT WT A een in Figure 5., Figure 5.3 and Figure 5.4, the initial operating point of the SCIG wind turbine ha a minor effect on it hort-circuit current. In particular, the peak current are very cloe in all cae. However, the bet agreement with the detailed model i obtained when chooing the average wind for determining the initial condition of the aggregate model. I a (pu) I b (pu) I c (pu) time () Figure 5. Simulated SCIG hort-circuit current for a three-phae fault. Detailed model (black), aggregate model initialized with maximum (blue), minimum (red) and average (magenta) wind peed.

95 8 Chapter 5. Fault Current from Wind Farm time () Figure 5.3 Simulated SCIG hort-circuit current for a phae-phae fault. Detailed model (black), aggregate model initialized with maximum (blue), minimum (red) and average (magenta) wind peed. I a (pu) I b (pu) I c (pu) time () Figure 5.4 Simulated SCIG hort-circuit current for a phae-a to ground fault. Detailed model (black), aggregate model initialized with maximum (blue), minimum (red) and average (magenta) wind peed. I a (pu) I b (pu) I c (pu)

96 5.5. Simulation of a SCIG Wind Farm 83 The high frequency component after the fault i due to reonance between the capacitor bank and the inductance in the grid. To invetigate the importance of a correct choice of the cable parameter for the aggregate model, a comparion with the hort-circuit current delivered by the aggregate model diregarding the internal cable i hown in Figure 5.5. The main difference between the hort-circuit current delivered by the detailed and the aggregate model without cable repreentation i the decay of it DC component. The DC component in the current of phae b and c decay fater in the detailed model cae. The decay of the DC component i affected by the reitance of the cable in the wind farm model. Ignoring completely the cable in the aggregate model reult in longer time contant for the DC component of the hort-circuit current. A too long cable, and hence high reitance, in the aggregate model reult intead in horter time contant for the DC component of the hort-circuit current. I a (pu) I b (pu) I c (pu) time () Figure 5.5 Simulated SCIG hort-circuit current for a for a three-phae fault. Detailed model (black), aggregate model with cable parameter according to Table 5.3 (blue) and with no cable model (red). The aggregate model i initialized with the average wind peed. Thee eem to be however minor iue. The ue of a ingle machine equivalent lead to accurate prediction of the total SCIG wind farm fault contribution at the PCC.

97 84 Chapter 5. Fault Current from Wind Farm 5.6 Simulation of a DFIG Wind Farm The SCIG in the wind farm are now replaced by DFIG with crowbar protection. Each wind turbine i initially loaded according to Table 5.5. The different loading of the wind turbine i caued by different incoming wind, and thi lead to a different initial lip for each wind turbine generator. The fault occur after 4 m. It i here aumed that the crowbar remain connected for a time duration of 1 m after the fault and only thi period i reported in the next figure. The period with the crowbar inerted i alo the mot intereting period during which the highet fault current are delivered. When the crowbar i diconnected the RSC regain the control of the tator current and the DFIG can be een a a contant current ource, in a imilar way to a FSC (Walling 9). Notice that the crowbar may diconnect before 1 m, depending on the magnitude of the current and the DC-link capacitor voltage. In the cae of unymmetrical fault, pecial control algorithm to reduce torque or DClink voltage ocillation can be ued. The delivered fault current would then depend on thi control cheme when the crowbar i diconnected. Here, it i aumed intead that the crowbar remain connected through the whole duration of the unymmetrical fault. Thi i a reaonable aumption at leat for evere unymmetrical fault (Semaan 6(b)), ince the negative equence voltage in the grid would induce high rotor current through the whole duration of the fault. The fault current from the detailed wind farm model i reported in the next figure along with the fault current from the aggregate model with different initial condition. In cae of a three-phae hort-circuit, the fault current from the detailed and aggregate model are reported in Figure 5.6. The cable parameter in the aggregate model are choen a in Table 5.3. Alo in thi cae, the wind peed given to the aggregate model to determine the initial condition of the aggregate model affect marginally the hort-circuit current. The average wind peed reult however in better accuracy. From the analyi of imulation reult, it i alo concluded that in the detailed model, the intant in which the crowbar i inerted i roughly the ame for all the generator in the wind farm.

98 5.6. Simulation of a DFIG Wind Farm 85 Table 5.5 Initial condition for DFIG wind farm P (MW) Q (Mvar) v wind (m/) r (pu) WT WT WT WT WT WT WT WT WT I b (pu) I a (pu) I c (pu) time () Figure 5.6 Simulated DFIG hort-circuit current for a three-phae fault. Detailed model (black), aggregate model initialized with maximum (blue), minimum (red) and average (magenta) wind peed. The fault current delivered by the detailed and the aggregate model for a phae-phae and a ingle-phae-to-ground fault are hown in Figure 5.7 and Figure 5.8.

99 86 Chapter 5. Fault Current from Wind Farm I a (pu) I b (pu) I c (pu) time () Figure 5.7 Simulated DFIG hort-circuit current for a phae-phae fault. Detailed model (black), aggregate model initialized with maximum (blue), minimum (red) and average (magenta) wind peed. I a (pu) I b (pu) I c (pu) time () Figure 5.8 Simulated DFIG hort-circuit current for a phae-a to ground fault. Detailed model (black), aggregate model initialized with maximum (blue), minimum (red) and average (magenta) wind peed.

100 5.7. Summary 87 The ame concluion drawn for three-phae ymmetrical hort-circuit alo hold for unymmetrical hort-circuit. In thi cae, it i eaier to appreciate how chooing the average wind peed in the wind farm to initialize the aggregate model reult in better agreement. In concluion, aggregate model of DFIG wind farm can be ued to accurately predict the fault current contribution of DFIG wind farm. 5.7 Summary The ue of a ingle-machine approach to model the total fault current contribution from a SCIG and a DFIG wind farm ha been checked through imulation. The maximum difference between the lightet and heaviet loaded generator in the farm i aumed to be around 3 % of the nominal power of a generator. Both for SCIG and DFIG wind farm, the ue of an aggregate model to predict the hort-circuit current contribution of the entire wind farm lead to accurate reult for ymmetrical a well a unymmetrical fault. A long a the aggregate model i initialized with a wind peed in the range between the minimum and maximum wind peed experienced in the wind farm by the individual generator, the initial condition of the aggregate model doe not ignificantly affect the hort-circuit current. Initializing the aggregate model with the average wind peed in the wind farm lead to bet accuracy. For DFIG wind farm, thi i true at leat in the firt cycle after the fault, during the period the crowbar i inerted and the RSC i blocked. Ignoring completely the cable network in the aggregate model lead to a lower decay of fault current DC component. The aggregate model hould therefore be interconnected to the grid with a cable, whoe parameter are to be choen to achieve ame DC component decay rate a for the detailed wind farm model.

101

102 Chapter 6 GFRT for Fault below Tranmiion Network Swedih grid code (SvKFS 5:) require GFRT of wind turbine for fault at tranmiion level, above and included the kv network, cleared within maximum 5 m by the protection ytem. Different requirement are et for wind farm below and above 1 MW. Document (TR ) by the Swedih TSO tate that line fault on the tranmiion level hould be cleared intantaneouly with a maximum fault clearing time of 13 m. Thi aume that communication between the two line end i ued. In cae of intervention of the circuit breaker failure protection, the maximum clearing time i required to be 5 m. Fault clearing time below the tranmiion level are often longer than at tranmiion level, due to different employed protection ytem. At ub-tranmiion level communication i not alway ued, reulting in longer fault clearing time. Therefore, it i not certain that a wind turbine which repect grid code requirement for fault at the tranmiion level i alo able to ride-through fault below tranmiion level. Thi iue i further treated in thi chapter. 6.1 Introduction An invetigation on the voltage dip profile reulting from fault at ubtranmiion and medium voltage level and on their effect on GFRT of wind turbine i performed in thi chapter. In Sweden there i actually no requirement for wind turbine to ride-through fault below tranmiion level. The voltage-time curve defined in the Swedih grid code i in fact applicable to point on the tranmiion network following a fault at the ame network level. Wind turbine connected below the tranmiion network mut ride-through uch fault cauing voltage dip at the connecting tranmiion network bue above the Swedih GFRT curve. To ae whether the GFRT of wind turbine i compromied for fault at ub- 89

103 9 Chapter 6. GFRT for Fault below Tranmiion Network tranmiion and medium voltage level, reaonable aumption on their ridethrough capability mut be made. The voltage-dip profile obtained from imulation can then be checked againt thee aumption to evaluate the GFRT of wind turbine. A poible way to ae GFRT of wind turbine i to ue the Swedih voltagetime curve and let it repreent the minimum GFRT requirement at the connection point for fault below tranmiion level. Uing the Swedih voltage-time curve a a reference for GFRT aement in thi chapter i utified ince wind turbine connected ufficiently cloe to the tranmiion network may experience voltage dip not very different from the one experienced in the tranmiion network during fault in that network. A a matter of fact, for uch fault the voltage dip experienced by wind turbine connected cloe to the tranmiion network may even be lower than the one experienced in the tranmiion network. In thee cae, wind turbine complying with the Swedih grid code can alo be aumed to be capable of GFRT for fault below tranmiion level cauing voltage dip profile at the connection point above the Swedih voltage-time curve. It i reaonable to aume that commercial wind turbine are able to ridethrough voltage-time curve defined in other national grid code and applicable at the connection point, which may be either at tranmiion or ditribution level. The voltage-time curve in the Danih and German (E.ON) grid code are two uch example. The aim of the Swedih GFRT curve i to afeguard the integrity of the power ytem for large diturbance, a a fault on the tranmiion ytem. A already aid, for wind turbine connected ufficiently cloe to the tranmiion ytem, the Swedih GFRT curve can alo be aumed to be repreentative of the minimum GFRT requirement at the connection point. It i not urpriing if wind turbine complying only with thi minimum requirement would not ride-through ome of the fault cae at ub-tranmiion and medium voltage level conidered in thi chapter. It i expected intead that wind turbine complaining with the Danih and E.ON voltage-time curve, defined even for fault below tranmiion level, would ride-through the fault conidered in the imulation in thi chapter. Thee three curve will be ued a a reference in thi chapter to evaluate GFRT of wind turbine for fault below tranmiion network. A imilar analyi to the one in thi chapter ha been carried out in (Coter 9) for ride-through of CHP plant, but without taking into account the

104 6.. Network and Component Modelling 91 load dynamic which i intead central in thi work. The author of the mentioned paper do not focu on the pot-fault voltage recovery and they do not relate the reulting voltage dip to any GFRT curve. In Section 6., an analyi i performed on the extenion of the network area where the wind turbine GFRT capability could be endangered during a fault on ub-tranmiion network. To addre thi quetion, the value of the voltage dip at different bue in a network can be calculated uing tandard power ytem analyi method. Thi analyi can be ued to etimate the total amount of wind power expected to be diconnected following uch a fault, once the total amount of wind power connected to a ub-tranmiion bu i known. Thi analyi i however not ufficient and mut be completed by dynamic imulation to account for the dynamic load effect on voltage recovery. The influence of the load mix compoition on the voltage recovery after fault clearing i then addreed. It i hown that load mix compoition and it dynamic behavior play a crucial role for GFRT when the fault i below tranmiion network. The preence of dynamic load, induction motor, woren the voltage recovery in the network after fault clearing (Taylor 1994). A realitic analyi hould moreover conider alo the amount of dynamic load that i tripped during the voltage dip. Indutrial induction motor may trip during voltage dip due to the releae of the trip contact of the AC contactor (Taylor 1994). However, the amount of dynamic load diconnected during a dip, the value of the voltage and time at which diconnection happen are all unknown variable and therefore reaonable cenario mut be ued intead (Taylor 1994, Tranpower 9). Simulation of ome tet cae are performed in DigSilent PowerFactory (DigSilent PowerFactory 1) to find out the voltage dip reulting from different fault and load cenario. The reult are hown after an introduction on the ytem, protection and load modelling aumption. 6. Network and Component Modelling The network ued in thi chapter i the Nordic 3 network model (CIGRÉ 1995). However, thi tudy focue only on a mall part of thi network, a 13 kv ring compoed of 5 node, 141 through 145. Thi part of the network i hown in Figure 6.1. Dynamic imulation are carried out in DigSilent PowerFactory. The 13 kv ring i connected with the 4 kv tranmiion network at two

105 9 Chapter 6. GFRT for Fault below Tranmiion Network point. Some modification to the original Nordic 3 model have been done in thi 13 kv ytem. Two extra 13 kv node, 6 and 7, have been added compared to the original model. A medium voltage network with two feeder ha been added to bu 141 with a wind farm model connected to it. The load in the ytem were originally of tatic type. They have been modified o that a certain percentage i repreented a impedance load and the remaining part a induction motor load. The total teady-tate active power drawn by the load i till unchanged and extra capacitor bank have been added to compenate the extra reactive power needed by the motor load. The motor load i repreented uing a tandard model in PowerFactory with a mechanical load varying quadratically with rotor peed. Figure kv ytem ued in the imulation. The ytem i originally part of the Nordic 3 model, but it ha been modified by adding bue 6, 7 and a 5 kv network with two feeder. Protection ytem The protection ytem on the 13 kv network i aumed to be made up primarily of ditance protection without any communication. Ditance protection with 3 different zone etting i conidered. For hort-circuit without earth involved, the delay time of the ditance protection are aumed

106 6.. Network and Component Modelling 93 to be a in Table 6.1. Table 6.1 Intentional delay time for different zone of ditance protection Zone 1 Zone Zone 3 Delay time.4 1. Zone 1 i aumed to cover up to 8 % of the line length. Zone cover up to 1 % of the line length. Two main protection are aumed to be intalled o that the probability of protection failure i negligible. Failure of a circuit breaker i dealt with circuit breaker failure protection, CBF (Horowitz 1995). It i aumed that the total clearing time with CBF protection operation i 5 m. Ditance protection with communication or differential protection may be ued or needed in cae of hort line, when the line ditance i too hort to achieve electivity with only ditance protection. However, thi kind of protection i not conidered in thi tudy. The total fault clearing time i given by the um of the intentional delay time of the protection relay, plu the time neceary for the protection relay to detect a fault and the breaker opening time. The time neceary for the protection relay to detect a fault may vary depending on the type of relay and the algorithm it i baed on. A realitic value i 1.5 cycle, i.e. 3 m (ABB REL65). Here, it i aumed that the protection relay operating time i 4 m. A realitic value for a 13 kv circuit breaker interrupting time i 3 cycle, i.e. 6 m (ABB Type PMI). The total um of relay operating time and circuit breaker interrupting time i therefore aumed here to be equal to 1 m. Thi time mut be added to the intentional delay time indicated in Table 6.1. Load modelling Three different cae for the load compoition mix are conidered. In the firt cae, the load i aumed to be entirely of impedance type. In the econd cae, 3 % of the load i aumed to be dynamic load and the remaining 7 % impedance load. In the third cae, half of the load i modelled a dynamic and half a impedance load. Modelling of dynamic load with tatic model i not accurate for large voltage variation (Taylor 1994). Therefore, the dynamic load i repreented with the detailed model of an induction motor driving a mechanical load with torque

107 94 Chapter 6. GFRT for Fault below Tranmiion Network increaing quadratically with peed. In reality, indutrial motor are automatically diconnected from the network when the voltage drop below a certain level for a certain time. The diconnection i due to the deenergization of the AC contactor ued to connect the motor to the grid. The drop-off voltage may vary between 3 and 65 % of nominal voltage (Taylor 1994). Induction motor ued in mall appliance, mainly ingle-phae, will not be diconnected due to under-voltage but they are olely protected againt overload. Dynamic load will woren the voltage profile during a fault becaue of the high amount of reactive power they aborb. The electromagnetic torque produced by induction motor i proportional to the quare of the voltage, thu a voltage drop trongly decreae the produced motor torque leading to a deceleration of the motor with conequent higher aborption of reactive power. If the voltage i not retored, the motor could come to a tall. Diconnection of motor load during voltage drop i beneficial to voltage tability. It i aumed that either 5 % or 5 % of the total motor load i diconnected during the voltage dip. The diconnection i performed when the voltage drop below.5 pu with a time delay of 1 m. Only thoe motor connected to the bue experiencing ufficient voltage drop are hence diconnected. The amount of dynamic and tatic load i choen in each cae, o that in teady-tate the total active power drawn by the load in the ytem i unchanged, i.e. if a certain amount of dynamic load i added, a correponding amount of tatic load i diconnected. The reactive power drawn in teadytate by the motor load i compenated locally with capacitor bank. 6.3 Voltage dip calculation In (Souza 1) it i hown how the vulnerability area technique can be ued to ae the effect of voltage dip in the grid on the GFRT of wind turbine. The technique baically give the voltage dip magnitude at a grid location for hort-circuit applied throughout the whole ytem. If thi technique i combined with tatitical data on fault rate, the expected frequency for wind turbine diconnection following a fault can be found. In general, to calculate the voltage dip magnitude at different bue in the ub-tranmiion network given a fault at a certain bu, tandard matrix calculation method can be employed (Glover ). The reult will give a firt indication on which bue are eriouly affected by the voltage dip and therefore an etimation of the amount of wind power that may experience

108 6.4. Fault at ub-tranmiion network 95 GFRT problem. Once the network tructure i known, the bu admittance and impedance matrice, Y and Z, can be calculated. If the fault i at bu k, the voltage at bu n i given a: V Z kn n 1 Vpre fault Z (6.1) nn Thi approach ha been followed here and the reult are hown in Figure 6. along with the value obtained through imulation in PowerFactory. It i noted that in thee imulation, only tatic load have been conidered. Figure 6. Voltage dip magnitude for two different three-phae fault. A een, the calculated reult are quite cloe to the imulated value. Thee reult indicate that in the conidered network, only the bue cloe to the fault experience voltage dip that could endanger the GFRT of wind turbine. Thee reult are however only rough and do not take into account any dynamic apect. Alo, the everity of the dip mut be combined with the duration of the dip to make any aertion on GFRT of wind turbine. The duration of the fault depend on the protection ytem. Fault not correctly cleared by the protection ytem may caue failure to ride through the fault even if the voltage dip i not particularly deep. 6.4 Fault at ub-tranmiion network The aim of thi paragraph i to find out how voltage dip following a et of conidered fault event on the ub-tranmiion network look like and to evaluate if they poe any rik to GFRT of wind farm eventually connected in the vicinitie of the fault.