Voltage security assessment with high penetration levels of utility-scale doubly fed. induction generator wind plants.

Transcription

1 Voltage ecurity aement with high penetration level of utility-cale doubly fed induction generator wind plant by Ryan Konopinki A thei ubmitted to the graduate faculty in partial fulfillment of the requirement for the degree of MASTER OF SCIENCE Major: Electrical Engineering Program of Study Committee: Venkataramana Ajjarapu, Major Profeor Chen-Ching Liu Dougla Gemmill Iowa State Univerity Ame, Iowa 2009

2 Copyright Ryan Konopinki, All right reerved. ii

3 iii TABLE OF CONTENTS LIST OF FIGURES LIST OF TABLES v vii CHAPTER 1. Introduction Trend, Regulation, and Concern with Wind Generator Fundamental of Doubly Fed Induction Generator 5 CHAPTER 2. Static Modeling of DFIG Wind Plant Introduction DFIG Plant Capability Curve DFIG Wind Park Load Flow Model Wind Park Aggregation Reactive Capability Validation 15 CHAPTER 3. DFIG Plant Dynamic Modeling and Control Dynamic Modeling Principle Aerodynamic Characteritic Mechanical Drive Train Characteritic Generator Characteritic Vector Control Pitch Control Characteritic Dynamic Model Validation Dynamic Model Validation DFIG Control Enhancement Electrical PEC Control Structure Controller Development 31 CHAPTER 4. Voltage Security Aement Methodology Introduction Static Security Aement for High Wind Penetration Dynamic Validation of Wind Penetration Level 48 CHAPTER 5. Dicuion and Analyi of Reult Power Sytem Decription Penetration Level Characteritic Tranfer Margin with Extended Reactive Capability Static Analyi Reult Dynamic Analyi Reult Voltage Control Application Dynamic Scenario Setup Sytem Performance Validation 62 CHAPTER 6. Project Concluion Introduction Increaed Sytem Tranfer Suggeted Order 661-A Reviion 71

4 iv APPENDIX A. Maximum Power Tracking (MPT) Strategy 73 APPENDIX B. Machine Parameter 74 APPENDIX C. Publication 75 Accepted for publication 75 BIBLIOGRAPHY 76 ACKNOWLEDGEMENTS 84

5 v LIST OF FIGURES Figure 1. Yearly intalled type of turbine by percentage from [8] 3 Figure 2. High-level ytem overview of a DFIG wind turbine 6 Figure 3. Power electronic converter architecture 6 Figure 4. DFIG high level power balance chematic 7 Figure 5. DFIG tatic machine model 9 Figure 6. DFIG wind park tatic power capability curve in per unit 12 Figure 7. DFIG Main Block Diagram [30] 18 Figure 8. Rotor performance coefficient (cp) v. tip-peed ratio ( λ ) for variou pitch angle ( β ). 20 Figure 9. Pitch Controller Block Diagram [30] 26 Figure 10. DFIG Dynamic Model Validation 29 Figure 11. DFIG Power Electronic Control Block Diagram 31 Figure 12. FERC Order 661-A (Black) v.. POI Bu Voltage (Red) 33 Figure 13. DFIG Crowbar Protection Repone Schematic 34 Figure 14. Current Flow in Power Electronic during Crowbar Operation 35 Figure 15. Coordinated Converter Control Schematic during Crowbar Operation 36 Figure 16. Impact of Grid Side Reactive Booting with (black) and without (red) Control 37 Figure 17. PEC Current with and without Crowbar Re-trip Prevention 38 Figure 18. Crowbar Re-trip Prevention Implementation 39 Figure 19. DC Link Capacitor Voltage with and without Chopper Control 40 Figure 20. DC Link Chopper Circuit Schematic 41 Figure 21. Simplified two bu power ytem 42 Figure 22. Voltage Security Aement Flow Chart for High Wind Penetration 48 Figure 23. Simulated Power Sytem with Wind Park Interconnection at Bu Figure 24. Power Tranfer Margin from Increaed Plant Output with No Reactive Compenation 54 Figure 25. Power Tranfer Margin from Increaed Plant Output and 50 MVAr Support at Figure 26. Power Tranfer Margin from Increaed Plant Output 50 MVAr upport at 204 & Figure 27. DFIG Wind Plant Voltage Controller Schematic 60 Figure 28. Comparion of bu 153 voltage (p.u.), reactive power (MVAr) (park: 3008/5 red/black), and rotor current (ka) (park: 3008/5 pink/brown) from 20% penetration at cut-in peed with RPF (left) and with CC (right) control 65 Figure 29. Comparion of bu 153 voltage (p.u.), reactive power (MVAr) (park: 3008/5 red/black), and rotor current (ka) (park: 3008/5 pink/brown) from 20% penetration at 15% output with RPF (left) and with CC (right) control 66 Figure 30. Comparion of bu 153 voltage (p.u.), reactive power (MVAr) (park: 3008/5 red/black), and rotor current (ka) (park: 3008/5 pink/brown) from 20% penetration at 50% output with RPF (left) and with CC (right) control 67

6 Figure 31. Comparion of bu 153 voltage (p.u.), reactive power (MVAr) (park: 3008/5 red/black), and rotor current (ka) (park: 3008/5 pink/brown) from 20% penetration at 100% output with RPF (left) and with CC (right) control 69 Figure 32. DFIG wind park electrical output (p.u.) and rotor peed (p.u.) veru wind velocity (m/) 73 vi

7 vii LIST OF TABLES Table 1. Converter Sizing for Theoretical Reactive Operation 16 Table 2. Bae Cae Scenario Lit 45 Table 3. MW injection for given wind penetration and park output 53 Table 4. Power Tranfer Margin at Different Penetration Level with No Reactive Support 55 Table 5. Power Tranfer Margin at Different Penetration Level (50 MVAr at 204) 57 Table 6. Power Tranfer Margin at Different Penetration Level (50 MVAr at 204 and 3008) 59 Table 7. DFIG Wind Park Machine Simulation Parameter 74

8 1 CHAPTER 1. INTRODUCTION Thi work detail the impact and development of utilizing a capability curve for a doubly fed induction generator (DFIG) wind plant. The teady tate and dynamic power ytem repone to the reactive capability of the machine are teted with high wind penetration level. Thi work wa motivated by the interconnection requirement et forth by FERC in order 661-A which mandate the operation of wind park within a power factor range of 0.95 leading and lagging. Thi retricted operation dratically under-utilize the reactive output of the machine and hence create a potential negative impact to the power ytem. The reult preented in thi work outline a methodology for determining the penetration level of a ytem baed on voltage and power tranfer criteria defined for the ytem. DFIG wind model control enhancement were alo developed which improve pot fault voltage profile and damp ytem wing, preventing overhoot immediately after a diturbance. In the preented cae tudy the extended reactive limit of the plant were found to prevent ytem collape. 1.1 Trend, Regulation, and Concern with Wind Generator In today world, with riing global fuel price and concern relating to climate change, there i an increaing focu on renewable ource to atify energy requirement. Among the renewable reource, wind generation i gaining prominence due to available technologie that allow for large-cale power generation [1], [2]. The year 2007 wa a record

9 2 year for wind generation in the United State with a total increae of 45% in wind intallation that accounted for 30% of all newly added generation [3], [4]. Federal policy in the form of production tax credit and tate regulation in the form of renewable portfolio tandard (RPS) [5] have contributed to encouraging the development of wind generation in the United State [6]. Over 25 tate have accepted RPS by requiring a ubtantial contribution from renewable to their power generation portfolio [7]. Conidering thee incentive and mandate, wind generation technology appear to be the mot popular renewable option for developer. Although, the ame type of regulatory policy that i helping timulate contruction, may be counter productive to power ytem operation. In 2005, FERC order 661 and 661-A [23] [24] were releaed which mandate the interconnection requirement for large wind park over 20MW. According to thi order, a key requirement for plant operation i that the power factor at the point of interconnection (POI) mut remain between 0.95 leading and 0.95 lagging. Reference [24] tate that the reaon for thi ruling i that reactive power capability for a wind plant i a ignificant additional cot compared to conventional unit that poe inherent reactive capability. A more wind intallation appear in year to come, policy ruling may need modification to allow for healthy performance of the power ytem. The majority of the newly intalled wind generation conit of doubly fed induction generator, but it ha not alway thi way [8]. Figure 1 how the percentage of the different type of wind turbine intalled each year from 1995 through There are four type of wind unit: type A Fixed Speed (FSG), type B FSG with Variable Rotor Reitance (WRIG), type C Doubly Fed Induction Generator (DFIG), and type D Full Scale Frequency Converter (WRSG).

10 3 Figure 1. Yearly intalled type of turbine by percentage from [8] It hould be noted that traditional wind generating unit (type A&B) do not poe reactive capability. They were, contrat to DFIG unit, reactive power conumer whoe inherent downfall lead to the inability to regulate terminal voltage [51]. To mitigate thi reactive demand, many FSG and WRIG wind park were equipped with external ource of reactive power. Static ource intallation uch a hunt capacitor are relatively inexpenive a compared to dynamic reource uch a SVC [18]. On the other hand, newer wind park coniting of DFIG unit have reactive power capability [19]. The preence of power electronic control in DFIG make them a fat acting dynamic reactive reource a compared to ynchronou generator [20]. Thi allow for voltage regulation and reactive power control by a wind park [21], [22]. A more and more DFIG come online, the threat of increaing level of wind penetration have generated a widepread concern over potential impact to power ytem performance.

11 4 There are primarily two reaon for uch a concern, the variability of the reource and the nature of the generator, both of which imply the retructuring of ytem operation and the incluion of non-traditional ytem equipment [9]. Therefore, the effect of high wind penetration level on ytem performance become a critical factor in power ytem operation and planning of future unit. To accommodate thi new tyle of generation, precie modeling of DFIG unit i important for both tatic and dynamic analyi of power ytem performance [1]. When conidering performance criterion, ytem frequency and voltage level mut be maintained within trict tolerance for proper operation [53]. Compared to other turbine configuration DFIG wind unit poe the uperiority to operate during low voltage condition a well a regulate ytem voltage level [52]. Therefore, it wa decided to tudy the ytem impact of DFIG wind turbine with a focu on the voltage regulating feature within the machine. The following work demontrate reliability improvement in both tatic and dynamic power ytem operation through application of an extended capability curve. It i hown that, contrary to FERC order 661-A, DFIG wind park may and hould operate at much lower power factor without incurring additional cot to the plant owner. Thi extended reactive capability allow for not only improved voltage recovery following a grid diturbance, but may aid in the prevention of a ytem collape. The reult provide valuable information that help to accurately ae the tability of a ytem and to prevent voltage violation with tudie involving high penetration level [10]. It will be hown that impoing thi power factor retriction limit the performance of DFIG wind park, which directly influence the ytem performance. Thu utilizing the DFIG capability curve can lead to improved tatic and dynamic ytem repone a well a reduce

12 5 the amount of committed conventional reactive reerve. The FERC order 661-A give general guideline for interconnecting wind park, but for pecific park employing DFIG unit the retriction on power factor hould be lifted becaue additional performance may be obtained at no extra cot to the wind farm owner. 1.2 Fundamental of Doubly Fed Induction Generator Doubly fed induction generator are a type of variable peed wind turbine that include operational advantage of both ynchronou and induction machine. A with any induction machine variable peed operation can eaily be achieved. In DFIG turbine thi feature allow the capture of maximum wind potential over a wide range of wind peed uing a control trategy termed maximum power tracking (MPT). Thi tracking characteritic i accomplihed by regulating the power extracted by the rotor blade at a given operating peed from the wind [54]. The MPT uperiority i oberved economically when compared to a fixed peed turbine with upward of a 10% increae in energy efficiency [1]. Figure 2 lay out a high level diagram that contain the ytem and ubytem involved with a DFIG machine. A with other electrical machine, DFIG have both electrical and mechanical component. The main mechanical ytem contain rotor blade that are fixed to a prime mover that are controlled with a pitching mechanim to adjut the aociated aerodynamic characteritic. The prime mover then connect to a tep up gearbox that drive the generator. The electrical ytem i compried of a wound rotor induction generator and power electronic converter (PEC) control that function a the generator exciter [55].

13 6 Figure 2. High-level ytem overview of a DFIG wind turbine Since the PEC act a the machine excitation ytem, complete control of reactive power i handled here. Thi i accomplihed through a back-to-back converter that couple the rotor winding of the machine to the grid ide of the machine. Figure 3 diplay the converter architecture that conit of IGBT controlled witche that have a common DC bu linking the AC circuit [56]. Figure 3. Power electronic converter architecture A uch the PEC i compried of two independently controlled three-phae inverter that can function in four-quadrant operation. The name of thee converter will be referred to a the RSC and GSC, which are the repective rotor and grid ide converter. The back-to-back

14 7 RSC and GSC configuration allow for bi-directional power flow between the rotor ide winding and the tator ide output terminal of the machine. Thi ytem of power tranfer reemble an HVDC et up wherein two aynchronou ytem are interconnected. Thi veratility lend itelf to the nature of the machine a the mechanical peed i contantly changing with fluctuation in the wind. Therefore, in order to maintain an internal rotating magnetic field with repect to the grid frequency (a in ynchronou machine) the RSC i contantly adjuting it current injection to the rotor. Thu, excitation of the generator i upplied by the RSC hown in Figure 4. The frequency of the current injection i defined by equation 1. ω = ω + ω (1) r m In thi manner the RSC maintain a rotating magnetic field at ynchronou peedω by injecting current at a frequencyω given the mechanical rotation of the machineω. r m Figure 4. DFIG high level power balance chematic Figure 4 alo how the flow of power within the machine, where the total mechanical, tator and rotor electrical power are repectively defined by P m, P, P r. Since the generator i an induction machine the lip () i given by:

15 8 p ω ω m = 2 (2) ω where p i the number or pole. Relation of the above power can be approximately expreed uing the lip of the machine to derive the total electrical real (P tot ) power output. P tot = P + P P P ( 1 ) (3) r m The next chapter will preent detail regarding the derivation of the total electrical active and reactive capability from the machine and power electronic converter.

16 9 CHAPTER 2. STATIC MODELING OF DFIG WIND PLANTS Reference [34], [35] make claim of 20% penetration level from wind reource in the U.S. At thee level, mirepreenting the actual power capability of DFIG plant at high penetration level may caue large dicrepancie between the actual power flow and mirepreent limit on the power ytem. Although DFIG park may be modeled in variou way, conidering the true reactive ability of the plant i paramount for tudying future ytem operation. Power flow imulation are ued to derive bu voltage and network line flow for given load condition. Typically generating unit are modeled with a finite operating region for real and reactive power output. In recent year appropriate repreentation of wind plant have been of increaing interet. 2.1 Introduction Proper machine modeling i neceary to undertand the actual reactive power capability of a DFIG wind park. Thee machine can be repreented uing the implified traditional induction machine T-model (figure 5) with an additional voltage ource connected to the rotor. Thi voltage ource repreent the rotor ide converter (RSC) that excite the machine. Figure 5. DFIG tatic machine model

17 10 In the above figure the tator and rotor voltage (V, V r ) and flux ψ, ) equation can ( ψ r be derived from the machine current (I, I r ) [25]. In equation 4-7, (R,L l ) and (R r, L lr ) are the repective tator and rotor reitance and leakage inductance with L lm defining the magnetizing inductance. V = R + jω L ) I + jω L ( I + I ) (4) ( m r V r R r = + jω L r I r + jω L m ( I + I r ) (5) = L I L lm I ψ (6) r r r lm r ψ = L I L I (7) The grid frequency i given by w, the machine lip i, and L =L l +L lm and L r =L lr +L lm. The RSC ize i computed from the rating of the rotor ide voltage ource by combining the above baic KVL loop equation with the tator and rotor flux equation. Eliminating the tator flux from thee equation, expreion for the rotor current, voltage, and converter MVA rating can be computed [25]. I = ( ψ L I ) L (8) r r m r V r = ψ r[( Rr Lr ) + jω] [ I ( Lm Lr ) Rr ] (9) * S = 3( V I ) (10) r The converter that excite the DFIG machine i contructed with a back-to-back power electronic converter (PEC) that i fed from a DC voltage ource [26]. Given the r r

18 11 highet AC voltage connected to the bi-directional converter the DC link voltage can be obtained uing the following relation where P m i the modulation index. V = 2 2 dc link ac (11) 3P m V DFIG Plant Capability Curve It i well known that electro-mechanical machine have inherent limitation that allow for a fixed amount of power production. The operating characteritic of any generator are important for repreenting a machine true power capability. The limitation that define a DFIG electrical power capability are influenced by two factor the generator and power electronic. Referring to the maximum power tracking characteritic (Appendix A) it hould be noted that only a finite amount of power i able from a given wind peed. Therefore the real power limit are et by the availability of the wind. The maximum reactive capability of the machine i determined by the generator deign where the applied current and voltage et limitation on the tator and rotor. Therefore, the maximum power capability i completely limited by the deign of the machine although the back-to-back power converter, which will be addreed in ection 2.1.4, define the actual capability. A capability curve for a DFIG wind park wa formulated uing the method followed in [31] with a maximum power tracking characteritic given in Appendix A. Thi technique i given for only a ingle machine, but it i aumed that the power capability of one machine can be caled up to accurately aggregate the behavior of a DFIG wind park.

19 12 Figure 6. DFIG wind park tatic power capability curve in per unit Thi i made under the aumption that the DFIG wind park network i not conidered uch a each machine feeder line and tranformer. Additional impedance may be added to the model, which account for thee implification [32]. The plot in figure 6 diplay the operation of a DFIG within the pecified 0.95 leading and 0.95 lagging power factor. Superimpoed i the capability curve for the DFIG at different wind peed correponding to variable level of power output. Given in the plot are the capability curve for lip 0.25, 0.1, -0.05, and Thi pan the entire pectrum of wind peed from cut-in peed that correpond to 0.25 lip to jut before cut-out peed that correpond to lip. Thu by utilizing the capability curve in network analyi additional reactive power and hence improved power ytem performance may be attained over a regulated power factor. It i evident from the figure that at 100% plant output the ue

20 13 of the capability curve doe not give much additional reactive upport compared to the 0.95 leading operation. In contrat additional reactive conumption may be realized in lagging operation. Wind park will never continuouly operate at 100% output and therefore in the period of operation below 100% there i ignificant additional reactive power available that could aid in improved ytem performance DFIG Wind Park Load Flow Model For any power ytem analyi it i important to appropriately model the characteritic of a ytem device. In the cae of DFIG machine and machine that make up the repone of a DFIG wind park, model for tatic analyi are till being developed. The following ection will briefly dicu everal model repreentation of DFIG wind plant Negative Load Repreentation One of the mot implified DFIG wind park repreentation can be defined by a negative load [36]. During the load flow thi repreentation will inject real power (P) and either leading or lagging reactive power (+/-Q) into the grid. The farm i modeled a a PQ bu. Thi model aume that the wind farm operate at a fixed power factor and cannot regulate it reactive output Synchronou Machine Repreentation A wind park may be modeled a a ynchronou generator with either fixed real and reactive power limit or by employing a capability curve. Both model are repreentative of a PV bu that contain terminal voltage control. The capability curve i the more accurate repreentation. The only diadvantage of thi trategy i that the teady tate et point cannot

21 14 directly be ued in dynamic imulation. Therefore, the initial et point mut be recomputed for the dynamic DFIG model DFIG Repreentation Several oftware manufacturer now appropriately account for DFIG wind park model in load flow tudie. Notable are PSS/E [37], [38], PowerFactory [30], and Eurotag [39] imulation oftware for including model that can automatically be initialized for dynamic analyi. For the trict purpoe of teady tate analyi, ynchronou machine repreentation can accurately portray the behavior of a wind plant. A uch, employing a DFIG capability curve uing a ynchronou machine model prove advantageou when tudying high level of wind penetration Wind Park Aggregation Although model of DFIG turbine have been well tudied [27] there i no indutry tandard and a uch each oftware package may contain it own DFIG model. Moreover ince larger cale wind park will contain upward of everal hundred unit, modeling of individual unit for power ytem dynamic tudie would reult in large imulation time and require greater computation capability. Therefore model implification and park aggregation i jutified in certain power ytem analyi. The incluion of a wind farm into a power ytem for imulation purpoe i often bet repreented by a condened model. Aggregation technique of variable peed wind turbine have been thoroughly dicued and their ignificance decribed in [28]. Studie

22 15 comparing the reult between detailed and aggregated model conclude that an aggregated electrical ytem and non-aggregated mechanical ytem i an efficient and accurate model for mid and long term imulation [29]. For hort-term imulation both electrical and mechanical ytem may be aggregated. For the purpoe of dynamic imulation, an aggregated park of approximately 100 DFIG unit wa modeled in PowerFactory imulation oftware by DIgSILENT [30]. The park wa contructed with a fully aggregated technique that condene the behavior of each individual turbine electrical and mechanical model into a ingle machine model repreenting both electrical and mechanical characteritic Reactive Capability Validation The DFIG park ued in thi tudy are 1.5 MW unit that contain a power electronic converter rated to 30% of the machine rating (Appendix B). Thi aumption i jutified by calculating and comparing the required PEC rating neceary to operate between a +/ power factor a well a over the entire range of the capability curve in figure 6. The procedure outlined in [25] wa followed to compute I r, V r, V dc, and S conv correponding to everal operating point given repective active and reactive power. Row 1-5 of table 1 detail the converter calculation for everal operational point taken from the DFIG capability curve. Row 6 how the rating neceary to operate at 0.95 leading power factor at rated output. Oberve in figure 6 that at each real power (P) operating point the reactive limitation (Q) of the retricted power factor regulation i le than the capability curve (except near 100% real output). It i apparent in table 1 that a the DFIG active output (P tot )

23 16 increae likewie doe the current magnitude (I r ). Near 50% output the machine reache it ynchronou peed and the voltage applied to the rotor i minimal and therefore the required converter rating i at it lowet. Near 100% output the required rotor current, voltage, and converter rating are at their highet value. Thi implie that the leading reactive output (Q tot ) determine the converter rating and ize at 100% output. Hence, only the maximum operating point for the power factor regulation wa diplayed. Table 1. Converter Sizing for Theoretical Reactive Operation P tot [p.u.] Q tot [p.u.] lip [%] V rotor [V] I rotor [A] V dc-link [V] S converter [kva] The required DC link capacitor voltage baed on the tator ide voltage i calculated uing equation 11 with a maximum modulation index of 0.9. The tator ide rated voltage i 575 V and therefore the correponding DC link voltage i 938 V. From the above table the required V dc-link from the rotor ide voltage i lower compared to that required by the tator ide voltage. Thi implie that no change in the actual DC link capacitor voltage rating i neceary to implement the capability curve. The implemented PEC rating wa derived uing a margin of afety baed on the highet kva rating from table 1. The DC capacitor voltage wa deigned from the tator ide voltage and rated to 1150 V with the PEC rated at 450 kva.

24 17 It i very important to note that operation of a DFIG with a power factor regulation mut produce a Q tot of 0.33 and for the capability curve it mut produce 0.37 at rated active power. Comparing I r and S converter for row 5 and 6 in table 1 how that only a 2% increae in the rating i neceary to implement thi capability curve over a regulated power factor cheme. Therefore converter intalled in operational and newly commiioned DFIG wind farm abiding by the FERC 661-A order have additional reactive capability that may be utilized. Thi demontrate that operating the DFIG within a power factor regulation greatly under-utilize the machine overall reactive ability.

25 18 CHAPTER 3. DFIG PLANT DYNAMIC MODELING AND CONTROL To acertain the dynamic performance of a power ytem with large amount of DFIG wind penetration, appropriate modeling of DFIG wind park i neceary. Each machine in a wind park ha everal ub ytem model in order to accurately reproduce the repone of each machine. Since wind park are made up of any where from everal to hundred of turbine, tep mut be taken to lump the individual repone into one large or aggregated model that repreent the behavior of the wind park. The ub ytem that will be dicued in thi chapter are the aerodynamic, mechanical, electrical, and control ubytem that comprie the DFIG wind turbine ytem. The block diagram of the DFIG machine that relate the all the ytem and ub ytem i diplayed in Figure 7. Figure 7. DFIG Main Block Diagram [30]

26 Dynamic Modeling Principle All wind turbine perform the function of converting mechanical to electrical energy for tranportation via the electric tranmiion network. The energy converion proce begin with the energy input ource: the wind, which obey certain aerodynamic characteritic. The following ection introduce the principle of how a turbine extract power from the wind Aerodynamic Characteritic A wind flow pat the blade of a machine, whether rotating or tationary, mechanical lift i produced and an aerodynamic torque (T ae ) i applied to the blade. Depending on the rotation of the blade, thi torque produce a mechanical power ( P m ). The mechanical power that the turbine extract from the wind and applied to the electrical ubytem for converion i given by the following relationhip P m 1 2 λ β πρ 2 3 = cp (, ) air R V w (12) Thi power (Pm) i function of the peed of the wind (V w ), the blade radiu (R), the denity of the air ( ρ air ), and the performance coefficient of the rotor blade (cp). Due to the phyical nature of the energy capturing proce, not all the power available from the wind may be extracted by the turbine blade a decribed by Betz law [57]. Thi fact i quantitatively illutrated a the performance coefficient of the rotor (cp). It i a manufacturer pecified parameter that i determined from a method outlined in [57]. The performance coefficient i a highly non-linear variable defined by a function baed on the parameter ( λ, β ) a hown in 12.

27 20 The tip-peed ratio (λ) i given in equation 13 a the peed at the tip of the rotor blade divided by the uptream wind velocity (V w ) whereω i the angular peed of the turbine rotor. ω w r, tur r, tur λ = (13) The parameter β i the pitch angle of the rotor blade with unit in degree. The pitch angle of the turbine blade can be controlled to influence the value of cp, which in turn limit the amount of mechanical power available for extraction from the wind. V Figure 8 preent a graphical repreentation of cp with repect to a continuouly varying λ over different dicrete value of β. From thi plot it can eaily be oberved a the pitch angle ( β ) increae the performance coefficient i reduced and thu decreae the extracted power from the wind. S. Heier explain how to analytically calculate the performance coefficient of a wind turbine [57]. R Figure 8. Rotor performance coefficient (cp) v. tip-peed ratio ( λ ) for variou pitch angle ( β ).

28 21 Uing thi method, equation 14 and 15 were derived and the above rotor performance coefficient curve were developed for the DFIG model ued in thi tudy. It hould be noted that a turbine mechanical power depend on cp and therefore i deemed a deign parameter. C p λi = e ( λ, β ) β β (14) λi λ 1.01 = 3 λ 0.002β β + 1 (15) i Since the calculation of thee curve can be computationally cumberome with an explicit function, dicrete point from ( λ, β ) indice may be defined by a two-dimenional lookup table (Α). λ β λ, β 1 cp (, ) = Α (16) For value not located in dicrete entrie of the table an interpolation method, uch a a cubic pline function, may be ued to determine thee point. Thu, the aerodynamic curve developed in MATLAB were then entered a dicrete input point into the dynamic imulation oftware Mechanical Drive Train Characteritic The mechanical ytem of the Doubly Fed machine ued in thi project only model the dynamic repone of the drive train inertia. Thee implification are made due to the fact that the drive train ha the mot ignificant impact on power fluctuation [59]. Any other mechanical element are thu not conidered in thi model. Additionally, the mechanical

29 22 model i more general in nature and may apply to any type of wind turbine. A uch, the purpoe of the drive train i to convert the aerodynamic torque (T ae ) from the wind into mechanical torque on the low peed haft (T haft ) that couple to the generator via a gearbox [58]. The gearbox i aumed to be lole with a gear ratio of 1:n gear. PowerFactory imulation oftware conider a two ma lumped model with one large ma ymbolizing the rotor inertia and a maller ma repreenting the generator inertia [55]. Thee general detail regarding the drive train lead to expreion for the ytem of equation that make up the mechanical model [60]. 2 H & ω = T kθ D ω (17) m g m g m mg mg e g m 2 H & ω = kθ T D ω (18) g m & θ = ω ω ω ) (19) mg o ( m g T m i the accelerating torque from the wind and T e i the decelerating electric torque. K i the haft tiffne and θ mg i the angular torion in the haft. Dmω m and D ω are the damping g g torque of the turbine and generator where ω ω, ω o, are the ytem, rotor, and generator m g peed. Thi ytem of differential equation comprie the two-ma drive train model ued in the imulation Generator Characteritic The electrical ytem i made up of a ytem of differential equation modeling the dynamic repone of the DFIG machine. The differential equation that make up thi machine are very imilar to that of a traditional induction machine, but do not contain hort-

30 23 circuited rotor winding. Intead a voltage ource generated by the RSC i connected to the rotor winding of the wound rotor induction machine. A fifth order et of differential equation can be written in dq0 coordinate a diplayed below. The voltage, current, and flux indice are referred by the direct (d) and quadrature (q) axe while () and (r) repreent the repective tator and rotor quantitie [60]. V d = R i ω ψ + ψ& (20) d q d V q = R i ω ψ + ψ& (21) q d q V dr = R i ω ψ + ψ& (22) r dr qr dr V qr = R i ω ψ + ψ& (23) r qr dr qr 1 ω& = ( T m T e ) (24) 2 H Of thi et of equation 24 decribe the motion of the machine. From expreion the flux linkage can be derived a hown in ψ = ( L + L ) i L i (25) d m d m dr ψ = ( L + L ) i L i (26) q m q m qr ψ = ( L + L ) i L i (27) dr m dr m d ψ = ( L + L ) i L i (28) qr m qr m q Combining and if the tator tranient, the flux derivative in 20-23, are neglected algebraic definition for voltage may be derived [61]. v = R i + ω L + L ) i + L i ) (29) d d (( m q m qr

31 24 v = R i ω L + L ) i + L i ) (30) q q (( m d m dr v = R i + ω L + L ) i + L i ) (31) dr r q (( r m qr m q v = R i + ω L + L ) i + L i ) (32) qr r qr (( r m dr m d Vector Control Conider a rotating reference frame rotating in at the ame angular velocity ( ω 1) a the power grid ( ω ). Reviiting equation et 4-7 and rewriting expreion 4 and 5 in the time domain the following tator and rotor expreion may be derived [25]. I R + V = ψ& jω ψ (33) I r R r + V = ψ & j( ω ω ) ψ (34) r r r r Under the ame aumption above, the d axi of the d-q coordinate i elected to be in line with the tator fluxψ. Now, ψ d remain contant ince the tator flux i contant, and thu ψ& = 0 andψ = 0. Thee term can be input to the above expreion (33,34) to obtain d q V d = 0 and Vq = ω 1 ψ d. If the tator flux equation are rewritten in term of I then we have I L I m =ψ r where l m L L = L + L (35) Combining the voltage and flux equation (25-28 and 29-32), one can generate expreion that relate dq voltage and current quantitie to real and reactive power output of the machine. P = v i + v i + v i + v i (35) d d q q dr dr qr qr Q = v i v i + v i v i (36) q d d q qr dr dr qr

32 25 Under the aumption above the tator real and reactive power may be formulated a follow. P Q Lm = 3/ 2( Vd I d + Vq I q ) = 3/ 2( Vq I q ) = 3/ 2( ω1ψ d I qr ) (37) L Lm = 3/ 2( Vd I d Vq I d ) = 3/ 2( Vq I d ) = 3/ 2( ω 1ψ d ( ψ d I qr ) (38) L It hould be noted that the expreion for the tator real power i regulated by the I qr component wherea reactive power i controlled by I dr. Now if the rotor flux equation (27 and 28) are ued by ubtitution into 35 for I the following can be derived. 2 L Lm ψ + ( Lr ) I r (39) L m r = ψ L If expreion (34) i conidered in the teady tate ψ& = 0 and uing equation (39) with r L c 2 Lm = ( Lr ), the rotor voltage equation may be derived a follow. L V = R I + L ( ω 1 ω ) I (40) dr r dr c r qr V qr Lm = Rr I qr ( ω 1 ωr ( LcI dr + ψ d ) (41) L From the above equation it hould be oberved that applying appropriate control to the rotor ide converter can produce the neceary voltage to control the tator power. Therefore, the active and reactive power of the DFIG machine may be controlled with the decoupled vector control trategy.

33 Pitch Control Characteritic The pitch controller of a DFIG wind turbine i ued to regulate the mechanical power extracted from the wind. Thi trategy become ueful when mechanical power input to the turbine become exceive at high wind peed. Reviiting expreion 12 it i apparent that the mechanical power can be approximated a the cube of the wind peed. Thu a mall change in wind peed produce a large change in mechanical power. In order to help control the turbine input within it deign pecification the performance coefficient (cp) in 12 mut be regulated. It wa dicued earlier that the function for cp i derived with two variable, ( λ, β ). The pitch controller can be repreented by the tranfer function in a block diagram a hown in Figure 9. The entire ubytem repone i compried of a PI controller that govern the output of the controller, which i then feed into a pitch-actuated ervomotor. Figure 9. Pitch Controller Block Diagram [30]

34 27 The pitch controller modeled in PowerFactory i a generic model that doe not repreent any individual turbine manufacturer. The model wa adjuted to produce a reaonable repone a compared with other documented reult [59]. Detail regarding the deign of the pitch controller can be found in numerou literature [62]. 3.2 Dynamic Model Validation All of the following imulation tudie were baed on the GE 1.5 xle MW wind turbine. All data regarding the electrical and mechanical machine model can be found in Appendix B. The default model given in DigSILENT PowerFactory wa a 7 MW DFIG wind turbine preumably repreentative of an off hore machine. In order to appropriately addre the effectivene of adjutment to the default model, a wind peed ramp wa applied to the newly configured 1.5 MW machine. Thi tet would demontrate the accurate operation of the machine by treing the electrical converter and blade pitch control ytem from cut-in to cut-out operation Dynamic Model Validation The below imulation were run to model 400 econd or approximately 6.5 minute of a contant linearly increaing wind gut. Thi wind gut wa made to begin at cut-in peed around 4 m/ and increae until cut-out wind peed of 20 m/. Figure 10 contain eight time imulation plot that contain variable of interet to validate the imulation over the 400 econd. The output i compried of 100 wind turbine, which repreent the aggregated repone of a wind plant and not jut one unit.

35 28 A the wind ramp gradually increae it can be noticed that the performance coefficient increae from 0.05 at cut-in until around 0.40 at rated wind. At thi point the pitch controller activate at around 150 econd and cp begin to exponentially decay until cut-out wind peed. Throughout the entire imulation the total electrical power output of the machine poitively increae from around zero. It hould alo be noted that the rotor power i negative during the ub ynchronou operation and poitive for uper ynchronou operation.

36 29 Figure 10. DFIG Dynamic Model Validation And at exactly 1.0 pu haft peed (ynchronou peed) the rotor power i 0. The haft peed cut in around 0.75 pu and hit rated peed around 1.25 pu. Thi correpond to the +/-

37 30 25% peed range deigned for thi machine. Thu baed on the reult in Figure 10 it wa determined that a wind plant model had been appropriately developed for the cope of thi project. 3.3 DFIG Control Enhancement Thi ection provide information regarding the control tructure of the power electronic converter and control enhancement to the default model. The control over the power electronic converter i independent on both the rotor and grid ide converter. Information regarding detail on the control theory behind thi can be found in literature [63] Electrical PEC Control Structure The control tructure ued in the PowerFactory model i depicted in Figure 11. Thi figure how three ubection control function. The Current and Power ubection were default control loop for the PowerFactory DFIG wind turbine model. The Protection ubection control loop wa developed a a part of thi work and i meant to be a mean of protecting the machine. The block labeled DC OV, GSC Boot, and Re-trip will be covered in the following ection. DC OV i an overvoltage protection circuit and control logic built into the PEC DC link. The GSC Boot allow for the grid ide converter to be utilized to provide reactive power when the RSC circuit i in a protected tate. And the Re-trip i another hardware circuit and logic to prevent the default DFIG protection circuit from retriggering in the preence of a evere diturbance.

38 31 Figure 11. DFIG Power Electronic Control Block Diagram Controller Development In thi ection the development of enhancement made to the DFIG wind turbine control tructure will be covered. The control development are in regard to improvement made to the machine, which increae the likelihood of the plant to contribute to ecure operation of the ytem. In order to tet the claimed improvement to the control tructure the

39 32 DFIG wind plant wa ubject to a evere diturbance a defined by the Federal Energy Regulatory Commiion (FERC) Zero Voltage Ride Thorough A mandated by FERC Order 661-A wind-generating plant are to remain online in the preence of evere voltage diturbance for a defined period and voltage profile [24]. Thi ection will addre iue regarding the Pot-tranition Period that take affect on all newly intalled wind generation from January 1, 2008 to preent. The regulation tate that wind plant are to remain on-line during three-phae fault with normal clearing time (4-9 cycle) and ingle line to ground fault with delayed clearing. The pot-fault voltage recovery mut return to the pre-fault voltage level. The clearing time requirement for a three-phae fault will be determined by the tranmiion provider and may not have a clearing time greater than 9 cycle (150 m). In the event that a fault remain greater than the clearing duration or the pot-fault voltage doe not recover above the determined value, the wind park may diconnect from the tranmiion ytem. Wind generating plant mut remain connected during uch fault with voltage level down to 0 volt, a meaured at the high ide of the POI tranformer. Figure 12 ummarize the fault ride through criteria graphically.

40 33 Figure 12. FERC Order 661-A (Black) v.. POI Bu Voltage (Red) All DFIG control enhancement were teted with a 3-phae hort circuit at the high ide POI uing the tet ytem decribed in Chapter 5. The following ection elaborate on each control enhancement individually and then draw a comparion between the imulation model before and after the control improvement Protection Repone of DFIG Wind Turbine The main problem inherent in a DFIG machine i the current enitivity of the IGBT that make up the power electronic converter [66]. Thee device may be ubject to damage if converter current limitation are exceeded. Since the RSC of the PEC i connected to the

41 34 rotor via lip ring a crowbar circuit may be hort circuited in parallel to the rotor winding a depicted in Figure 13. Figure 13. DFIG Crowbar Protection Repone Schematic Thi action electrically iolate the RSC from damaging tranient current that may be induced into the rotor wind from the tator ide of the machine during a diturbance. Figure 14 how a time dependent imulation of the current through the RSC from pre to pot fault. It hould be noted that at time 0.0 a hort i placed at the high ide POI. Thi action induce large current tranient in the RSC jut before the RSC protection i triggered. Small tranient are oberved at 150m when the protection circuit i cleared and the RSC reynchronize the machine with the ytem. The tranient current are thu maller a the bu voltage ha returned to it pre-fault tatu.

42 35 Figure 14. Current Flow in Power Electronic during Crowbar Operation Although thi protect the RSC from potentially damaging current, it unfortunately diable all excitation and power control that i needed for operation of the DFIG. During the protection period the machine effectively i functioning a a traditional induction generator that become an inherent reactive power conumer due to the nature of the machine. Thu, control improvement are preented which are ued a an aid to overcome ome of thee hortcoming dicued during fault condition in the machine Grid Side Reactive Power Booting During a grid fault where the RSC converter i diconnected from the rotor, a decribed in the previou ection, the machine loe all controllability of real and reactive power. During thi protection period the Grid Side Converter may be controlled to provide a

43 36 reactive power injection a depicted in Figure 15. Uing thi trategy effectively utilize all component of the PEC ytem and increae the internal bu voltage within the wind farm collector ytem. Figure 15. Coordinated Converter Control Schematic during Crowbar Operation During thi period the GSC i effectively functioning a a STATCOM device that may inject or conume a dynamic ource of reactive power baed on the local ytem need [66]. Figure 16 how the imulation reult where the zero-voltage FERC fault criteria were applied and the DFIG protection ytem triggered a noticed by the diminihed collector ytem terminal voltage. The upper graph in the figure how the intantaneou repone in the GSC reactive injection with and without the GSC reactive power booting control trategy.

44 37 Figure 16. Impact of Grid Side Reactive Booting with (black) and without (red) Control The redlined imulation are uing the default DFIG plant model wherea the black-lined reult are after the reactive booting wa added. It can be noted that an approximate increae of 7% in the machine terminal voltage wa attributed to the GSC booting during the hortcircuited crowbar protection period Crowbar Protection Re-trip Prevention A noted earlier the natural repone of the DFIG protection upon placing and clearing a fault i induced current tranient from the tator to rotor. The tranient current upon fault clearing are much le than the fault initiation, but nonethele till preent in the

45 38 rotor and poibly through the RSC. Under certain condition the tranient current produced during the protection clearing ha been hown to re-trigger a diplayed in Figure 17. Figure 17. PEC Current with and without Crowbar Re-trip Prevention The redlined reult i the current flow through the RSC before, during, and after the fault without any control modification to the exiting DFIG model. If uch a cenario were to occur the DFIG plant would remain inoperable for another 150m until the protection would again reynchronize with the grid. Thi utained lo of generation could inevitably repeat itelf or have already produced an uncontrolled deterioration to the ytem during heavily loaded period. Either ituation may prove devatating; a uch the idea of a witched erie reitance wa mentioned a a potential method to contribute to fault-ride through [67]. Figure 18 how the chematic repreentation along with the traditionally included crowbar

46 39 protection. That tudy concluded that hardware cot would out way the benefit of uch invetment and wa not purued any further. Figure 18. Crowbar Re-trip Prevention Implementation Intead thi invetigation delve into the poible effectivene of thi implementation while neglecting the cot burden aociated with the equipment. A imulation equivalent of the hardware hown in Figure 18 wa programmed into the PowerFactory default model and teted with a hort circuit at the POI. The poitive effect of reducing the tranient rotor current wa oberved in the previou Figure 17. It hould be noted that upon reynchronization at 150m the RSC rotor current (black line) ha a ubtantially reduced magnitude a compared with the original model. Thu the erie crowbar implementation complement the already exiting parallel crowbar to prevent the re-trip of the machine protection circuit.

47 DC Link Capacitor Over Voltage Mitigation The DC link capacitor in the PEC circuit i integral for bidirectional power flow between the rotor and grid. Proper voltage level mut be maintained at the DC link in order to effectively control the excitation of the machine and to avoid damage to the capacitor. It ha been documented that during certain condition the DC link may encounter a potentially diatrou cenario when the RSC protection i triggered [64]. Figure 19 how the voltage magnitude (red-lie) of the link capacitor when the crowbar trigger due to a 3-phae hort at the POI for 150m. It hould be noted that nominal DC link voltage i at 1.3 kv and the tranient voltage that occur produce level of around 1.9 kv. Figure 19. DC Link Capacitor Voltage with and without Chopper Control Thi cenario preent itelf when the machine i functioning in it ub-ynchronou region of operation at high lip. In thi regard there i a relatively large amount of energy tranfer from the grid through the PEC and to the rotor. When the RSC crowbar trip, the

48 41 GSC control cannot intantaneouly repond which caue overcharging of the capacitor. Thi ituation can be damaging to PEC circuitry. To mitigate uch a problem the chopper circuit hown in Figure 20 ha been implemented a found in literature [64]. Figure 20. DC Link Chopper Circuit Schematic The control methodology contantly ene the DC link voltage and activate the chopper circuit at a turn on / off threhold. The voltage level (black-line) in Figure 19 how the reduced voltage tranient the reult from the implemented chopper hardware and control logic uing PowerFactory.

49 42 CHAPTER 4. VOLTAGE SECURITY ASSESSMENT METHODOLOGY Reactive power i eential for the table operation of the power ytem. It facilitate flow of active power from generation ource to load center [11]-[13] and maintain bu voltage within precribed limit [14]. Stable operation of power ytem require the availability of ufficient reactive generation [15]. Figure 21 how a implified two-bu power ytem model that demontrate the relationhip between real and reactive power flow given the machine angle and terminal voltage. Auming negligible line reitance and tranforming the above figure uing a Thevenin equivalency the following relation are derived. Figure 21. Simplified two bu power ytem The ending and receiving end terminal voltage are given by, E and E r with an equivalent line reactance (X) and angle of difference between the machine (δ ). From thee

50 43 definition the real and reactive power tranfer (P, P r ) acro the line can be define by equation E Er P = Pr = inδ (35) X Q r = E E r coδ E X 2 r (36) Q E E Er coδ = (37) X 2 Notice that if the voltage magnitude are fixed the real power tranfer i predominantly governed by the tranmiion angle (δ ). Converely, the terminal voltage trongly influence the reactive power tranfer. Therefore, appropriate reactive capability of machine i eential to maintaining proper voltage level. Although the above approximation are derived from teady tate condition, the availability of dynamic reactive power ource following a diturbance play an important role in voltage tability [16], [17]. 4.1 Introduction In thi chapter a methodology i outlined which wa developed for aeing the ecurity of a power ytem with high wind penetration. Thi framework i a two-tage proce that firt determine the maximum wind penetration allowable into a given location from tatic analyi. The econd tage of the proce conider the power flow olution and analytical reult from the firt tage in order to perform dynamic imulation of the ytem. The dynamic part of the methodology aee the integrity of the ytem following a elect lit of critical contingencie.

51 44 The two-part procedure ue indutry grade reliability tandard a a metric for evaluation of ytem health. Following the completion of the outlined methodology, it can be determined if the propoed wind penetration level are valid baed on the voltage ecurity aement Static Security Aement for High Wind Penetration A Voltage Security Aement (VSA) wa developed uing criteria et forth by the Wetern Electricity Coordination Council (WECC) and North American Electric Reliability Corporation (NERC) [65]. The VSA methodology integrate a tatic ecurity tandard that uniquely aid in identifying ytem deficiencie when large wind penetration i preent. Dynamic imulation are then run to validate that the propoed penetration level will be appropriate for the power ytem. A detailed decription of how to perform the VSA will be introduced next Methodology Decription The VSA i formulated around a teady tate power flow olution that i deemed ecure by WECC tandard. Thi proce i the heart of the VSA, a the dynamic are only to validate an appropriate level of wind penetration. The teady tate analyi i ued to identify the level to be teted. The following decription will elaborate on the VSA flow chart depicted in Figure A matrix cae lit of penetration level and park output mut be defined which will be teted on the bae cae ytem. Table 2 define 16 poible combination

52 45 of bae cae cenario that could potentially be elected. There are four penetration level (15%-30%) containing four plant output (0%-100%) that correpond from cut-in to cutout wind peed. All output level hould be run for the lowet penetration level being teted before moving on to the next penetration level. Penetration Level Plant Output Table 2. Bae Cae Scenario Lit 15% 20 % 25% 30% 0% % % % For each of the above 16 cenario a lit of critical contingencie i defined through a creening and ranking procedure [68]. For the purpoe of thi VSA only voltage level are conidered a a criterion for electing the mot crucial outage. All generator and line outage were conidered a poible n-1 contingencie. To creen the n-1 outage, the bae cae bu voltage were compared againt the repective bu voltage in each n-1 cenario and ranked in order of magnitude. A elect number of the larget voltage difference were ued for each of the 16 cae cenario.

53 46 3. PV analyi i performed uing the identified lit of critical contingencie [68]. At each incremental increae in load all critical contingence are run and all performance metric are checked againt the repective criteria limit. The PV analyi will proceed until a violation i reached. Thi tudy conidered proportionally increaing the entire ytem load although a ingle tudy area may be of interet for large ytem. 4. Stopping criteria for the PV analyi i baed on a combination of limit including WECC teady tate voltage criteria and a atifactory power tranfer margin of the ytem. A tranfer margin hould be determined baed on engineering judgment and a uch, the bae cae cenario without wind wa ued a the metric of comparion. 5. After completing the cenario for all plant output level and all given penetration level, hunt placement can trategically be determined. For each PV analyi of cenario, a certain load level wa determined. The load level increae wa topped due to voltage violation at one bu. Thi bu i then deemed the critical bu and each cae will contain one. Again uing engineering judgment or given a repeated critical bu baed from all cenario, hunt compenation of a elected amount can then be applied.

54 47 6. When the PV analyi for one cae i completed a uitable tranfer margin and voltage level are aumed to be etablihed. Thu the next penetration and output cae file from the initial cae lit matrix hould be loaded and run. Thi procedure hould be repeated until all cae have tried with the VSA method. 7. The maximum penetration level can then be identified for the teady tate. Thi penetration level i determined baed on a uitable tranfer margin from each of the teted output level for a given teted penetration level. Here i a ample example. For a teted penetration level of 25%, tranfer margin of 12%, 25%, 28%, and 20% were found from each of the four plant output percentage. Therefore, if an acceptable ytem tranfer margin of 10% i allowable, then a 25% penetration level i acceptable. But if a tranfer margin of 15% i neceary, 25% wind penetration would no be appropriate ince one of the four tranfer margin wa only 12%. 8. Latly, dynamic imulation of everal critical fault hould be run on the ytem for variou wind plant output level. Thi can validate whether the ytem i table at the now intalled penetration level.

55 48 Figure 22. Voltage Security Aement Flow Chart for High Wind Penetration 4.2 Dynamic Validation of Wind Penetration Level Once a penetration level i defined a ecure in the teady tate, then dynamic imulation of thoe given load and generation condition can be validated. The dynamic ecurity apect incorporate the reult from the tatic aement and therefore may validate

56 49 whether the teady tate wind penetration level i acceptable. Thu evere or critical contingencie are identified and teted with appropriate imulation model.

57 50 CHAPTER 5. DISCUSSION AND ANALYSIS OF RESULTS Thi chapter conit of a two-part dicuion involving the analyi of the reult from implementing the VSA on a tet power ytem. Section 5.1 cover the tatic reult of the Voltage Security Aement and 5.2 detail the dynamic validation of the elected penetration level from the VSA. The Voltage Security Aement methodology wa performed uing the VSAT oftware program found in the DSA Tool package developed by PowerTech [68]. The dynamic validation of the choen wind penetration level wa performed in PowerFactory oftware imulation package developed by DIgSILENT. 5.1 Power Sytem Decription A ample power network available in the PSS/E oftware wa imported into VSAT and PowerFactory for ytem analyi. The original network conit of 6 conventional machine and 26 bue. The total load wa modified to 3035 MW and 1230 MVAr with 305 MW of motor load that ha been ditributed between bue 3005, 153, 203. Refer to figure 23 for the chematic of the network. Shunt compenation (950 MVAr um) i located at variou bue throughout the ytem with a large 600 MVAr reactor at bu 151. The tranmiion voltage range from kV and the line parameter have been modified to reflect appropriate tranmiion ditance [11]. In the bae cae the majority of generation i concentrated in the Northern region of the grid. The load center are located in the South and Southeat portion of the ytem with

58 51 major concentration at bue 154 and 206. The Southwet part of the network contain low load and low tranmiion capacity. Typical high wind region have thee characteritic and hence it i aumed a potential ite for large-cale wind facilitie [43]. Figure 23. Simulated Power Sytem with Wind Park Interconnection at Bu 3008 Since one of the underlying theme of thi reearch i to addre the implementation of large DFIG penetration level, unit 3018 ha been taken off line. Intalled in place of thi unit are 3 DFIG wind facilitie trategically placed at bue 3005/ 7/ 8. The replacement of thi unit wa to imulate iolated wind generation that would emphaize the impact of high DFIG penetration on ytem performance. To facilitate the tranfer of energy from thee high wind region to the load center the line ( ), ( ), and ( ) are upgraded to have ufficient tranmiion capacity.

59 Penetration Level Characteritic Additionally, to analyze the impact of increaed DFIG wind penetration, variou penetration level at 15, 20, 25, and 30% are imulated. In thi tudy wind penetration i defined a the total capacity of wind generation compared to the total load. Intalled wind capacity Penetratio n Level = (38) Load Tranfer Margin with Extended Reactive Capability The tranfer margin (TM) of a ytem wa defined by a the percentage increae in load (MW) beyond the bae cae found in the PV analyi. Equation 39 wa ued to define the tranfer margin of a ytem for thi tudy. TM MW PV ba ec ae = (39) MW MW ba ec ae 5.2 Static Analyi Reult The reult from the VSA are preented in the equence that follow the flow from the chart in Figure 22 from Chapter 4. The VSA inner iterative Shunt Compenation loop wa entered on three occaion. Thi method trategically placed compenation that allowed for an increae in the power tranfer margin and thu an overall increae in wind penetration. The bae cae lit of cenario wa compiled baed on the MW injection of the wind park output a hown in Table 3. Although, the MW injection are 0 for each penetration

60 53 level in the 0% plant output cae, thi i doe not mean that the power tranfer of the ytem will be the ame for each penetration. Table 3. MW injection for given wind penetration and park output Penetration Level Plant Output 15% 20 % 25% 30% 0% % % % Since the wind park are modeled after modeled after the GE 1.5 MW wind turbine, reactive capability i aumed even when the plant i not producing real power [68]. It i aumed that reactive production i approximately 30% of 1.5MW due to a partial cale PEC converter. Thi additional reactive capability can be dratically noticed when comparing the power tranfer margin for 15, 20, and 25% penetration level in the 0% output cae in Figure 24.

61 54 Power Tranfer Margin from Increaed DFIG Penetation (No Additional Reactive Support - Limited by 204) Power Tranfer Margin (%) % 20% 15% 20% 25% 13.5% (Bae) 15% 13.5% (Bae) 7.3 0% 33% 66% 100% Wind Power Output (% Nameplate) Figure 24. Power Tranfer Margin from Increaed Plant Output with No Reactive Compenation The firt round of the VSA wa teted with only 15, 20, and 25% penetration level and 30% wa initially neglected. Thi i due to a limited power tranfer margin a can be noticed at 25% penetration with 100% plant output. Therefore, 30% penetration wa not invetigated further for the bae cae with no hunt compenation. The bae power tranfer (or cenario without wind) wa ued a a metric of comparion for all wind penetration cae. Table 4 how the numerical reult of Figure 24. The table how that the power tranfer margin wa lower at 0% output production for both 15 and 20% penetration level a compared with the 13.5% bae cenario. In contrat 25% penetration wa everely limited at 100% plant output. The identified contingency that breached the ecurity contraint wa the lo of unit 206 (contingency A30 or B3).

62 55 Table 4. Power Tranfer Margin at Different Penetration Level with No Reactive Support Penetration Level Plant Output 15% 20 % 25% 0% % % % Following the VSA methodology, an SVC of 50 MVAr reactive upport of wa located at bu 204 a an aid to increae the power tranfer margin. The firt round of the VSA provided valuable information regarding the current ytem ability to handle an increae in DFIG wind penetration. It i obviou that any wind penetration amount between 15-25% would not be feaible from a reliability tand point a the bae cae conventional ytem ha a 13.5% power tranfer margin. Given the new reactive compenation placed in the ytem the econd iteration of the VSA will be elaborated on. After the additional hunt wa placed at bu 204 ubtantial overall increae in power tranfer for all wind penetration level (15-25%) wa noticed. At a 15-25% penetration level, bu 204 again become the limiting factor at 0% output with the lo of unit 206. Therefore, additional hunt compenation may need to be intalled at 204 before reaching higher tranfer margin. Figure 25 graphically depict the tranfer margin for 15-25% penetration againt the bae cae (0% wind penetration) cenario. The additional of 50 MVAR at bu 204 allow for around a low of 14% and high of 19% power tranfer for a 20%

63 56 wind penetration level cenario. At thi level an increae in wind penetration allow for a higher tranfer than the bae cenario. Power Tranfer Margin from Increaed DFIG Penetation (50 MVAr Support at Limited by 3007 & 3008 at 25%) Power Tranfer Margin (%) % 15% 20% 25% 13.5% (Bae) 25% 15% 13.5% (Bae) 9.5 0% 33% 66% 100% Wind Power Output (% Nameplate) Figure 25. Power Tranfer Margin from Increaed Plant Output and 50 MVAr Support at 204 It hould be noticed that a 25% DFIG penetration level i limited by at both 0% and 100% park output paired againt the bae cenario. The limiting bue are found by the VSA to be 3007 and 3008 which coincidently are in the location of the intalled wind facility Following the VSA procedure another 50 MVAr of hunt compenation wa intalled. Thi time the location of the intallation wa at the high ide of the DFIG POI at bu Since the ytem wa oberved to handle 20% wind penetration reliably above the bae cae, the 15% penetration level wa not conidered in the next round of VSA imulation. Table 5 mark the numeric reult from the above Figure 25.

64 57 Table 5. Power Tranfer Margin at Different Penetration Level (50 MVAr at 204) Penetration Level Plant Output 15% 20 % 25% 0% % % % Again, after following the VSA with the placement of 50 MVAR at bu 3008, imulation for penetration level of 20, 25, and 30% penetration were conducted. At thi point after running the imulation with input from the VSA it can be concluded from Figure 26 that the total allowable DFIG wind potential in thi ytem i at leat 20% and no more than 25% penetration. Thi i baed on the aumption that no other form of reactive upport are invetigated. Uing engineering judgment, it wa concluded that a limit of 30% DFIG wind penetration in the tet ytem wa ufficient. Figure 26 how that penetration level in the range of 20-25% have a power tranfer margin equal to or greater than the bae cae margin of 13.5%. Since the VSA ha provided that all n-1 voltage level are atifactory for thee load increae and the 20-25% penetration tranfer margin are equal to or above the 13.5%, thee concluion are deemed valid.

65 58 Power Tranfer Margin from Increaed DFIG Penetation (50 MVAr Support at 204 & Limited by 3007/8 at 25%) Power Tranfer Margin (%) % 25% 20% 20% 25% 30% 13.5% (Bae) 13.5% (Bae) 0% 33% 66% 100% Wind Power Output (% Nameplate) Figure 26. Power Tranfer Margin from Increaed Plant Output 50 MVAr upport at 204 & 3008 It i very intereting to decipher what caued a 0% tranfer at 100% plant output from 30% DFIG penetration. In thi cae the poweflow olution would not converge and wa determined to be untable. Thi wa invetigated further and a concluion wa reached that the reactive capability neceary to tranfer all 700 MW from bu 3008 into the ytem wa not ufficient. Therefore a larger ource of reactive capability would be needed in the local vicinity in order to hip thi amount of power out of the wind corridor. A poible olution that wa not invetigated would be to utilize unit 3018 that wa taken off line a a ource of reactive compenation. Table 6 how the numeric reult from thi lat iteration of the Voltage Security Aement for large-cale wind production facilitie.

66 59 Table 6. Power Tranfer Margin at Different Penetration Level (50 MVAr at 204 and 3008) Penetration Level Plant Output 20% 25 % 30% 0% % % % Untable 5.3 Dynamic Analyi Reult Thi ection provide information regarding the reult of dynamic imulation run on the tet power ytem for a 20% DFIG wind penetration level. Thee imulation are to validate the intallation of a large amount of DFIG unit put in place of unit 3018 a found in the VSA methodology. Thee imulation not only tet the validity of a 25% wind penetration, but alo the performance of utilizing an extended reactive capability curve a decribed in the earlier Chapter 1. Section explain the detail behind the reactive capability comparion for the DFIG plant Voltage Control Application There are two voltage control trategie that are implemented to demontrate a comparion between DFIG park repone on ytem performance. The firt trategy utilize

67 60 the +/ power factor regulation et forth by FERC order 661-A, where the reactive limit are defined by the park real output [24]. Q tan(co 1 max = P output (0.95)) (40) The econd trategy utilize the reactive capability that i detailed in the developed capability curve in figure 6. Both cheme ue a proportional-integral (PI) controller that regulate the POI voltage a outlined in figure 27 [31]. Figure 27. DFIG Wind Plant Voltage Controller Schematic The input to the controller are the et point voltage (V ref ) and the voltage meaurement (V mea ) at the POI. The error ignal between the voltage i ued by the controller to compute the reactive power et point (Q ref ) of the DFIG. The range of control

68 61 depend on the reactive limit of the DFIG (Q max, Q min ). Thi voltage controller wa incorporated into the exiting DFIG control model available in PowerFactory. The controller deign wa developed uing DIgSILENT Simulation Language (DSL) to compare the dynamic repone of the trategie. The main difference between the control trategie i the reactive power limitation (Q max, Q min ) placed on the controller for a given real output. Thi variation in the limit will tet the ytem repone between the extended reactive limit of the capability curve over the regulated power factor limitation Dynamic Scenario Setup At each penetration level the total wind generation i imulated at 2, 15, 50, and 100% output in order to conider variou production condition from cut-in to cut-out wind peed. Since wind i not a contant reource thi tudy aim to capture the effect of wind variability on ytem reliability. At 2% park output it i conidered that the wind unit have jut cut-in and the real power output i at a minimum. When employing the capability curve, the reactive limit of the machine are the greatet at thi output a compared the other output level tudied. A wind peed increae the park real output increae and conequently the reactive capability of the DFIG park reduce. In contrat, the FERC regulation allow wind unit to increae their reactive capability a the real output i increaed. Again referring to figure 6, at 100% real output the leading reactive capability of both trategie i approximately equal.

69 62 Dynamic ytem data include tandard IEEE exciter and governor model imported from PSS/E for all ynchronou machine. The motor load are repreented by tandard induction machine model Sytem Performance Validation Given the tatic reult gathered from the VSA analyi, dynamic imulation wa carried out in DIgSILENT PowerFactory to analyze the tranient repone of the tet ytem. Simulation were performed uing RMS value (3 rd order imulation model) that capture the electromechanical tranient [29]. The inherent DFIG model in PowerFactory wa modified to incorporate the deigned voltage controller and derived converter rating from the previou ection. The parameter of the modified DFIG model are given in Appendix B. Four power output cenario are teted in order to ae the impact of the reactive control trategie with a high level of DFIG wind penetration. Wind park output of 2% (cutin wind peed), 15%, 50%, and 100% generation are imulated on the previouly decribed power ytem with a 20% penetration level of ditributed DFIG wind generation. The area around bu 3001 wa identified a a critical fault location. A 3-phae hort circuit wa applied at thi bu for a duration of 0.14 econd to compare the repone between the voltage control trategie. It wa determined in the VSA tudy that a penetration level increae the reactive capability of the DFIG park wa inufficient for ecure ytem operation. Hence, additional witched hunt capacitor bank of 150 MVAr were placed at bue 3005 and 3008 to enure an accurate comparion of both trategie.

70 63 For each control trategy the park were initialized to 0.95 leading power factor. Thi enured that pre-fault condition of all machine are identical in each cenario for both control trategie. Before each imulation i performed the correponding reactive limit are placed on the controller hown in Figure 27. A uch, the (P, Q) coordinate are taken from the reactive power curve in Figure 6 and the regulated power factor limit are computed from (40) for the repective real output. The upper and lower reactive bound, Q max and Q max in per unit, are the limit for the voltage controller. The controller parameter are K=1 and T= The following plot (Figure 28-31) detail the DFIG wind park repone to the diturbance near the fault at park 3005 and further away at park Each figure contain two et of plot to compare the voltage control trategie. The three quantitie of comparion are: voltage at load bu 153, reactive power output, and rotor current magnitude through the PEC. Bu 153 ha been elected for monitoring it voltage performance due to the fact that it contain 42% of the ytem motor load. Thi bu voltage will be ued a a metric of comparion between the voltage control trategie. Other quantitie of interet are the total park reactive injection and rotor current, which are integral in determining whether the wind plant electrical control i diabled. The power electronic converter protection limitation were trictly taken into account with an over current etting of 600A (1.27 p.u.) given the 470A nominal current magnitude found from the converter izing. Once triggered a reitive crowbar hort-circuit the rotor winding for 0.15 econd.

71 Scenario 1: 2% Output (Cut-in peed) Thi cenario tet the ytem repone for minimum wind level when the turbine have jut cut-in (4 m/). The Q limit (p.u.) ued in the controller baed on the capability curve (CC) were (0.72,-0.92) and (0.0, 0.0) uing the retricted power factor (RPF) mode. It can be oberved from the bu 153 voltage in figure 28 that with the RPF control cheme the ytem voltage are unable to recover pot fault. In both trategie the PEC crowbar protection doe not activate thu allow reactive injection through the fault. At fault clearance in the CC cae the DFIG plant are able to dynamically compenate for the reactive burden placed on the ynchronou generator by the induction motor. Thu, utilizing the extended reactive capability in the CC cae tabilize the ytem and prevent collape.

72 65 Figure 28. Comparion of bu 153 voltage (p.u.), reactive power (MVAr) (park: 3008/5 red/black), and rotor current (ka) (park: 3008/5 pink/brown) from 20% penetration at cut-in peed with RPF (left) and with CC (right) control Scenario 2: 15% Output At 15% park output the correponding Q limit for thi tudy are (0.70, -0.90) for the CC cae and (0.08, -0.08) for the RPF cae. The PEC protection again doe not activate at the fault initiation and over loading of the converter doe not occur during the tranient. Oberving the bu voltage plot in figure 29 demontrate that the CC control cae provide enhanced pot fault clearance voltage repone. Thi i noticed in the reduced voltage