Comparison of Field Measurements and EMT Simulation Results on a Multi-Level STATCOM for Grid Integration of London Array Wind Power Plant

Transcription

1 Downloaded from orbit.dt.dk on: Oct 4, 28 Comparison of Field Measrements and EMT Simlation Reslts on a MltiLevel STATCOM for Grid Integration of London Array Wind Power Plant Glasdam, Jakob; Kocewiak, kasz Hbert; Hjerrild, Jesper; Bak, Clas Leth; Zeni, Lorenzo Pblished in: CIGRE Session 45, 24 Pblication date: 24 Link back to DTU Orbit Citation (APA): Glasdam, J., Kocewiak,. H., Hjerrild, J., Bak, C. L., & Zeni, L. (24). Comparison of Field Measrements and EMT Simlation Reslts on a MltiLevel STATCOM for Grid Integration of London Array Wind Power Plant. In CIGRE Session 45, 24 [B4_26_24] International Concil on Large Electric Systems. General rights Copyright and moral rights for the pblications made accessible in the pblic portal are retained by the athors and/or other copyright owners and it is a condition of accessing pblications that sers recognise and abide by the legal reqirements associated with these rights. Users may download and print one copy of any pblication from the pblic portal for the prpose of private stdy or research. Yo may not frther distribte the material or se it for any profitmaking activity or commercial gain Yo may freely distribte the URL identifying the pblication in the pblic portal If yo believe that this docment breaches copyright please contact s providing details, and we will remove access to the work immediately and investigate yor claim.

2 2, re d Artois, F758 PARIS B4_26_24 CIGRE 24 http : // Comparison of Field Measrements and EMT Simlation Reslts on a Mlti Level STATCOM for Grid Integration of London Array Wind Power Plant Jakob Glasdam,2, Łkasz Hbert Kocewiak, Jesper Hjerrild, Clas Leth Bak 2, Lorenzo Zeni,3 ) DONG Energy Wind Power, 2) Aalborg University, 3) Technical University of Denmark Denmark SUMMARY Simlation reslts are widely sed in the design of electrical systems sch as offshore wind power plants (OWPPs) and for determination of grid compliance. Measrements constitte an important part in the evalation process of the OWPP, inclding passive and active components sch as the static compensator (STATCOM). Qality measrement data allow the system designer to obtain reallife knowledge on the operating characteristics of the electrical component(s) for varios operating scenarios of both the OWPP and the external grid. Frthermore, measrement data constitte an indispensable part of the evalation of the simlation models of the OWPP components. The prpose of this paper is to develop and validate an electromagnetic transient (EMT) generic model of the modlar mltilevel cascaded converter (MMCC) STATCOM based on comparison with field measrements. For this prpose, measrement data have been acqired on a commercial ±5 MVAr stateoftheart (SOA) MMCC STATCOM. The STATCOM is located at the point of common connection (PCC) at the world s largest OWPP, London Array (LAOWPP). According to the athors knowledge, the presented paper will be the first of its kind to compare a detailed model of the STATCOM with actal field measrements for a commercial MMCC STATCOM. Frthermore, the paper offers the reader, being a researcher, transmission system operator, converter designer or a WPP developer etc., the niqe opportnity to gain indepth knowledge of the operating characteristics of the STATCOM for wind power integration, as well as of the validity of applying a generic model of the STATCOM withot knowledge of the actal implemented control system. The proposed model is integrated into an aggregated EMT model of LAOWPP, which will be sed to investigate possible resonance phenomena that will be shown in the paper to affect the harmonic distortion level. The STATCOM distortion level will be shown to be highly affected by the nmber of wind trbine generators (WTGs) in service. It will be shown that the inclsion of band rejection filters (BRFs) in the WTGs control loop lowers the STATCOM distortion level. It is fond that the total interharmonic distortion (TIHD) index calclated according to IEC Standard 647 is sefl for assessing possible and ndesired control interaction between the power electronic devices (PEDs) in OWPPs. The total harmonic distortion (THD) index, on the other hand, is fond to contain very little information on possible PED controller interaction. KEYWORDS Electromagnetic transient (EMT), field measrements, harmonic stability, model evalation, modlar mltilevel cascaded converter (MMCC), offshore wind power plant (OWPP), static compensator (STATCOM), total interharmonic distortion (TIHD) index, voltagesorce converter (VSC). JAKGL@dongenergy.dk

3 Introdction Nowadays, offshore wind power penetration into the electrical grid is rapidly increasing and the crrent trend is to locate the OWPPs frther from shore []. Flexible AC transmission system (FACTS) devices sch as the STATCOM are commonly installed at the PCC in order to meet the reactive power reqirements [2]. Modelling and simlation analysis has become an accepted and integral method in the design of electrical systems sch as the OWPP [3]. The modelling approach mst look for a right compromise between accracy and comptational speed. Frthermore, details regarding the internal behavior of MMCC PEDs are often nknown or only partially known to the transmission system operator (TSO) and the OWPP developer. It is therefore crcial for these parties to assess the reqired modelling details and to gain confidence on the accracy of the applied model. Therefore, field measrements have been collected on an MMCC STATCOM, which will be sed in this paper to assess the MMCC STATCOM modelling reqirements. Previos research on the modelling of the MMCC STATCOM has mainly been focsed on the control system design and performance of the converter sing a switching model (e.g. [4]), an analytical model (e.g. [5]) or a detailed model of the PED (e.g. [6]). None of these pblications consider the accracy of the model representation based on comparison with measrement data. The intention of this paper is to fill the gap, by providing a comparison between a generic, yet detailed STATCOM model and field measrements. Using the validated STATCOM model, the harmonic distortion level in the OWPP will be addressed in the time domain. Section. briefly otlines the pecliarities of possible control interaction between PEDs in an OWPP and possible methods to assess the controller interaction.. Harmonic Distortion in OWPPs OWPPs are ssceptible to socalled harmonic instability, where the extensive sbmarine cabling and possible low available shortcircit power (ASCP) at the PCC may create resonance(s) within the bandwidth of the WTG controller bandwidth [7], [8], [9]. The reslt is nacceptable high harmonic distortion, which may case disconnection of the WTG, nless contermeasres are taken. Disconnection of the WTG(s) is of corse nacceptable, and harmonic stability stdies are now an integral part of OWPP design stdies. Crrently, the stdies are mainly done in the freqency domain [7], [9], where the assessment can be made based on algebraic formlation of the system and conventional indices sch as the Nyqist stability criterion (NSC) can be applied. In [7] it was shown that the harmonic impedance in the OWPP is highly affected by the nmber of WTGs in service (#WTGiS). According to the athors experience it is therefore necessary to perform in the range of thosands of cases to cover all possible operating points of the OWPP and the external grid. This is straightforward in the freqency domain, whereas it is more challenging in the time domain, since the observed reslts do not contain information on the relative stability of the system. Frthermore, timeconsming model initialisation is also needed in the time domain. On the other hand, there are limitations to freqency domain analysis, as linearisation is done on both the nonlinear switching devices as well as the control system. Frthermore, satration effects of e.g. transformers and the integrator part of the PED proportionalintegral (PI) controller are neglected. It is therefore sefl to assess the controller interaction in both domains, i.e. calclate the NSC indices for the e.g. 2 reqired stdy cases and then perform sensitivity analysis on the most critical cases in the time domain (e.g. for 5 cases). The athors are crrently working on developing a procedre for addressing the controller interaction in OWPPs, taking into accont the abovementioned limitations in both the time and freqency domains. This paper will present how the harmonic distortion is affected by the #WTGiS and demonstrate how the time domain approach can be tilised in the assessment..2 Paper Otline The remainder of the paper is organised as follows: Section 2 briefly describes the LAOWPP and the OWPP representation in the simlation tool. The installed STATCOM and the measrement setp are also briefly described. Section 3 presents the proposed model of the STATCOM, which will be evalated in section 4 based on comparison with measrement data. Section 5 presents simlation reslts, in which the #WTGiS is varied and how this affects the STATCOM s performance. Section 6 otlines conclding remarks. Section 7 otlines ftre work. 2

4 2 System Description The LAOWPP is located 2 km off the east coast of the UK at the moth of the Thames, see Figre a. The OWPP is bilt in two phases, where Phase I was partly commissioned at the end of 22. LAOWPP Phase I comprises MW fllscale converter (FSC) WTGs, yielding a total capacity of 63 MW, making it the largest OWPP in the world. LAOWPP I is made p of for sections, and a simplified singleline diagram of one of these sections is shown in Figre b. As indicated in the figre, a combination of a STATCOM and a mechanically switched reactor (MSR) is connected to the tertiary winding of a 4/5/3.9 kv 8/9/9 MVA onshore transformer in each of the for OWPP sections. The STATCOM is eqipped with 22 fllbridge sbmodles (SMs) per phase, where two are redndant, bt still in service dring normal operating mode []. The efficient operating nmber of SMs is therefore N = 2, reslting in a 4level line voltage waveform. Figre b also indicates the locations of the installed voltage and crrent transdcers (VTs and CTs, respectively). Flexible Rogowski coil CTs with a minimm bandwidth of 55 mhz 3 MHz [] were sed together with cstom made capacitive voltage dividers with bandwidth of Hz MHz [2]. The sampling rate was set at 5.2 ks/s/ch for approximately half the dration of the measrement campaign and at 2.4 ks/s/ch for the remainder of the onemonth measrement campaign, reslting in approximately TB of data (2 channels). Figre c shows a more detailed threephase diagram of the STATCOM. The locations of the installed transdcers as well as the symbols of the electrical qantities sed in the paper are shown. 33/ 5 kv Shore Filter 4 kv PCC = ab a b bc ca i = i i i b a ba, ac,2 i ba,2 b A B C i c =temporary installed transdcers =Existing power qality measrement eqipment MSR C filter 3.9 kv i ba, a i ac,2 a2 ac, Phase A L R C C2 i cb, Phase B bc, London a) Geographical location of the LAOWPP. MMCC STATCOM Wye connected ngronded b) Simplified single line diagram of one of the for sections of the LAOWPP. D Phase C cc, Measrement points Voltage: Crrent: b2 c2 i cb,2 i c ac, c) Simplified threephase diagram of the STATCOM and associated filter. Figre Location and single line diagram of the LAOWPP and threephase diagram of the STATCOM. 2. London Array OWPP Representation A simplified representation of the LAOWPP has been implemented in an EMT platform, as will be briefly described. Only one of the for sections is modelled. The WTGs located in the section are aggregated into one eqivalent twolevel FSC WTG. This is considered appropriate, as explicit representation of the individal WTGs significantly increases the modelling order and compromises the simlation speed dramatically. Frthermore, the main focs in this paper is on the performance of the proposed STATCOM model. The athors intend to evalate this simplification in ftre work. The generator side of the converter is represented by a firstorder DC Norton eqivalent with a time constant of T = ms, adapted from [3], see Figre 2b. Figre 2a shows the inner crrent control system, operating in the rotating reference frame (RRF). Band rejection filters (BRFs) have been inclded in the control loop according to the recommendation in [9] to improve the relative harmonic stability of the system. The transfer fnction of the BRF is given in canonical form in () [7], where the damping coefficient is selected as ζ = 2. s is the Laplace operator. ω n,brf is the tned anglar freqency in [rad/s], which has been selected as 3

5 ω n,brf = 2π 8 Hz based on Forier transformation on the simlation reslts. Tning of the BRF is normally done sing freqency domain analysis [7], [9], for reasons described in section.. As will be shown in section 5, the BRF tned at this freqency significantly improves the harmonic content in the STATCOM s otpt crrents and voltages over the fll range of #WTGiS of the LAOWPP. G BRF (s) = s ω n,brf 2 s ω n,brf 2 2ζ s ω n,brf () The 33 kv cable collection grid is aggregated into one nominal PI section with a length of km, which is considered sfficient for the freqency range considered in the crrent work. The approximately 5kmlong 5 kv export cable is modelled sing the FreqencyDependent Phase Model in the EMT simlation tool. No information of actal ASCP has been available, hence the 4 kv grid is represented by a 22 GVA Thevenin eqivalent with X R = 22, given in the asbilt docmentation of the LAOWPP. i c, j θ PLL PCC, j RRF abc RRF abc BRF BRF Crrent controller dq, i dq, ( j= abc,, ) dq, i dq, RRF abc θ PLL j P mech st P dc i dc a) Innerloop crrent controller. b) Simplified WTG model. Figre 2 Aggregated model of the WTG and innerloop controller in the RRF. Command signals i d and i q in a) are provided by two oterloop controllers responsible for controlling dc and U PCC,RMS, respectively. 3 STATCOM MMCC Modelling The MMCC STATCOM consists of a nmber of distribted DC voltages, which are incrementally inserted or bypassed in order to synthesise a highqality sinsoidal voltage waveform. The MMCC technology was extended to the voltage sorce converter highvoltage direct crrent (VSCHVDC) system in [4] and is now considered SOA []. A detailed representation of the MMCC STATCOM has been implemented in an EMT simlation tool, where all the internal dynamics sally relevant for EMT stdies (i.e. generated harmonics de to switching of the converter, individal SM capacitor charging etc. [5]) are taken into accont. Losses de to the commtation process are not inclded, as only the SM terminal and capacitor conditions are of relevance in this work, which is in accordance with the gidelines given in [6] for power system stdies. The relatively high nmber of switching elements in the MMCC STATCOM possesses some challenges in EMT simlation tools, as a high comptational effort is reqired for retrianglarisation of the electrical network sbsystem admittance. Based on the Nested Fast and Simltaneos Soltion [7], an efficient and accrate representation of each of the phase legs of the MMCC VSC HVDC was proposed in [5]. A similar modelling approach has been taken in this work for the STATCOM phase leg, as will be described with reference to Figre 3. dc i dc dc R chop j i c, j Filter bank PCC, j SM n = n = 2 dc, dc,2 leg Main EMT electrical solver i in EQ, leg leg R EQ, leg REQ, leg n= N dc, N Norton Eq. a) Series connection of N=2 fllbridge SMs in the STATCOM phase leg. b) Electrical eqivalent of figre a. c) Aggregated Norton eqivalent of the STATCOM phase leg and interface with the main EMT solver. Figre 3 Derivation of Norton eqivalent representation of the STATCOM phase leg. Firstly, the large nmber of IGBT/diode pairs and distribted capacitances in the phase leg in Figre 3a are represented by their electrical eqivalent, as in Figre 3b. The SM capacitance can be represented by its Norton eqivalent by applying the trapezoidal integration rle according to 4

6 Dommel s formlation [5]. By calclation of the inpt impedance of the n th (n =,2, N = 2) SM and calclating the SM capacitor voltage ( dc,n ) in each SM, it is possible to aggregate the N SMs into a single Norton eqivalent as shown in Figre 3c, where EQ,leg is given in (2): N EQ,leg = dc,n Similarly, R EQ,leg is the smmation of the calclated inpt impedances of the N SMs in Figre 3b. By sing the Norton eqivalent in Figre 3c, the nmber of freqently switched branches in the reslting network admittance matrix is thereby significantly redced, while all branch crrent and node voltage information is retained within the Norton eqivalent and accessible to the main EMT solver. The athors are crrently preparing the description of the derivation of the Norton eqivalent, which will be presented in ftre work. 3. STATCOM Control System One of the main challenges related to MMCCs is the control of the distribted SM capacitor voltages dynamically as well as in steadystate []. A brief description of the implemented controller will be given in the following, with reference to Figre 4. Figre 4a shows the pperlevel controller, which is a conventional cascaded PI controller with the innerloop crrent controller operating in the RRF. The two oter loops are responsible for controlling the PCC reactive power (Q PCC ) (qaxis) and the average SM capacitor voltage of the three phases ( dc,av ) in the daxis, respectively. dc,av is calclated as in (3) and is compared with rated SM voltage ( dc,sm ) in the oter control loop. n= J=3 dc,av = 3N dc,j,n N j= n= The converter phase leg voltage controller in Figre 4b controls the average converter leg voltage ( dc,av,j = N N n= dc,j,n ) in the j th phase leg (j = a, b, c) to follow dc,av from (3). The otpt of the P 4 controller in the j th phase is then mltiplied by a cosine wave which is in phase with c,j obtained from Figre 4a. Smming the otpt of the three P 4 phase controllers is then a zeroseqence reference crrent i. In the existence of an imbalance between the phase leg voltages, e.g. dc,av > dc,av,a, the prodct of a,c i forms a positive active power charging the capacitors in phase leg A in Figre c. The crrent loop forces i to follow its command signal i. Q PCC Q PCC dc, SM dc, AV i c, j θ PLL PCC, j abc abc PI PI 2 RRF RRF i d i q i d d i q q Lp Lp PI 3 PI 3 ( j= abc,, ) SRF= stationary reference frame d α SRF abc c, j q β RRF SRF a) Upperlevel controller with crrent controller operating in the RRF. dc, AV dc, AV, a Σ θ PLL dc, AV, b cos( θ 2π 3 ) Zero seqence crrent controller P 4 dc, AV, c cos( θ 2π 3) θ = tan ( β α) Valve DC voltage controller P 4 cos( θ ) P 4 Σ i ac, i ba, i cb, i 3 i P 5 ' c, j π dc, AV dc, j, n P 6 i k,2 ( k = ba, cb, ac) dc, j, n ci) Individal SM controller. c, j Common reference signal for all SMs in the j th phase dc, j, n jn, Individal reference signal for the n th SM in the j th phase (2) cii) Reference signal for the n th SM. Carrier fors,n and S 2,n jn, S,n S3,n b) Phasevalve voltage and zeroseqence crrent controller. d) POD PSCPWM for the n th SM. Figre 4 Implemented STATCOM controller. Figre c) i) shows the individal capacitor voltage controller for the n th SM in j th phase (j = a, b, c). Figre c) ii) shows the reference signal for the n th SM. [ p.] π 2 S S 3,n 4,n π 2 S 2,n S 4,n Time π (3) [ s] yt ( ) 5

7 A distribted SM voltage controller has been inclded as shown in Figre 4ci. The controller is responsible for maintaining the capacitor voltage of the n th SM in the j th phase ( dc,j,n ) at the reference vale dc,av. The controller forms an active power between dc,j,n and the leg crrent (i k,2 (k = ba, cb, ac), see Figre c) [8]. In [9] it was shown that no copling exists between the zeroseqence crrent controller in Figre 4b and the individal SM controller in Figre 4ci. Figre 4ci shows the addition of the reference signals from the three main controllers in Figre 4a, b and c. for the n th SM (n =,2, N) in the j th phase (j = a, b, c) is then The reslting reference signal j,n fed to a modlator, where it is compared to a trianglarwave carrier (TWC n, n =,2, N) that is individal for the n th SM (Figre 4d). Phase opposition disposition (POD) phaseshifted carrierbased plsewidth modlation (PSCPWM) [2] with a phase shift of θ ps = 2π(n )/N between two adjacent carriers (e.g. TWC n and TWC n ) has been sed. The freqency of the carriers has been selected as f TWCn = Hz. This vale of f TWCn has been chosen as a noninteger plse nmber has a balancing effect on the SM capacitors [2]. It shold be noted that no information on the applied modlation techniqe of the actal STATCOM is pblicly available and was ths not available to the athors. Based on the comparison between measrement and simlation reslts in section 4 (Figre 6b), it is evident that the model with the described modlation techniqe is highly capable of replicating the switching actions of the actal STATCOM. The applied modlation techniqe in the model is therefore considered viable. The modlation techniqe will be more careflly investigated based on postprocessing of the collected onemonth measrement data in ftre work. 4 Model Evalation The proposed model of the MMCC STATCOM described in section 3 will be validated based on comparison with obtained measrement reslts dring steadystate operating mode and dring a transition from Q =.73 p. to Q =.79 p.. The evalation will be based on a qalitative approach comparing measrement and simlation reslts and will be done in the time domain. Figre 5 shows the measred converter terminal voltages and the line crrents dring steady state (see Figre c). Sbscripts m and s denote measrement and simlation, respectively. As evident from the figre, the measrement system is capable of captring the measred waveforms with high resoltion. The figre also clearly shows that the converter terminal voltages with a good approximation appear to be sinsoidal waveforms de to the MMCC technology. Very little harmonic reside is present in the otpt line crrent as also replicated in the simlation. Voltage [p.].8.6 a,c,m.4 a,c,s.2 b,c,m.2 b,c,s.4.6 c,c,m.8 c,c,s i.6 a,m.4 i a,s.2 i b,m.2 i b,s.4 i.6 c,m.8 i c,s a) Measred and simlated leg voltages. b) Measred and simlated line crrents. Figre 5 Comparison between measrement and simlation dring steadystate operating mode. Sbscripts m and s denote measrement and simlation, respectively. Figre 6a shows the phase A converter leg and phase A to B line voltages ( a,c and ab, respectively, see Figre c) dring a shorter time span, in order to more clearly show the high qality of the measrements and the validity of the proposed model. The incremental step increase in the converter voltage is clearly observable in the measrement and well replicated by the simlation reslts ( a,c,m and a,c,s, respectively). The model is to a high extent capable of replicating the measred variation in a,c,m cased by the switching action of the STATCOM. It shold be noted that an imbalance in the line voltages was sstained throghot the measrement period as shown in Figre 6b. The voltage imbalance has been compensated for by adjsting the amplitde of the individal Thevenin phase voltage sorces. This is considered appropriate since no detailed information of the external grid was available. Crrent [p.] 6

8 Figre 7a shows a comparison between measred and simlated phase A line crrent dring the transition in the STATCOM operating point, as previosly defined. A good agreement can be observed between measrements and simlations. Voltage [p.] ab,m ab,s a,c,m a,c,s Days [] a) Phase A voltages. b) STATCOM minte average RMS voltages. Figre 6 a) Comparison of measrement and simlation reslts for phase A. b) minte average RMS line voltage at the STATCOM 3.9 kv bsbar (see Figre b) for a sample of days dring the onemonth measrement campaign. In order to frther validate the proposed STATCOM model, a comparison between measred and simlated phase A line (i a ) and converter leg crrent (i ba,2 ) is shown in Figre 7b for a different STATCOM operating point. Comparing the waveforms with the crrents in Figre 5b, it is evident that the harmonics are more prononced at the low crrent otpt in Figre 7b, which is well replicated in the simlation. Crrent [p.] i a,m i a,s a) Phase A line crrents. b) Phase A line and converter leg crrents. Figre 7 Comparison between simlated and measred phase A line and delta crrents a) dring reactive power transition and b) at Q =.2 p. steadystate operating point. 5 Application of the Proposed STATCOM Model for Harmonic Analysis An assessment on how the #WTGiS affects the qality of the STATCOM otpt waveforms will be made in the following. This is done by performing simlation with the BRF activated (G BRF (s) from ()) and deactivated (G BRF (s) = ). Only the STATCOM s generated waveforms will be analysed and presented, since the STATCOM is the main focs of the paper. Secondly, the prodced waveforms of the WTG reqire more attention, since they are mch more distorted de to the twolevel strctre, shown in Figre 2b. This will be addressed in ftre work. Figre 8a shows the STATCOM phase A leg voltage and crrent with the BRF activated and deactivated (sbscripts BRF and NonBRFrespectively) for #WTGiS = 4 WTGs. The waveforms are significantly distorted when the BRF is disabled, whereas the BRF efficiently trncates the distortion. Figre 8b shows the simlated SM capacitor voltages with the BRF activated and deactivated (top and bottom, respectively) for 5 ms. In both cases the SM voltages are in general well centred arond the reference vale, whereas there is some divergence in the case where the BRF is disabled. The controller interaction between the WTGs and the STATCOM therefore affects the internal dynamics of the STATCOM, implying that simplified STATCOM models sch as those otlined in section cannot be sed to accrately assess the controller interaction between mltiple PEDs. The athors will investigate this in ftre work. According to the athors experience it is not straightforward and relies on a nmber of simplifications and omissions to derive the transfer fnction between the n th SM voltage and the SM s control signal ( dc,j,n and j,n, respectively, see Figre 4). This therefore highlights the need to not only perform the harmonic stability assessment in the freqency domain, bt also to perform a specified nmber of stdy cases in the time domain, as described in section.. Voltage [p.] Crrent [p.] ab bc ca i a,m i a,s i ba,2,m i ba,2,s

9 Calclating the total harmonic (THD) indexes over cycles on the converter leg voltages in Figre 8a sing the method described in the following reslts in THD BRF =.83 p. and THD NonBRF =.77 p. for the two cases. This is very misleading, as observed from the figre. Calclating the total interharmonic distortion (TIHD) indexes yields TIHD BRF =.65 p. and TIHD NonBRF =.24 p., which is in better correlation with Figre 8a. The THD index alone does therefore not indicate possible harmonic controller interaction. The TIHD index will therefore be introdced in the stability assessment. Voltage [p.] Voltage [p.] Voltage & crrent [p.].2.8 a,c,brf.4 a,c,nonbrf. i ba,brf.4 i ba,nonbrf a) Phase A leg voltage and crrent with the BRF activated and deactivated for #WTGiS = 4 WTGs b) SM voltages for phase A leg with BRF enabled and disabled (top and bottom, respectively). c) Calclated THD (top) and TIHD with BRF enabled and disabled (red and ble, respectively). Figre 8 Simlation reslts for the STATCOM and c) calclated THD and TIHD as a fnction of #WTGiS. The THD and TIHD are calclated according to IEC 647 [22], where the sampling interval (SI) is fixed at cycles for fndamental freqency (f = 5 Hz). Ths freqency bins in the discrete spectrm estimated sing discrete Forier transform are with Δf = 5 Hz resoltion. The standard proposes to grop the spectra into harmonic and interharmonic sbgrops (HSGs and IHGs, respectively). The HSG for the n th harmonic (n = 2,3, N = 5) consists of the n f harmonic component and the two adjacent freqency bins (e.g. {95,, 5 Hz} from the HSG of second harmonic (n = 2)). Similarly, the seven freqency bins {6, 65,, 9 Hz} from the first IHG (n = ). The THD is defined as the ratio of the RMS vale of the harmonic sbgrops (HSGs) to the RMS vale of the f component (U ) as in (4a), where U(n f i Δf) is the RMS vale of the n f i Δf freqency bin. The TIHD is calclated similarly, as in (4b). 2 N I= i= N I= [U(n f i Δf)] 2 n=2 [U(n f i Δf)] 2 n=2 i= [p.] (4a) THD = = U U I=8 THD [p.] TIHD [p.].4 BRF NonBRF N n= [U(n f i Δf)] 2 i=2 [p.] TIHD = (4b) U In order to correlate the harmonic stability of the OWPP with #WTGiS, Figre 8c shows the calclated THD BRF and THD NonBRF (top figre) and TIHD BRF and TIHD NonBRF (bottom figre). A resoltion of Δ#WTGiS =5 WTG/bin is sed. Frthermore, the case with #WTGiS = 36 WTGs is also inclded, as this is where the controller interaction becomes critical, as can also be noted from the spike at TIHD NonBRF (36) in the bottom figre in Figre 8c. There is a relatively good correlation between the THD BRF and THD NonBRF for all considered cases p to #WTGiS = 55 WTGs, which again implies that the THD index alone is insfficient in the assessment. A similar correlation between TIHD BRF and TIHD NonBRF is observed for #WTGiS # WTGs in service [] 8

10 WTGs. The TIHD BRF is relatively constant for #WTGiS 45. The TIHD BRF approaches TIHD NonBRF for #WTGiS > 45, indicating that the BRF is not capable to spress the controller interaction for this range of #WTGiS. The harmonic impedance of the OWPP (Z OWPP (ω, #WTGiS)) and the corresponding resonance freqency (ω r (#WTGiS)) are fnctions of #WTGiS, as described in section.. The high nmber of #WTGiS most likely cases ω r ( 5WTG) to be shifted away from ω n,brf, meaning that the BRF is not capable of affectively attenating the resonance. Freqency domain techniqes wold be sefl to assess the freqencydependent characteristics and the change in resonances. The TIHD BRF cold be improved by e.g. increasing the bandwidth of the BRF, which might deteriorate the NSC at other freqencies and is therefore not desired [7]. Another viable option wold be to apply adaptive BRFs in the WTG controller [7]. The inclsion of BRFs in the STATCOM control system cold also be considered. The stdied system has in this work been one section of the LAOWPP, where the average nmber of installed WTGs per section is 75/4 45. Based on Figre 8c, it can therefore be conclded that the OWPP section is stable for all possible #WTGiS, when sing the BRF tned at 8 Hz in the WTG and considering that generic models have been sed to represent both the aggregated WTG and the STATCOM. Frthermore, the highest possible ASCP at the PCC has been sed in this work. It is therefore also necessary to assess the harmonic distortion as a fnction of ASCP in order to assess the overall stability of the system. The possible effect of the remaining three OWPP sections shold also be investigated. The prpose of this section has been to investigate the application of time domain analysis in the harmonic stability assessment and not the actal stability of the system, which will be considered more careflly in ftre work. 6 Conclsion The paper has presented a generic, yet detailed model of the MMCC STATCOM and compared the reslts with highqality measrement data collected on an actal MMCCbased STATCOM for wind integration. It was fond that the generic developed STATCOM model is capable of replicating the measred waveforms with good accracy. A generic model of the STATCOM is therefore sitable in the preliminary design phase of an OWPP, before an agreement with a specific vendor is made. The paper has otlined the necessity to assess the harmonic stability in the design phase of an OWPP and pointed ot advantages of time and freqency domain analysis methods. Using the proposed model of the STATCOM, the harmonic stability assessment has been addressed in the time domain, where the #WTGiS was varied to investigate the controller interaction between the WTGs and the STATCOM. It was fond that the interaction becomes noticeable for #WTGiS 36 WTGs, nless contermeasres are taken. It was fond that the total interharmonic distortion (TIHD) index according to IEC Standard 647 is sefl to assess possible and ndesired control interaction between the PEDs in OWPPs. It was shown that an application of active filtering in the WTGs by means of BRF (i.e. notch filter) in the main control chain can potentially redce harmonic emission at the point of interest (e.g. PCC) and improve overall stability in OWPPs. This can be achieved by redcing the harmonic content generated by the converters as well as changing existing resonances (i.e. improving damping or shifting resonance freqencies). Improving the converters controllers rejection capability called active damping is a certain type of active filtering. The converter may be controlled adaptively or tned to sppress selected harmonic components. Ths there is no need to interfere with the OWPP design. 7 Ftre Work The athors intend in ftre work to more careflly correlate freqency and time domain analysis for harmonic stability stdies and are crrently developing a procedre for assessment of the harmonic stability in AC and VSCHVDCconnected OWPPs. More robst mitigation methods than the BRF sed here will be investigated for. An aggregated representation of the WTGs was sed in the crrent work, since this has a significant impact on simlation speed. The appropriateness of this simplification for harmonic stability stdies will be investigated in the time and freqency domains. The derivation of the detailed MMCC STATCOM model will be more careflly described, and the athors intend to investigate the applicability of more simplified generic RMS and EMT models of the MMCC STATCOM for harmonic stability stdies. A statistical analysis of the onemonth measrement data will be presented in ftre work, where the generated harmonics will be investigated based on e.g. power system freqency variation, OWPP power prodction level, STATCOM operating point etc. 9

11 BIBLIOGRAPY [] J. Glasdam, J. Hjerrild, L. H. Kocewiak and C. L. Bak, "Review on mltilevel voltage sorce converter based HVDC technologies for grid connection of large offshore wind farms" (Power System Technology (POWERCON), 22 IEEE International Conference on, 22, pp. 6.) [2] M. Pereira, D. Retzmann, J. Lottes, M. Wiesinger and G. Wong, "SVC PLUS: An MMC STATCOM for network and grid access applications," (PowerTech, 2 IEEE Trondheim, 2, pp. 5.) [3] J. Glasdam, C. L. Bak and J. Hjerrild, "Transient stdies in large offshore wind farms employing detailed circit breaker representation," (Energies, vol. 5, no. 7, pp , 22.) [4] S. Siriskprasert, A. Q. Hang and J. S. Lai, "Modeling, analysis and control of cascadedmltilevel converterbased STATCOM," (PES General Meeting, 23, IEEE, vol. 4, 23.) [5] J. Kmar, B. Das and P. Agarwal, "Modeling of Level Cascade Mltilevel STATCOM," (International Jornal of Recent Trends in Engineering, vol. 2, no. 5, 29.) [6] T. S. Yeh, H. F. Jh and H. W. Sng, "Modeling and control of threephase mltilevel inverterbased STATCOM," (Power Electronics for Distribted Generation Systems (PEDG), 2 2nd IEEE International Symposim on, 2, pp. 464.) [7] L. Kocewiak, "Harmonics in large offshore wind farms," (PhD Thesis, Department of Energy Technology, Aalborg University, Aalborg, 22.) [8] Ł. H. Kocewiak, J. Hjerrild and C. Leth Bak, "Wind trbine converter control interaction with complex wind farm systems," (IET Renewable Power Generation, vol. 7, no. 4, Nov. 23.) [9] P. Brogan, "The Stability of Mltiple, high power, active front end voltage sorced converters when connected to wind farm collector systems," (Proc. 2 EPEC, 2.) [] M. Pereira, M. Pieschel and R. Stoeber, "Prospects of the new SVC with Modlar Mltilevel Voltage Sorce Converter," (CIGRE Colloqim, October, 2.) [] Powertek, CWT and CWT LF wideband crrent probes for high freqency high crrent measrement, (Accessed online Dec, 23, [2] L. Christensen, et al., "GPS synchronized high voltage measring system," (Nordic Wind Power Conference, nov., 27.) [3] S. K. Chadhary, "Control and Protection of Wind Power Plants with VSCHVDC Connection," (PhD Thesis, Aalborg University, Aalborg, Denmark, 2.) [4] A. Lesnicar and R. Marqardt, "An innovative modlar mltilevel converter topology sitable for a wide power range," (Power Tech Conference Proc., 23 IEEE Bologna, vol. 3, 23, pp. 6) [5] U. N. Gnanarathna, A. M. Gole and R. P. Jayasinghe, "Efficient modeling of modlar mltilevel HVDC converters (MMC) on electromagnetic transient simlation programs," (Power Delivery, IEEE Transactions on, vol. 26, no., pp , 2.) [6] A. Gole, et al., "Gidelines for modeling power electronics in electric power engineering applications," (Power Delivery, IEEE Transactions on, vol. 2, no., pp. 5554, 997.) [7] K. Strnz and E. Carlson, "Nested fast and simltaneos soltion for timedomain simlation of integrative powerelectric and electronic systems," (Power Delivery, IEEE Transactions on, vol. 22, no., pp , 27.) [8] M. Hagiwara and H. Akagi, "Control and experiment of plsewidthmodlated modlar mltilevel converters," (Power Electronics, IEEE Transactions on, vol. 24, no. 7, pp , 29.) [9] M. Hagiwara, R. Maeda and H. Akagi, "NegativeSeqence ReactivePower Control by a PWM STATCOM Based on a Modlar Mltilevel Cascade Converter (MMCCSDBC)," (Indstry Applications, IEEE Transactions on, vol. 48, no. 2, pp , 22.) [2] D. G. Holmes and T. A. Lipo, Plse Width Modlation for Power Converters: Principles and Practice, (IEEE Press, 23.) [2] G. Asplnd, "Method for controlling a voltage sorce converter and a voltage converting apparats," (US Patent 2,,328,977, Dec. 3, 2.) [22] International Electrotechnical Commission, IEC 647, Testing and measrement techniqes General gide on harmonics and interharmonics measrements and instrmentation, for power spply systems and eqipment connected thereto. 22.