The Triple Spar Campaign: Implementation and Test of a Blade Pitch Controller on a Scaled Floating Wind Turbine Model

Transcription

1 Downloded from orit.dtu.dk on: Dec 2, 217 The Triple Spr Cmpign: Implementtion nd Test of Blde Pitch Controller on Scled Floting Wind Turine Model Yu, Wenye; Lemmer, F.; Bredmose, Henrik; Borg, Michel; Pegljr Jurdo, Antonio Mnuel; Mikkelsen, Roert Flemming; Lrsen, T. Stoklund; Fjelstrup, T.; Lomholt, Anders Kjær; Boehm, L.; Schlipf, D.; Armendriz, J. Azcon; Cheng, P.W. Pulished in: Energy Procedi Link to rticle, DOI: 1.116/j.egypro Puliction dte: 217 Document Version Pulisher's PDF, lso known s Version of record Link ck to DTU Orit Cittion (APA): Yu, W., Lemmer, F., Bredmose, H., Borg, M., Pegljr Jurdo, A. M., Mikkelsen, R. F.,... Cheng, P. W. (217). The Triple Spr Cmpign: Implementtion nd Test of Blde Pitch Controller on Scled Floting Wind Turine Model. Energy Procedi, 137, DOI: 1.116/j.egypro Generl rights Copyright nd morl rights for the pulictions mde ccessile in the pulic portl re retined y the uthors nd/or other copyright owners nd it is condition of ccessing pulictions tht users recognise nd ide y the legl requirements ssocited with these rights. Users my downlod nd print one copy of ny puliction from the pulic portl for the purpose of privte study or reserch. You my not further distriute the mteril or use it for ny profit-mking ctivity or commercil gin You my freely distriute the URL identifying the puliction in the pulic portl If you elieve tht this document reches copyright plese contct us providing detils, nd we will remove ccess to the work immeditely nd investigte your clim.

2 Aville online t ScienceDirect Energy Procedi 137 (217) th Deep Se Offshore Wind R&D Conference, EERA DeepWind'217, 18-2 Jnury 217, Trondheim, Norwy The Triple Spr Cmpign: Implementtion nd Test of Blde Pitch Controller on Scled Floting Wind Turine Model W. Yu, *, F. Lemmer, H. Bredmose, M. Borg, A. Pegljr-Jurdo, R. F. Mikkelsen, T. Stoklund Lrsen, T. Fjelstrup, A. K. Lomholt, L. Boehm, D. Schlipf, J. Azcon Armendriz c, P. W. Cheng Stuttgrt Wind Energy, University of Stuttgrt, Allmndring 5B, 7569 Stuttgrt, Germny DTU Wind Energy, Nils Koppels Alle Building 43, DK-28 Kgs. Lyngy, Denmrk c CENER, Ciudd de l Innovcı on, no. 7, Srriguren (Nvrr), Spin Astrct In this project y the University of Stuttgrt, DTU Wind Energy nd CENER, rel-time lde-pitch control system ws implemented on scled model in comined wind-nd-wve tnk. A simplified low-order simultion model including erodynmics, hydrodynmics, mooring dynmics nd structurl dynmics ws used to design the controller. Some effort hs een mde to investigte the influence of different gin scheduling methodologies of the collective lde-pitch controller on the dynmic ehvior of the floting wind turine. The issue relting to the negtive erodynmic dmping is lso investigted in order to find out whether the effects seen in simultion models cn e eqully reproduced y model tests. Additionlly, wind nd wve-induced responses with different gin scheduling methodologies nd the difference to the tests without lde-pitch control re discussed. A solution for the hrdwre implementtion of the rel-time controller hs een introduced. The developed controller is proven to function throughout the test cmpign, which lso proves the reliility of the simplified simultion model for controller design. It hs een shown tht with the low-reynolds rotor it is possile to control the rotor speed t Froude-scled frequencies y ctuting the lde pitch ngle. 217 The Authors. Pulished y Elsevier Ltd. Peer-review under responsiility of SINTEF Energi AS. * Wei Yu. Tel.: ; fx: E-mil ddress: yu@if.uni-stuttgrt.de The Authors. Pulished y Elsevier Ltd. Peer-review under responsiility of SINTEF Energi AS /j.egypro

3 324 W. Yu et l. / Energy Procedi 137 (217) Keywords: Floting wind turine; scled model test; controller design; controller implementtion 1. Introduction The wind nd wve induced response of floting offshore wind turines (FOWT) is rther complex nd reflects the rotor erodynmics, tower elsticity, lde-pitch control, incident wves, floting pltform dynmic properties, nd the mooring dynmics. Theoreticl developments, simultion models nd vlidtion dt re needed to understnd these coupled dynmics. Severl coupled ero-hydro-servo-elstic codes hve lredy een developed for the nlysis nd design of FOWTs. The verifiction of these codes y direct code-to-code comprison through the Offshore Code Comprison Collortion (OC3) nd the Offshore Code Comprison Collortion Continution (OC4) projects hd enchmrking success. A significnt numer of model tests for FOWTs hve lso een performed to vlidte the simultion codes. One open question of these model tests is the scling methodology of the erodynmic forces. As solution discussed in [1], newly designed ldes, which yield the correctly (Froude-) scled thrust force, re mostly used for comined wind-nd-wve model tests of FOWT. The Froude-scled model requires re-designing the ldes for low Reynolds numers, which implies tht the erodynmic torque is not correctly scled nd the control system should e redesigned. Due to the complexity, in the pst tests s for full-scle turines, lde-pitch control hs not een included ut the rotor speed ws kept constnt through servomotor. Recent reserch hs investigted the soclled instility prolem in the pltform-pitch mode ecuse of the negtive erodynmic dmping, which is due to the significntly reduced frequency of the tower-top displcement mode. Severl solutions hve een introduced nd verified with simultions, through which the effect of the controller on the dynmic response of FOWTs hs een indicted. To tke this into ccount, comined wind-nd-wve model test with fixed lde-pitch nd constnt rotor speed is not sufficient. One method to include the erodynmic dmping is the Softwre-in-the-Loop method, which is sed on the use of ducted fn sustituting the wind turine scled rotor [2]. The fn yields relistic force to represent the erodynmic thrust through the fn rottionl speed governed y controller. However, the inertil properties of the rotor re not represented correctly, which chnge the nturl frequencies nd the modes shpes of the tower. The motivtion of this pper is to demonstrte solution to include the erodynmic dmping, introducing the controller design nd implementtion for FOWT model tests. For this, the controller is not scled-down from the full-scle model ut re-designed for the scled model nd the low-reynolds irfoils. The pper descries first the simultion model nd the controller design procedure nd finlly comprison of the numericl nd experimentl responses with discussion. 2. Simultion model The simplified simultion model used for the design of the controllers in the experimentl tests is riefly descried in this section. It ws lso clirted through severl identifiction tests Reduced simultion model Figure 1 presents the 1:6 scled DTU 1MW reference wind turine (RWT) test model [3], which is mounted on generic concrete floting pltform model. The pltform concept Triple Spr is hyrid etween spr nd semisumersile, developed originlly in the project INNWIND.EU (see [4]) nd ws uilt t the University of Stuttgrt. The wind turine model, which represents the DTU 1 MW RWT, is constructed t the Dnish Technicl University (DTU). The rotor ldes hve low-reynolds-numer irfoils to produce the Froude-scled thrust force. An overview of the cmpign will e shown in the compnion pper [5]. The rted wind speed of the scled model is defined t 1.47 m/s nd rotor speed t 71 rpm. A simultion model is needed to design controller specificlly for the scled model. Thus, prior to the tests t DHI/Denmrk, simplified low-order simultion model ws set up with only 3 rigid odies: pltform, tower, ncelle nd totl of 5 degrees of freedom (DOFs): surge, heve, pitch, tower top displcement in downwind direction nd the zimuth of the rotor. The joints nd the DOFs re mrked with red color in the sketch in Figure 1 (). A fixed coordinte system with its origin on the se wter level nd t the initil center of flottion is used to descrie the

4 W. Yu et l. / Energy Procedi 137 (217) pltform s position nd orienttion, which is lso presented in Figure 1 (). BEM theory is used to otin the erodynmic coefficients of the rotor. Frequency dependent liner hydrodynmic coefficients, e.g. first order hydrodynmic rdition (dded mss, rdition dmping) nd diffrction forces re clculted with ANSYS-AQWA nd then scled to the model size ccording to the Froude similrity, which is the common procedure in offshore engineering. The dded mss is lso clirted lter in preceding free-decy test. The mooring dynmics re solved using qusi-sttic nonliner model. The simultion models were ugmented with the discrete-time properties of the selected hrdwre nd the noise properties of the sensors in the experiment in order to relisticlly simulte the model eforehnd. Figure 1: () Test model setup; () Simultion model nd coordinte system Clirtion of simultion model A Froude-scled rotor with low-reynolds irfoils is designed to mtch only the thrust, which mens tht the erodynmic power cnnot e scled with the Froude scling lw. This is due to the difficulty in clculting the lift nd drg coefficients t reltively low Reynolds numers. A high uncertinty is ssocited with erodynmic losses nd the drivetrin mechnicl friction et ceter, nd therefore clirtion with experimentl mesurements is necessry. Thus, the erodynmic power coefficient cccc PPPP is firstly nlyzed with the BEM method nd then clirted through rotor-identifiction test with clmped tower se, the test setup of which is presented in Figure 2. The power coefficient ws mesured t severl opertion points with different comintions of lde-pitch ngle nd tip speed rtio. The mesurements were compred to simultion model with BEM-model, whose result is presented in Figure 3. As erodynmic losses re significntly incresed through the low Reynolds numers it ws s expected tht the simultion model power coefficient differs from the test, depicted through yellow lines in Figure 3. To clirte the erodynmic model with less complexity, totl loss efficiency fctor of 65% ws chosen y minimizing the totl devition etween the results of the simultion nd the test, which cn e seen in Figure 3. As result, this fctor ws dded to the BEM simultion model, which ws used for the controller design.

5 326 W. Yu et l. / Energy Procedi 137 (217) Figure 2: Setup of ottom fixed rotor identifiction test. Figure 3: Clirtion of erodynmic power coefficient. Since potentil flow theory neglects the viscosity of wter nd hence does not consider non-liner viscous dmping, the hydrodynmics were clirted with free decy tests in surge, heve nd pitch directions. As n exmple, pitch decy is shown in Figure 4. The nturl frequency ws estimted nd the dmping rtio ws clculted y fitting the envelope line of the mplitudes. The non-liner viscous dmping ws simplified s liner dmping nd then dded to the simultion model. To verify the clirtion, free decy tests were simulted nd compred to the tests. Figure 4 () shows good mtch for the pitch decy test.

6 W. Yu et l. / Energy Procedi 137 (217) Controller design Figure 4: () Free decy test in time nd frequency domin; () Result of hydrodynmic clirtion. Two different methods of controller design re used for the lde-pitch controller for the model test; one is sed on the NREL 5MW seline controller nd detuned ccording to [6] nd [7], nd the other is sed on coupling design method which is presented in [8] Principle concept of the pitch controller The controller for the scled test model is sed on the NREL 5MW seline controller [9], which is ginscheduled proportionl-integrl (PI) controller. The ojective is to otin the optiml tip speed rtio y regulting the generte torque under rted wind speed nd keep stle rotor speed y lde-pitching ove rted. A servomotor produces the genertor torque. Regrdless of the 1:5 ger rtio, the following discussion only uses the corresponding rotor speed (low speed shft) to void confusion. Figure 5: Blde-pitch control lock digrm. The torque controller elow the rted wind speed is implemented in the sme wy s the seline controller to ensure the sfe trnsition etween under-rted nd ove-rted wind speeds, lthough the focus of the tests is ove-

7 328 W. Yu et l. / Energy Procedi 137 (217) rted conditions. The principle concept of the pitch controller is descried in Figure 5. When only 1-DOF is considered, i.e. the drivetrin rottionl speed, the nonliner eqution of the wind turine drivetrin motion is: IIII DDDD φφφφ = MMMM (VVVV, Ω, θθθθ) MMMM gggg, (1) where IIII DDDD is the inerti of the isolted drivetrin, φφφφ is the ngulr position of the servo motor, MMMM nd MMMM gggg re erodynmic torque nd motor torque, nd VVVV, ΩΩΩΩ nd θθθθ re wind speed, rotor speed nd lde-pitch ngle, respectively. At stedy operting point, the erodynmic torque equls the motor torque, such tht the rotor speed is kept constnt, i.e. IIII DDDD φφφφ =. In the region ove rted wind speed, when the genertor torque is usully kept constnt nd there is disturnce from the wind, the motor speed will chnge due to the chnge of erodynmic torque. For negligile fluctution of the rotor speed nd the incoming wind speed, the erodynmic torque cn e linerized s: Considering tht MMMM, = MMMM gggg,, the eqution of the drivetrin motion is: The lde-pitch controller feeds ck the speed error to the pitch ngle MMMM (VVVV, ΩΩΩΩ, θθθθ) = MMMM, + δδδδmmmm ΔΔΔΔθθθθ. (2) δδδδδδδδ IIII DDDD φφφφ = δδδδmmmm ΔΔΔΔθθθθ. (3) δδδδθθθθ tttt Δθθθθ = KKKK PPPP ΔΩ gggg + KKKK IIII ΔΩ gggg dddddddd = KKKK PPPP φφφφ + KKKK IIII φφφφ, (4) where KKKK PPPP nd KKKK IIII denote the coefficients of the proportionl nd integrl terms respectively. Finlly, the liner closedloop eqution, vlid round the operting point cn e represented s: IIII DDDD φφφφ + δδδδmmmm δδδδδδδδ KKKK PPPPφφφφ + δδδδmmmm δδδδδδδδ KKKK IIIIφφφφ =, (5) which hs the form of n ordinry second-order system, where the nturl frequency ωωωω Ω nd the dmping rtio ζζζζ Ω depends on the controller prmeters ωωωω 2 Ω = δδδδmmmm KKKK δδδδδδδδ IIII 1, IIII DDDD 2ωωωω Ω ζζζζ Ω = δδδδmmmm KKKK δδδδδδδδ PPPP 1. IIII DDDD (6) 3.2. Negtive erodynmic dmping Erlier work [6] hs shown the potentil instility of the pltform-pitch mode of FOWTs with conventionl lnd-sed pitch controller. This results from the erodynmic negtive dmping, when the nturl frequency of the drivetrin closed-loop is higher thn the nturl frequency of the pltform pitch mode. The reson cn e descried with the eqution of pltform-pitch motion ccording to [7] s: (MMMM 55 + AAAA 55 )ββββ PPPP + BBBB 55 ββββ PPPP + CCCC 55 ββββ PPPP = FFFF LLLL TTTT, (7) where ββββ PPPP is the pltform-pitch displcement, ββββ PPPP is the velocity, ββββ PPPP is the ccelertion, MMMM 55 is the totl structurl moment of inerti (pltform, rotor nd tower together re regrded s one rigid-ody), AAAA 55 is the dded inerti (from dded mss), BBBB 55 is the rdition dmping nd linerized viscous dmping, CCCC 55 is the linerized pitch restoring stiffness from hydrosttics nd mooring lines, FFFF is the erodynmic thrust nd LLLL TTTT is the hu height. The product of the thrust force nd the hu height will ct s torque in the pitch mode of the pltform. For smll pitch ngles, the pltform pitch ngle cn e presented with the trnsltionl displcement xxxx TTTT nd the height of the hu LLLL TTTT s ββββ PPPP = xxxx TTTT /LLLL TTTT, the eqution of motion in pitch mode cn consequently e written s:

8 W. Yu et l. / Energy Procedi 137 (217) MMMM 55+AAAA 55 LLLL TTTT 2 xxxx TTTT + BBBB 55 LLLL TTTT 2 xxxx TTTT + CCCC 55 LLLL TTTT 2 xxxx TTTT = FFFF. (8) At stedy stte ove the rted wind speed, disturnce in hu displcement (tower motion leds to the chnge in effective wind speed) cuses the chnge of the erodynmic thrust: FFFF = FFFF, FFFF xxxx TTTT. (9) So the eqution (8) of 1-DOF motion in pltform-pitch mode cn e descried s: MMMM 55+AAAA 55 LLLL 2 xxxx TTTT + ( BBBB 55 TTTT LLLL 2 TTTT which is typicl second-order system with dmping rtio: ζζζζ = + FFFF )xxxx TTTT + CCCC 55 LLLL TTTT 2 xxxx TTTT = FFFF,, (1) (BBBB 55+ FFFF LLLL TTTT 2 ) 2 CCCC 55 (MMMM 55 +AAAA 55 ). (11) For lnd-sed wind turines, the pitch ction is reltively slow when compred to the chnge of the wind speed cused y the tower top motion. The thrust coefficient cccc TTTT cn therefore e regrded s constnt when the wind speed chnges. The increse in wind speed will result in n increse in thrust, which mens it is positively dmped, hence the thrust sensitivity FFFF / VVVV contriutes s positive prt. Tht is why the instility prolem cused y the negtive dmping is not significnt for lnd-sed wind turines. However, the first tower nturl frequency of the FOWT decreses significntly. With lnd-sed lde-pitch controller, the ction of pitching is fster. cccc TTTT is reduced y the increse of the lde pitch nd the decrese of the tip speed rtio, which consequently reduces the thrust, so the thrust sensitivity FFFF / VVVV will reduce the totl dmping of the dynmic system, which cn consequently led to negtive dmping in pltform-pitch mode. As conclusion, if FFFF / VVVV is kept ove resonle minimum limit, this instility prolem cn e solved. Qulittively speking, the controller should rect slower, which cn e chieved y reducing the proportionl gins of the controller Gin scheduling Since the erodynmic torque doesn t follow the Froude scling, scling of the controller prmeters from the full-scle turine is not possile. The controller tuning is re-done for the scled model nd is sed on the nlysis of simultions results. Different gin scheduling methodologies re developed using linerized versions of the simultion models. The control design followed the procedure of de-rting the proportionl gin in order to limit the ndwidth of the controller. With the gol of nlyzing this negtive dmping effect in the experiment, the ndwidth ws vried for different controllers y tuning the proportionl gins. For lnd-sed wind turines, the gins re usully scheduled such tht constnt drivetrin closed-loop nturl frequency ωωωω Ω nd dmping rtio ζζζζ Ω re mintined, see Eqution (6). For offshore wind turines, ecuse of the negtive erodynmic dmping, which ws introduced in Section 3.2, [6] hs suggested tht this nturl frequency ωωωω ΩΩΩΩ should e lower thn the pltform-pitch nturl frequency to ensure the pltform remins positively dmped. Hence, severl controllers with different gins re investigted (i.e. with different closed-loop nturl frequencies). As is presented in Figure 6 (), the drivetrin closed-loop nturl frequency of the controllers C1, C2, C3, re.7,.9 nd 1.1 times the pltform-pitch nturl frequency, respectively. The dmping rtio is kept constnt t ζζζζ Ω =.7. Since the pitch sensitivity MMMM / θθθθ chnges over stedy stte, the gins lso chnge over the stedy stte to ensure constnt dynmic ehvior of the closed-loop s Figure 6 () shows, where the corresponding lde-pitch ngles represent the stedy sttes.

9 33 W. Yu et l. / Energy Procedi 137 (217) Figure 6: Overview of test controllers with different gins; () Proportionl gins over lde-pitch ngles in stedy sttes. Another methodology, the coupled gin-scheduling methodology is discussed in [8]. Insted of considering the isolted drivetrin closed-loop, the motion of the whole FOWT is tken into ccount. The FOWT system is regrded s rigid multiody system with 5 DOFs nd set up in simplified model with liner wve excittions nd qusisttic mooring model. The erodynmics including the lde-pitch controller re linerized t different wind speeds, with which it is fesile to get the pole-zero plot nd determine the stility of the coupled system. The principle ide of the coupled gins is to keep the rel prt of the pole in pitch mode constnt for ech wind speed nd find the corresponding proportionl gin K P. For C4, the poles t wind speed 1.6 m s with KKKK PPPP vrying from.1 to.4 nd different time constnts TTTT iiii re plotted in Figure 7 (), in which the influence of KKKK PPPP nd TTTT iiii on the stility of the pitch mode is indicted. First, some of the poles with lrger KKKK PPPP, hve positive rel prt, which mens the pltform-pitch mode is not stle. Second, n incresed TTTT iiii tends to mke this mode more stle, however improvement of the stility is limited. Especilly for TTTT iiii greter thn 2.9 s, the position of the poles doesn t chnge so much to ring remrkle improvement on the pltform-pitch mode stility. Thus, TTTT iiii = 2.9 s is chosen, with which the poles t different wind speeds, i.e. stedy sttes re nlyzed s Figure 7 () shows. It is cler tht the influence of the proportionl gin KKKK PPPP on the pitch mode s stility differs for different wind speeds. At higher wind speeds, the negtive erodynmic dmping hs much less influence. For exmple, when VVVV = 1.9 m/s, if the rel prt of the pole R(λλλλ) is limited within (mrked with red dshedline in Figure 7 ()), the restriction on KKKK PPPP is not strict, since ll the gins stisfy the stility requirement. By interpolting within proportionl gins nd wind speeds, the gins with plced poles cn e clculted, which re plotted in Figure 6 (). To simplify the hrdwre implementtion, KKKK PPPP is polynomil function for smller lde-pitch ut kept constnt when the lde-pitch is greter thn 12 deg. Figure 7: Poles in pitch mode with K_P=.1.4 t wind speed 1.6 m s () nd T_i=2.9 s ().

10 W. Yu et l. / Energy Procedi 137 (217) Hrdwre implementtion of the controller The designed controllers re implemented on n Arduino ord. Detiled implementtion is presented in this section, including the hrdwre setup, the signl cquisition nd the determintion of the loop frequency Hrdwre setup of the control loop Figure 8 shows the finl hrdwre setup of the control loop, including two JVL MAC5 integrted servomotors, microcontroller, n Ethernet-shield, router nd power supply. The two servomotors ct s the torque- nd the ldepitch- ctutor, respectively, which re equipped with the EM4 expnsion module nd le to communicte with the industril stndrdized Modus protocol over TCP/IP. Figure 8: Hrdwre setup of the control loop. The Arduino-Due ord is used s microcontroller ecuse of its open-source prototyping pltform nd ffordle price. LVIEW is used only to log test dt oth from Arduino nd nlog-signl dt cquisition system t DHI. Emedded Coder y Mtl is used to generte the C++ code for ech step of the control loop. This generted onestep code is ssocited with rel-time clock nd executed in Arduino. Signls for the controller re cquired y the seril communictions protocol Modus, which is sed on mster nd slve rchitecture. In the test, the Arduino works s mster which trnsmits requests (lde-pitch ngle or motor torque) to the slve (servomotors) Signl cquisition Initil tests on signl cquisition of the servomotors re performed y ccelerting the servomotor stepwise from rpm to 17 rpm. The mesured motor speed from the speed register shows lrge fluctutions, nd even jumps s Figure 9 () shows, with n pproximte stndrd devition of 65 rpm. Figure 9: () Mesured speed with speed register; () Comprison of mesured speed in frequency domin with different register.

11 332 W. Yu et l. / Energy Procedi 137 (217) The prolem discovered on this issue is tht the mesurement resolution of the speed register is pproximtely + 8 rpm, which ultimtely involves quntistion noise mounting to 8 rpm [1]. Since the position register hs 32 it dt formt (motor speed register only 16 it) nd its unit is encoder counts (encoder smpling frequency is 52.8Hz), it ws decided to get the motor speed y differentiting the motor s current position insted of requiring dt from the speed register, which is clculted s: Ω = φ Loop nnnn φφφφ LLLLLLLLLLLLLLLLnnnn 1 δδδδtttt LLLLLLLLLLLLLLLL, (12) where φφφφ is the rel-time motor position nd δδδδtttt Loop is the exct time intervl etween two loops. The time which is needed for requesting register dt vries y up to 2 ms nd therefore time stmps re used to clculte the time intervl δ Loop. A low-pss filter is implemented to smooth out high frequencies. Regrding the existing system, discretized first order filter is used, which cn e derived y discretizing the Lplce trnsfer function. A common discretiztion method in control pplictions is the (Euler) Bckwrd differentition method, which is: Ω ffff (dddd nnnn ) = TTTT TTTT + δδδδdddd Ω ffff(dddd nnnn 1 ) + δδδδdddd TTTT + δδδδdddd Ω(dddd nnnn) (13) The filter is first order low-pss filter. According to Eqution (13), the present filter output Ω f (t n ) is function of the present filter input Ω(t n ) nd the filter output t the previous discrete time, Ω f (t n 1 ). The filter output is stored in Arduino so tht it is ville for the filter lgorithm t the following execution. Regrding the filtering in the test: At the low-speed shft the 1p-frequency is equl to 1.17 Hz for rted. The finl cut-off frequency is chosen s 7 Hz Loop frequency The communiction etween the Arduino nd the servomotors vi Modus in ech loop turned out to e timeconsuming, which results in mximum 45 Hz loop frequency. At the end of ech loop, wit function is pplied to dely the execution of the next loop, which ensures constnt loop time. However, since the time gp etween the end of one execution of the loop nd the strt of the next execution requires few clock cycles, which mens tht the ction of mesuring how long it tkes ffects how long it tkes, the rel-time clock is not exctly in rel-time, in other words, the loop time is not kept exctly constnt. Figure 1: () Encoder mechnism [11]; () Comprison of mesured speed with efore (left)- nd fter (right)-chnging loop frequency. The signl mesurement frequency ws designed to e the sme s the loop frequency (45Hz), which mens the δδδδtttt LLLLLLLLLLLLLLLL in (12) is 1 45 s. It is over 11 times slower thn the smpling frequency of the encoder (52.8Hz). For this reson lrge mesurement error is discovered (see Figure 1 ()), since the quotient of the encoder smpling frequency nd signl mesurement frequency 52.8Hz/45Hz = is floting point numer. Figure 1 () illustrtes simple rottionl solute encoder. The numer of light-sources nnnn decides how ccurte this encoder s smpling is ccording to the numer of distinct positions of the shft (2 nnnn ). For exmple, the JVL hs 12 light-sources, so there re 2 12 positions nd the resolution should e =.879 deg. These light-sources count 52.8 times per second to trck the unique positions to record the motor s rottionl position. However, Modus request the

12 W. Yu et l. / Energy Procedi 137 (217) dt of position 45 times per second. Tht mens, sometimes Modus requests register dt in the middle of the counting y light-source, so tht it gets the informtion fter the light-sources 11 counts, or gets dt fter 12 counts, there is lwys position difference even when the motor hs constnt speed. Tking this into considertion, the signl cquisition frequency is incresed to 4 Hz, which mkes the rtio n integer, i.e. 52.8Hz/4Hz = 12. The improvement is quite remrkle, which cn e derived from Figure 1 () with the smpling rte 4Hz on the right. 5. Wve tnk tests nd results Tests under wide rnge of se sttes nd wind speeds were crried out throughout the test cmpign. Some results relting to the controller will e presented nd discussed elow. The environmentl conditions for the lod cse (LC) discussed in this section re given in Tle 1. Tle 1: Lod cses for the tests. Model scle Full scle Lod cse H s [m] T p [s] V [m/s] H s [m] T p [s] V [m/s] Discussion of the test results The comprison of the response etween fixed lde-pitch test nd pitch-control test in LC2 is shown in Figure 11 (regulr wve) nd Figure 12 (irregulr wve), where Ω is the rotor speed, θθθθ is the lde-pitch ngle, MMMM gggg is the servomotor torque, ββββ P is the pltform-pitch rottionl displcement nd HHHH ssss represents the significnt wve height. Plots on the left side re in time domin nd those on the right side re in frequency domin. It should e mentioned tht pitch-fixed in the plot mens tht the lde-pitch nd rotor speed re oth preset nd constnt for ech test nd pitch-control implies tht the rotor speed is controlled with the controller C4 designed in Section 3.3. Oviously, the rotor speed is kept elow 11% of the rted rotor speed (71 rpm), which proves tht controller design procedure with simplified low-order simultion model is effective.

13 334 W. Yu et l. / Energy Procedi 137 (217) Figure 11: () Regulr wve response in time domin (LC2); () Regulr wve response in frequency domin (LC2). Figure 12: () Irregulr wve response in time domin (LC2); () Irregulr wve response in frequency domin (LC2). A prolem discovered in Figure 11 () is tht the rotor speed in the pitch-fix test is not fixed s expected, s is evident from the sinusoidl signl of the rotor speed Ω. The reson for this is tht the fixed rotor speed is chieved y servomotor. The servomotor keeps constnt rottionl speed s long s the resultnt torque cting on the shft is zero, which mens the erodynmic torque from one side nd the motor torque from the other side should e in

14 W. Yu et l. / Energy Procedi 137 (217) equilirium. This equilirium is chieved y djusting the motor torque through n internl PID-controller in the servomotor. Since the erodynmic torque chnges with the reltive wind speed, which is influenced y the wind, wve nd the movement of the pltform, the motor torque needs to e regulted ccording to the chnge of erodynmic torque. As consequence, the rotor speed cn only e kept qusi constnt. Regrding the unfixed rotor speed in control group, some conclusions cn e drwn. First, rotor speed control ove the rted wind speed with torque regultion hs much lower motion response in the pltform-pitch mode, which cn e explined y the disppernce of negtive erodynmic dmping, since the lde-pitch is kept constnt. Menwhile, the rotor speed in the pitch-control test hs less fluctution, which shows tht control with lde-pitch is more effective thn with torque. The power spectrl density (PSD) of the response in irregulr wve (LC1) is presented in Figure 13. The performnce of the four controllers, which re introduced in Section 3.3, is compred in frequency domin. The design of C1, C2 nd C3 uses 1-DOF model nd C4 is designed with coupled 5-DOF model. Identified resonnce frequencies corresponding to the modes of the pltform-pitch, wve nd rotor speed 3P re mrked respectively. In the pltform-pitch mode, the wind nd wve induced response with controller C1 is the gretest. Response with controller C2 is smller thn tht with C1, ut still much greter thn C3. This result grees well with the expecttion which is explined in Section 3. When the controller is so fst tht the nturl frequency of the drivetrin closed-loop exceeds the nturl frequency of the pltform-pitch, the instility prolem due to the negtive erodynmic dmping will proly pper. Thus, reducing the ndwidth of the controller (or mke the controller slower) cn mitigte the instility prolem in pltform-pitch mode. However, this method doesn t reduce the response within the rnge of wve frequencies, s evidenced y the similr mplitude in pltform-pitch mode. Another issue tht cn e oserved is tht detuning the gins chnges the system dynmic properties significntly. As cn e seen, the resonnce frequencies in the rotor speed, lde-pitch nd pltform-surge modes re chnged. It should lso e mentioned tht reducing the ndwidth comes t the cost of the control performnce. Aout the controller C4 which is coupled design, hs less wve-induced response in comprison to the other three with detuned gins. In conclusion, detuning sed on the principles of lnd-sed controller is insufficient; more conservtive controller (i.e. smller gins) chnges the coupled nturl frequencies more. Otherwise, it leds to the instility prolem with greter gins wve rotor 3P surge pitch C1 C2 C3 C4 Figure 13: Response in frequency domin with 4 different lde-pitch controllers in irregulr wve (LC1).

15 336 W. Yu et l. / Energy Procedi 137 (217) Vlidtion of simultion model Simultion models of FOWTs re vlidted normlly with fixed lde-pitch tests [1]. The underlying prolem is the question whether such vlidted simultion model cn lso reproduce the motions nd lods with controller t the sme level; in other words, whether simultions coupling the servo dynmic require higher fidelity model. Figure 14 shows the comprison of the vlidtion etween the reduced simultion model nd the test model in regulr wves (LC2). Figure 14: () Vlidtion with lde-pitch fixed test in regulr wve (LC2); () Vlidtion with lde-pitch control in regulr wve (LC2). The test with fixed lde-pitch ngle shows stisfctory greement in terms of pltform motion. A smll disgreement on mplitude is discovered, likely ecuse the motion-dependent hydrodynmic dmping ws simplified s liner dmping, which is determined through pitch decy test, in which the pltform motion frequency is much lower. The offset of the pltform-surge my come from the uncertinty of the nchor position during the test. The reson for the lrge mislignment on the thrust is tht the rotor speed is not exctly constnt ut chnges sinusoidlly, which is missing in the simultion model. Concerning the vlidtion with the controller which is shown in Figure 14 (), it cn e seen tht ll the disgreements re enlrged, i.e. the erodynmic thrust nd the motion of the pltform. The mplitude of the thrust with lde-pitch control in the test is much greter thn the one in simultion. This cn e explined y the sensitivity of the thrust coefficient cccc TTTT. In the pitch-fixed test, the chnge of thrust comes only from the tip speed rtio nd hence the erodynmic ehvior is less influenced since cccc TTTT is less sensitive to the tip speed rtio. However, with the ldepitch control, more uncertinty is included in the simultion model: e.g. for the pitch ctutor system, the time-dely of the control system nd the mesurement error of the rotor speed et ceter. These fctors cn led to greter error y predicting the erodynmics. Thus, vlidtion of simultion models with control system is more complicted. There is lso less reliility if the simultion model is vlidted without control system. Another issue tht will e discussed is whether high fidelity model is necessry to design the control system for FOWTs. It hs een shown previously tht model with only one single drivetrin DOF is not sufficient, ut it remins to e nswered how well the simplified low-order model with 5-DOFs mtches the test model, which is effective for the controller design. An extr comprison of wind nd irregulr wve induced response in oth LC1 nd LC2 is presented in Figure 15. The resonnce frequencies including surge, pitch nd the rnge of wve frequencies gree

16 W. Yu et l. / Energy Procedi 137 (217) well, which mens tht nturl frequencies of different modes of the dynmic system cn e predicted y the simplified model. In the rnge of wve frequencies, the response in the pltform-surge nd -pitch modes gree well, wheres there is slight difference regrding the lde-pitch nd rotor speed modes, which is elievly due to the time dely nd the reltively lower loop frequency of the controller introduced in Section 4.3. In the lower frequency region, in oth of the lod cses, gret difference etween the test nd the simultion exists due to the lck of second order wve excittion. In the higher frequency region, results in LC2 mtches etter. It should e mentioned tht the motion cpture sensors didn t work well in LC2 ecuse some lls for cmer trcking got wet, which ffected the reflection. As consequence, response in surge x P nd pitch β P of LC2 re quite different (for exmple, they don t hve the rotor 3P excittion). The rotor speed 3P excittion isn t replicted in simultion model since the rotor is modelled s n ctutor disk. To summrise, the simultion model cnnot predict the motions well due to the mny simplifictions. However, the controller C4 which is designed through this simplified model showed good performnce in controlling the rotor speed. Figure 15: () Response in irregulr wve (LC1); () Response in irregulr wve (LC2). 6. Conclusion This pper is prtil presenttion of the Triple Spr Cmpign with n ojective of investigting the wind- nd wve-induced response of scled rel-time controlled FOWT. Tests under series of wind-wve comined lod cses re crried out t the wve tnk of DHI in Hørsholm, Denmrk. Some difficulties re discovered from the hrdwre implementtion. Mesurement of rotor speed with higher resolution nd precision is recommended, s well s n ccurte clock to execute control loops. The low resolution of the Arduino with Modus communiction vi TCP/IP seems not sufficient, ecuse nonliner ehvior of the lde pitch nd rotor speed is discovered in regulr wves. The reduced simultion model cn t reproduce the response outside the wve frequency region well, however is efficient nd functionl for controller design. It hs een shown tht the dynmic ehvior of the FOWT with controller is different from the lde-pitch fixed tests, which hs een more commonly used in the pst. Thus, it is recommended tht wind-wve comined model tests should e performed with ctive control.

17 338 W. Yu et l. / Energy Procedi 137 (217) Acknowledgements This reserch ws finncilly supported y University of Stuttgrt, Technicl University of Denmrk nd Ntionl Renewle Energy Centre of Spin. We would lso thnk Myen Wigger for helping with the Modus communiction nd Florin Amnn for uilding the pltform model. References [1] K. Müller, F. Sndner, H. Bredmose, J. M. A. Azcon nd R. Pereir, "Improved Tnk Test Procedures For Scled Floting Offshore Wind Turines," in Interntionl Wind Engineering Conference IWEC, Bremerhven, 214. [2] J. Azcon, F. Bouchotrouch, M. González, J. Grcindí, X. Mundute, F. Kelerlu und T. A. Nygrd, Aerodynmic Thrust Modelling in Wve Tnk Tests of Offshore Floting Wind Turines Using Ducted, Journl of Physics: Conference Series, Volume 524, conference 1, 214. [3] R. F. Mikkelsen, "The DTU 1MW 1:6 model scle wind turine lde," 215. [4] F. Amnn nd F. Lemmer, "Design Solutions for 1MW Floting Offshore Wind Turines," INNWIND.EU, Deliverle D4.37, 216. [5] H. Bredmose, F. Lemmer, M. Borg, A. Pegljr-Jurdo, R. Mikkelsen, T. F. T.Stoklund Lrsen, W. Yu, A. Lomholt, L. Boehm nd J. A. Armendriz, "The Triple Spr cmpign: Model tests of 1MW floting wind turine with wves, wind nd pitch control". [6] T. J. Lrsen nd T. D. Hnson, "A method to void negtive dmped low frequent tower virtions for floting, pitch controlled wind turine," Journl of Physics: Conference Series, vol. 75, p. 1273, 27. [7] J. M. Jonkmn, Influence of Control on the Pitch Dmping of Floting Wind Turine, Reston, V.: Americn Institute of Aeronutics nd Astronutics, 28. [8] F. Sndner, D. Schlipf, D. Mth nd P. W. Cheng, "Integrted optimiztion of floting wind turine systems," in Proceedings of the 33rd Interntionl Conference on Ocen, Offshore nd Arctic Engineering OMAE, Sn Frncisco, USA, 214. [9] J. M. Jonkmn, Dynmics modeling nd lods nlysis of n offshore floting wind turine, ProQuest, 27. [1] A. JVL Industri Elektronik, "Integrted Servo Motors Technicl Mnul," 25. [11] J. Weidemnn, Instrumenttion: How does n encoder work?, 212.