Experimental Study on New Multi-Column Tension-Leg-Type Floating Wind Turbine

Deep-water regions often have winds favorable for offshore wind turbines, and floating turbines currently show the greatest potential to exploit such winds. This work established proper scaling laws for model tests, which were then implemented in the construction of a model wind turbine with optimally designed blades. The aerodynamic, hydrodynamic, and elastic characteristics of the proposed new multi-column tension-leg-type floating wind turbine (WindStar TLP system) were explored in the wave tank testing of a 1:50 scale model at the State Key Laboratory of Ocean Engineering at Shanghai Jiao Tong University. Tests were conducted under conditions of still water, white noise waves, irregular waves, and combined wind, wave, and current loads. The results established the natural periods of the motion, damping, motion response amplitude operators, and tendon tensions of the WindStar TLP system under different environmental conditions, and thus could serve as a reference for further research. Key words: floating wind turbine, model test, WindStar TLP, dynamic response


Introduction
As the development of wind energy continues, offshore turbines will be built in increasingly deep water. Floating wind turbines (FWTs) on platforms tethered to the seabed by cables are the most promising new method for developing deep-water sites, and have become a hot research topic in recent years. This novel type of structure comprises a complicated coupled aero-hydro-servo-elastic multi-body system that experiences multiple loads during the turbine operation. Wide-ranging research by teams around the world has led to various FWT design concepts that have been tested in wave basins Martin et al., 2014;Robertson et al., 2013). Unlike model tests of conventional offshore structures, tests of FWTs must establish the complicated coupled dynamics of the whole platformand-turbine system, and thus the model must include both a wind turbine and its platform to allow study of the system's overall dynamic performance. Therefore, it is crucial to design an appropriate wind turbine model for the FWT model testing. In 2012, a new multi-column TLP-type FWT concept (WindStar TLP system) was proposed by the State was mounted on the tower model. The model was a conventional three-bladed upwind variable-speed, variable bladepitch-to-feather-controlled turbine (Jonkman et al., 2009).
The support platform has a central column, three equivalent radiating corner columns, pontoons, tendon support structures (TSSs), and a vertical tension mooring system. The corner columns provide the buoyancy and external stability during the wet-tow transportation, installation, and operation of the system and also when the tendons are absent. Three TSSs support the tendons and increase the system's pitch and yaw stiffness. The vertical tension mooring system comprises three identical tendons, of which each has a pair of tethers made of polyester rope to implement secure redundancy. To better transfer the large overturning moment of the wind's force on the rotor and to minimize the stress concentration at the tower base, the conical tube tower structure is designed to be installed directly on the central column (Zhao et al., 2016b). The configuration of the proposed WindStar TLP system is illustrated in Fig. 1.

Scaling laws and parameters in the floating wind turbine basin tests
The FWT model was tested in a wave basin to investigate its hydrodynamic, aerodynamic, and elastic characteristics. Wave tank testing of offshore floating structures generally employs the scale models that aim to be similar to the reference system in terms of the geometry, kinematics, and dynamics without suffering from the scale effects. Given that the stiffness of the turbine tower greatly affects the pitching motion of a TLP-type FWT, maintaining the reference's elastic properties in the model is important.
It is considered that the Froude scaling law best preserves the inertial and gravitational forces in the wave basin tests. However, it leads to the Reynolds number being smaller than that at the full scale, which gives smaller lift coefficients and larger drag coefficients for model turbine blades. Therefore, the resulting rotor thrust of the model is much lower than the Froude-scaled value of a full-sized turbine (the Reynolds number scaling effect) (Martin, 2011;Du et al., 2013Du et al., , 2016. This work modified the blade design to give the model the same wind turbine thrust coefficient as the reference, and so eliminate the Reynolds number scaling effect. In summary, the following scaling relationships between the reference and the model were considered. (1) The model should be geometrically similar to the reference. Its linear dimensions (L: e.g., length, width, height, draft, location of the center of gravity, water depth, and wave height) should all be scaled by the same factor, where the subscript r refers to the reference and m to the model.
(2) The model should be kinematically similar to the reference, thus preserving the rotational character of the reference (e.g., rotor frequency and other resulting excitations) in the model. To ensure this similarity, the wind turbine tip speed ratio (TSR) should be maintained as follows: where V is the incoming wind velocity, ω is the wind turbine rotor rotational velocity, and R is the blade radius.
(3) The model should be dynamically similar to the reference. This includes maintaining the Froude scaling law and the rotor thrust coefficient. The Froude scaling law is defined as follows: where U is the characteristic velocity, including the wave particle velocity, structural motion velocity, wind velocity, etc. To keep the model rotor thrust coefficient the same as the reference, and so ensure the correct rotor aerodynamic load, a modified blade design is preferred.
(4) The turbine tower model should be elastically similar to the reference to ensure the natural frequencies scale consistent with time: where E is the Young's modulus of the tower structure. However, it is difficult to achieve a Froude scale stiffness by scaling the material Young's modulus alone. The natural frequency of a homogenous prismatic Euler-Bernoulli beam is given by (Martin, 2011) where β n is dependent on the beam end boundary condition, L is the beam length, I is the cross-section area moment of the inertia, and m is the beam mass. Application of the above scaling yields yielding Therefore, the combination of the material density, modulus and geometry were tuned together to achieve the desire stiffness of the tower model.
Given these scaling relationships, the proposed scaling ratios are summarized in Table 1.

Model and experimental set-up
Tests were conducted in SKLOE's wave basin (50 m× 40 m), which allowed a geometric scaling factor of 50. Water depth could be adjusted by the movable floor of the basin.
The design water depth was 160 m, and the model water depth was 3.2 m. A 1/50 scale model of the WindStar TLP system was designed and fabricated given the scaling ratios in Table 1 (Fig. 2). Data recorded during the tests included six degrees of freedom (DOF) of the tower top forces, turbine nacelle translational accelerations, turbine rotation velocity, platform deck translational accelerations, six DOF of platform motions, and tendon tensions. All data were recorded with a sampling rate of 40 Hz. The low-pass FFT filter was applied for the noise elimination of the raw experimental data. To eliminate the electromagnetic interference generated in the sensors, the servo motor and the controller were connected to the earth ground with a flat braided grounding strap.

Coordination system definition
The fixed global and moveable body coordinate systems are defined in Fig. 3a. The WindStar TLP is defined with a local coordinate system fixed on itself, and can move relative to the global coordinate system. It is considered as a rigid body with six DOFs: three translational (surge, sway, and heave) and three rotational (roll, pitch, and yaw), as shown in Fig. 3b. Wind and waves were considered as collinear in the test.

Wind turbine model
The NREL 5MW wind turbine was chosen as the reference (Jonkman et al., 2009). To meet the strict requirement of the gravity similitude, the model was fabricated with light materials: the blades were manufactured from carbon fiber, and the rest (nacelle, hub, and tower) were made of aluminium. To eliminate the Reynolds number scaling effect, three redesigned model blades based on the NACA4412 aerofoil were adopted. The model blade is shown in Fig. 4, and its parameters are listed in Table 2.
The tower model was fabricated with seven sections of 2024 aluminium alloy hollow rod. Its section properties were determined by the application of the aforementioned scaling laws. The first fore-aft bending natural frequency and the associated mode shape were calculated and compared with the full scale by using the Nastran software. Fig.  5 shows the calculation results of the first fore-aft bending natural frequency and the associated mode shape for both the tower model and the reference. It is found that the first fore-aft bending natural frequency of the tower model is only 0.98% lower than desired. Fig. 6 shows the fabricated 1/50 scale rotor model. An electrical servo motor with the gearbox was applied to control the rotor speed (to maintain the target TSR), and can produce or consume electrical power automatically. The main properties of the reference turbine and the model are summarized in Table 3.

WindStar TLP and tendon model
The WindStar TLP hull model was manufactured using fibreglass, and tethers were modelled using wire rope and springs. An aluminium flange was preinstalled in the hull model to connect it to the tower model. Table 4 gives the main particulars of the reference and the WindStar TLP model. Fig. 7 shows the layout of the mooring system.

Test cases
The IEC 61400-3 design standard (International Electro-technical Commission, 2009) was chosen for the WindStar TLP system. It requires the design of an offshore wind turbine based on the environmental conditions of the installation site, which was assumed to be 61°20′N latitude and 0°0′E longitude, near the Shetland Islands, north-east of Scotland, U.K. (Zhao et al., 2014(Zhao et al., , 2016a. The assumed water depth was 160 m below the mean sea level. The joint probability distribution for wind and waves is provided from 37992 samples, from approximately 13 years of data (Jonkman, 2007).    The experiment had three main parts: still water tests, white noise wave tests, and combined wind, wave, and current tests. The first part verified the horizontal stiffness of the TLP and obtained the natural periods and damping. White noise wave tests studied the RAOs of the TLP. The final set of tests included the turbine's operating conditions under low sea states and parked conditions under extreme sea states. Load cases with winds faster than the rated wind speed had the blade pitch angle and rotor rpm set manually to prescribed values to provide the design rotor thrust value. The model test categories and load cases are summarised in Table 5, and the corresponding environmental conditions are listed in Table 6.

Modelling of environment
A steady wind field was generated by nine axial fans in a 3×3 stacked square configuration installed on a towing carriage. A honeycomb screen was placed in front of the fans to reduce turbulence (Duan et al., 2016). Constant current, represented by the surface velocity, was generated by banks of submerged jets. The JONSWAP spectrum was applied to generate irregular waves. Fig. 8 shows both the measured data and the theoretical spectrum of the 50-year extreme wave condition. The wave spectrum generated in the basin by the high-precision wave generators closely matches the required specifications.

Still water tests
Horizontal static offset tests and six DOF decay tests   were carried out in still water. By prescribing a series of the surge displacements while measuring the restoring force and heave, static offset results (Fig. 9) and a set-down curve ( Fig. 10) were obtained for the WindStar TLP. As expected, the tendon system showed a linear stiffness trend, whereas the set-down curve is strongly nonlinear. The measurements agree well with the theoretical predictions (Chakrabarti, 2005).
To predict the natural periods and total damping of the WindStar TLP system, free decay tests were conducted in still water given an initial displacement in the surge, sway, heave, roll, pitch, and yaw directions. The measured natural periods and linear damping ratios are listed in Table 7. The natural periods of the WindStar TLP system remain well away from the range of wave excitation periods, thus avoid-ing the resonance and strong motion, especially for the heave and pitch.

White noise wave tests
White noise wave tests were performed over a range of frequencies from 0.2 to 1.5 rad/s. The significant wave height is 0.04 m, corresponding to a reference wave height of 2.0 m; the wave approach angle is 0° (heading). As shown in Fig. 11, the white noise wave spectrum generated in the basin shows a good agreement with the theoretical spectrum.
Figs. 12-14 show the surge, heave, and pitch RAOs of the WindStar TLP system, respectively. For the surge motion, the RAO is large at low frequency, whereas those for the heave and pitch mainly consist of high-frequency components.

Combined wind, wave, and current tests
The combined wind, wave, and current tests considered Fig. 10. Comparison of the calculated and measured platform set-down. Fig. 11. Comparison of the target and measured white noise wave spectra.    both operating (DLC01-03) and parked (DLC04) conditions. To identify the dominant load of the dynamic responses, each design load case (DLC) was separated into four sub-cases: each load applied separately and then all three together. The global response variables of the platform motions, tendon tensions, rotor thrust, and nacelle accelerations were examined for the given DLCs. Analysis of the surge and pitch motions (Figs. 15 and 16) show that they are governed by the rotor thrust load, and have the peak responses under the turbine's rated operating conditions. Under parked conditions, as the wave height and current velocity increase, the surge and pitch motions become dominated by the wave and current loads. Mean responses are mainly induced by low-frequency wind and cur-rent loads, and the variations of the amplitude are largely induced by wave loads. The maximum responses of the surge and pitch motion are 27.8 m and 1.62°, respectively, which occurred under the 50-year extreme load case (DLC04). Note that the variations of the surge and pitch amplitude under the combined load cases are lower than those under the wave load only case, suggesting that the wind and current loads stabilized the motions owing to the additional damping. Fig. 17 presents the analysis of the surge acceleration of the turbine nacelle. Unlike for an onshore fixed-bottom wind turbine, the nacelle surge acceleration of a FWT comprises not only the horizontal acceleration induced by the tower's fore-aft vibration, but also the platform's surge acceleration and pitch-induced horizontal acceleration. It is thus larger. The maximum nacelle surge acceleration under only a wind load occurred under the rated conditions and was about 1.71 m/s 2 . In the wave-only load case, the nacelle surge acceleration increased with the increasing wave height, and reached its largest value of 4.05 m/s 2 under the 50-year extreme condition. Low-frequency current load had little effect on the nacelle surge acceleration due to its steady nature. Under the combined loads, wind and current damping largely reduced the motion of the TLP system, and the largest nacelle surge acceleration was only about 3.3 m/s 2 , obviously smaller than that in the wave load only case. Overall, the wave load appeared to drive the nacelle surge     ZHAO Yong-sheng et al. China Ocean Eng., 2018, Vol. 32, No. 2, P. 123-131 129 acceleration. Figs. 18 and 19 present the analysis of the upwind and downwind tendon tension, respectively. The tension is highly influenced by the surge, heave, and pitch motions, and thus exhibits similar trends to them. Damping by wind and current (DLC04) reduces the magnitude of the variation of the tension relative to the wave-only case. Given the collinear wind and wave action considered in the model test, the upwind tendon shows larger tension variation than the downwind tendon: the maximum tension of 5986 kN, including the pre-tension, occurred under the 50-year extreme combined load case (DLC04). As a result, a minimum safety factor of 2.88 is maintained, and the minimum upwind tendon tension remains positive under all design load cases, thus satisfying the design requirement.
Figs. 20 and 21 compare the response spectra and standard deviations of the upwind tendon tension under cases of only wave load and the combined load (DLC04). The spectra show a complex combination of low-frequency, wavefrequency, and high-frequency components. The low-frequency component is mainly induced by the wind and current loads, whereas the wave-frequency and high-frequency components are primarily induced by the wave load. The pitch resonance response appears in the high-frequency component, indicating that the high-frequency tendon tension is greatly affected by the platform pitching. The most significant differences between the two spectra are in the low-and high-frequency components. Under the combined load, the wind and current govern the low-frequency tension response, whereas the damped pitching motion greatly influences the high-frequency tension response. The wavefrequency component has the largest standard deviation, which indicates that the fluctuations of the tendon tension are mainly induced by the wave load.

Conclusions
A comprehensive wave basin model test was carried out for the WindStar TLP system based on properly selected scaling laws and a redesigned rotor model. Results for the platform motion, nacelle surge acceleration, and tendon tension measured under different environmental conditions led to the following conclusions.
(1) The Reynolds number scaling effect can be eliminated by a modified blade design with the same wind turbine thrust coefficient as the reference.
(2) The wave load drives the nacelle surge acceleration in the WindStar TLP system under extreme conditions.
(3) Wind and current loads added to the wave load stabilize the surge and pitch motions. The pitch motion thus dampened by the extra loads had a high-frequency tension response greatly different from the wave-only case.
(4) Under the combined-load case, the wind and current governed the low-frequency tension response, and the fluc-    tuations of the tendon tension were mainly induced by the wave load.
(5) The upwind tendon was more dangerous than the downwind tendon under collinear wind and wave conditions. A sufficient safety factor was maintained, and the minimum tendon tensions remained positive under all the selected DLCs.
(6) In general, the proposed WindStar TLP system showed relatively small dynamic motion responses to environmental loads owing to its natural frequencies being kept well away from the dominant wave frequencies.