# State of the Art in Floating WindTurbine Design Tools浮动式风力机设计工具的研究现状

NREL is a national laboratory of the U.S. Department of Energy, Office of Energy Efficiency floating wind turbine; integrated design tools; state of the art; numerical simulation; testing; validation INTRODUCTION The offshore wind industry has experienced significant growth in re- cent years, and continues to expand worldwide. Nearly all of the off- shore wind turbines installed to date are located in North European Seas and are mounted on fixed-bottom support structures in water depths of 35 m or less. There are a limited number of suitable shallow water sites available in offshore locations for countries currently active in offshore wind. Much of the global offshore wind resource is in loca- tions where the water is much deeper than it is at the sites of current installations. The offshore resources also exist where fixed-bottom support structures are not feasible, for instance off the coasts of the United States, China, Japan, Spain, Portugal, and Norway. The possi- bility of mounting wind turbines on floating support structures opens up the potential to use such deepwater resources. The economic poten- tial of floating offshore wind turbines (FOWTs) is demonstrated in Musial et al. (2004). Realization of this potential, however, requires cost-effective floating wind turbine designs that can compete with other energy sources. The design and manufacturing of optimized and cost-effective floating wind turbines requires reliable and sophisticated design tools that can model the dynamics and response of floating wind turbine platforms in a comprehensive and fully integrated manner. Currently, several so- phisticated simulation codes are capable of modelling floating offshore wind turbines. This paper presents an overview of the current status of these codes, together with a description of the various modelling tech- niques employed by the different codes, and an analysis of the strengths and weaknesses of these methods. The testing and validation of these design tools is also reviewed, and conclusions are drawn about the development needs and future verification activities required to ensure that the tools continue to improve the accuracy of their loading and response predictions, thus providing the confidence required for de- tailed floating platform design. PREVIOUS RESEARCH Frequency-domain methods commonly are used in the offshore oil and gas industries to analyze and design floating structures. These methods also have been employed in a number of instances for the preliminary design of floating wind turbines. Bulder et al. (2002) used linear fre- quency-domain hydrodynamic techniques to find response amplitude operators (RAOs) to investigate a tri-floater concept. Lee (2005) used a similar process to analyze a tension-leg platform (TLP) design. Vijfhuizen (2006) used frequency-domain analysis to design a barge for a 5-MW turbine, including a wave energy device. Wayman (2006) performed calculations in the frequency domain to model various TLP and barge designs. Sclavounos et al. (2007) performed a parametric design study of floating wind turbine concepts and mooring systems using a coupled linear dynamic analysis in the frequency domain. There are numerous advantages to design calculations in the frequency domain. For example, the studies discussed above were useful in dem- onstrating the initial technical feasibility of floating wind turbines. They showed that turbines could be designed so that the natural fre- quencies are placed away from the wave-energy spectrum to minimize dynamic response. Frequency-domain calculations, however, also have important limitations. They cannot capture nonlinear dynamic charac- teristics and cannot model transient loading events—both of which are important for wind turbines because the nonlinear dynamics introduced through transient events and control system actions are significant for loads analysis. Matha (2009) performed a standard frequency-domain analysis for a floating wind turbine and showed that some couplings between the platform motion and the flexible tower and blades were not taken into account. This factor could lead to natural frequencies being wrongly predicted and critical system resonances not being identified. This result underscores the importance of performing calculations for floating wind turbines in the time domain. For the purposes of this paper, therefore, frequency-domain design tools are not considered and all the codes presented are based on a time-domain analysis. SUMMARY OF EXISTING DESIGN TOOLS A number of design tools available to the offshore wind industry have the capability to model floating offshore wind turbines in a coupled time-domain dynamic analysis. This section presents the methods em- ployed by those design tools known by the authors, and includes four 2 categories: structural dynamics, aerodynamics, hydrodynamics and mooring lines. The summaries presented here apply to the design tool capabilities available at the time of writing; future development is planned for most codes to expand their capabilities. The computational speeds of the various codes will depend on numer- ous factors. These include the discretization chosen by the user, the code features enabled, and the precise details of the coupling scheme (in the case of coupled codes). Without a full knowledge of these vari- ables a direct comparison of the computational speeds of the presented codes is not possible. However it can be said in general that the compu- tational speeds will be slower for codes with more complex coupling schemes. FAST with AeroDyn and HydroDyn by NREL The FAST code is a publicly available simulation tool for horizontal-axis wind turbines that was developed by the National Renewable Energy Laboratory (NREL), largely by Jonkman (2007). The FAST code was developed for the dynamic analysis of conventional fixed-bottom wind turbines, but has been extended with additional modules and to enable coupled dynamic analysis of floating wind turbines. Structural dynamics. The FAST code uses a combined modal and mul- tibody system dynamics (MBS) representation. The wind turbine blades and tower are modelled using linear modal representation assuming small deflections, with two flapwise bending modes and one edgewise bending mode per blade and two fore-aft and two side-to-side bending modes for the tower. The finite element method (FEM) pre-processor BModes (Bir, 2005) is used to calculate the mode shapes of the blades and tower. The floating platform upon which the tower is cantilevered has full six degree-of-freedom (DOF) rigid-body motion. The drivetrain is modelled using an equivalent linear spring and damper. Aerodynamics. The aerodynamic subroutine package AeroDyn is used to calculate aerodynamic forces in FAST. This model uses quasi-steady blade-element/momentum (BEM) theory or a generalized dynamic inflow model. Both of these models include the effects of axial and tangential induction. The BEM aerodynamic calculations include tip and hub losses according to Prandtl and skewed-wake corrections. Dy- namic stall is considered using the Beddoes-Leishman model. Further details can be found in Laino and Hansen (2002). Hydrodynamics. The hydrodynamic subroutine package HydroDyn is used to calculate applied hydrodynamic forces in FAST. Wave kine- matics is calculated using Airy wave theory with free-surface correc- tions. The hydrodynamic loading includes contributions from linear hydrostatic restoring, nonlinear viscous drag contributions from Mori- son’s equation, added mass and damping contributions from linear wave radiation (including free-surface memory effects), and incident wave excitation from linear diffraction. Full details are given in Jonk- man (2009). The linearized radiation and diffraction problems are solved in the frequency domain for a platform of arbitrary shape using WAMIT (Wave Analysis at Massachusetts Institute of Technology), a three-dimensional (3D) panel-based program for computing wave loads and motions of offshore structures (Lee, 1995). The resulting hydrody- namic coefficients are used in HydroDyn. Mooring lines. The FAST code uses a quasi-static mooring system module to represent the nonlinear mooring-line restoring forces. This module accounts for the apparent weight of the mooring line in fluid, the elastic stretching of the mooring line,and the seabed friction of each line. For a given platform displacement, the module solves for the ten- sions within each mooring line by assuming that each cable is in static equilibrium at that instant, and uses the resulting tensions to solve the dynamic equations of motion for the remainder of the system. Full de- tails of the quasi-static mooring line module are given in Jonkman (2009). Figure 1: Interface between modules in the FAST code for FOWTs (Jonkman, 2007) The FAST with AeroDyn and HydroDyn code has been used in a num- ber of research contexts to model coupled wind turbine and floating platform dynamics. The configuration described above is that used by Jonkman (2009). The various modules of the FAST code, however, have also been coupled with a number of other dynamic analysis pro- grams to model the dynamics and response of floating wind turbines. Two examples of this are presented below. FAST with Charm3D Coupling The FAST with AeroDyn code is coupled with the floater-mooring dynamic analysis program Charm3D by Shim (2008). Charm3D is an FEM program for the dynamic analysis of moored floating offshore structures. It was developed jointly by Texas A comparisons between the mooring-line force-displacement relationship calculated by the quasi- static method and that calculated by another code; and comparisons between time-domain results and frequency-domain results. The meth- ods are described in full in Jonkman (2009). The GH Bladed code has recently undergone development from a pure modal representation of structural dynamics to a MBS representation, as described above. The new code structure is released in Bladed v4.0. Several levels of testing and validation were carried out for the new code structure, including code-to-code comparisons and code-to-measurement campaigns. Full details are given in Witcher et al. (2010). The most extensive code-to-code comparison work in the offshore wind industry has been performed as part of the Offshore Code Comparison Collaboration (OC3) project within the International Energy Agency (IEA) Wind Task 23 (Jonkman et al., December 2010). In this project, a number of participants used different aero-elastic codes to model the coupled dynamic response of the same wind turbine and support struc- ture, with the same environmental conditions. The results were com- pared to verify the accuracy and correctness of the modelling capabili- ties of the participant codes, and to improve the predictions. Offshore Code Comparison Collaboration Phase IV In Phase IV of the OC3 project a floating offshore wind turbine was modelled (Jonk- man et al., April 2010). The turbine model used was the publicly avail- able 5-MW baseline wind turbine developed by NREL, and the floating platform was a modification of the Hywind spar-buoy developed by Statoil of Norway. The turbulent wind fields and irregular wave kine- matics were generated independently and were provided to all partici- pants to ensure tight control of all the inputs. A stepwise verification procedure then was used, and the complexity of both the model and the test cases was increased with each step. Figure 9: Illustration of NREL 5-MW wind turbine on OC3-Hywind spar (Jonkman et al., April 2010) A number of floating design tools were involved in Phase IV of the project, including FAST, ADAMS, Bladed, HAWC2, 3Dfloat, SIMO, SESAM and DeepC. The SESAM and DeepC tools were not included in the discussion above because currently they cannot model the cou- pled dynamics of the turbine with floating platform. A variety of differ- ent load cases were performed. These included a full-system eigenana- lysis; a static equilibrium test; free-decay tests for each of the six rigid- body degrees of freedom of the platform; time series response tests with regular waves and irregular waves modelled with a rigid rotor and no wind; time-series response tests with regular waves and irregular waves modelled with a flexible rotor and steady and turbulent wind; and “ef- fective RAOs” calculated with regular waves at varying frequencies. Not all of the codes were able to contribute results to every test case performed due to various limitations on their modelling capabilities. The test cases provided a number of interesting results, some of which are outlined below. Structural dynamics. The participating codes all employ different methods for modelling structural dynamics, which was illustrated in a number of differences in the results. The rotor-nacelle assembly was modelled rigidly in 3Dfloat and both the rotor-nacelle assembly and tower were modelled rigidly in SIMO, SESAM, and DeepC. This meant that these codes could not model structural deflections in these components. The FAST code predicted a higher natural frequency for the second blade asymmetric flapwise yaw frequency than that pro- vided by the other codes. This is because FAST does not account for a torsional mode in the tower, whereas other codes that include tower flexibility do account for this mode. The ADAMS code predicted less energy from the irregular wave simulations in the power spectra for tower-top shear and rotor torque at the second tower and blade bending natural frequencies than produced by FAST and Bladed. This might be because of an effect typical of ADAMS simulations, in which numeri- cal damping increases with frequency. Aerodynamics. Most of the participating codes use BEM theory for the calculation of aerodynamic loads with the exception of SESAM and DeepC, which did not model aerodynamics for the purposes of this project. The 3Dfloat, SIMO, SESAM, and DeepC codes modelled the rotor as rigid, which meant that the aero-elastic response was not rigor- ously modelled. One example of this was in the calculation of effective RAOs, for which the 3Dfloat code showed lower excitation in yaw, greater excitation in fairlead tensions, and greater excitation at the first tower bending frequency for all parameters. This was thought to be due to differences in aerodynamic damping due to the rigid rotor, although it also could have been related to the modelling of the rigid spar with beam elements of artificially high stiffness. The 3Dfloat code also gave a higher mean thrust in the simulations with regular wind and waves, which corresponded with higher platform surge and pitch displace- ments. Hydrodynamics. The free-decay tests showed a few differences between codes in their prediction of the amount of hydrodynamic damping pre- sent in the various modes. The HAWC2 predicted too much heave and pitch damping and ADAMS predicted too little pitch damping relative t