|dc.description.abstract||Offshore wind power is an active research topic motivated by today’s world energy crisis. While bottom-fixed turbines are relatively mature and commercially viable (though dependent on state-based subsidies to green energy), floating platforms still mostly remain at the research stage. The cost of offshore wind turbines is driven by many factors, and has to be reduced. The risk induced by the lack of knowledge of the environmental loads and the structural response of the turbine is one of these factors. Model testing has always played a significant role in decreasing this risk for offshore structures. Yet, for the case of offshore wind turbines model testing suffers from inherent modeling difficulties: The aerodynamic and hydrodynamic loads, which both contribute significantly to the response of the structure, are challenging to model simultaneously. It is impossible to maintain the Reynolds number from full to model scale, leading to erroneous aerodynamic loads if the turbine and environment are scaled upon the Froude numbers (which has to be used in hydrodynamic tests). Another issue is linked to the physical modeling of wind, which is hard to generate in a wave basin and inherently uncertain.
The solution explored in this PhD study is the artificial modeling of the aerodynamic loads by a numerical code (called numerical substructure), which takes as input the online-measured motions of the structure and whose outputs are applied to the structure in real time by means of actuators. The loads are computed in full scale in real time from a numerical wind field. This concept belongs to the vast family of hardware-in-the-loop testing and has been here given the name of real-time hybrid model (ReaTHM®) testing. It has been trademarked and selected by MARINTEK (Norway) as the way to model wind loads in the ocean basin model tests of the CSC NOWITECH semi-submersible platform supporting the NREL 5MW turbine.
This thesis aims at using this teamwork as a basis for the suggestion of a design and verification procedure for similar ReaTHM test campaigns. As this represents an innovative testing method in the field of offshore hydrodynamics, the design of the ReaTHM test setup had to be made from scratches and trial-and-error. The thesis relates the experience gained.
The definition of the quantities to be measured and the frequency range to be captured turned out to affect the design to a large extent. It first determined which components of the rotor load vector were to be actuated for an accurate modeling of the coupling effects between hydrodynamic, mooring, structural and aerodynamic/ generator loads. The choice was made to focus only on the rigid-body response of the structure, and to focus on frequencies up to the wave range. A dedicated sensitivity study to limited actuation concluded that, for this platform, all components except the vertical tangential aerodynamic force were to be actuated.
The actuation system was designed as a set of 6 servomotor-torsion spring-pulley-wire assemblies, pulling on the structure in a carefully chosen way. This actuator design enabled the development of an effective force control strategy, based on the stiffness of the transmission and the relative position difference between the motor’s rotor and the structure. A real-time control system was implemented, acquiring and processing measurements that were in turn used in the numerical substructure (including AeroDyn from NREL) and in the force controller driving the actuators. Observers were designed to estimate and filter the line tensions (to be used in the feedback force controller) and nacelle velocities (to be used by the numerical substructure).
The knowledge of the motions of the structure was essential for the performance of the controller. The delay in the position measurement was found to significantly affect the force tracking performance. A dedicated identification method has been developped, and the estimated delay was compensated by polynomial extrapolation. The need for exact modeling of the line kinematics also motivated the development of an emulated physical substructure emulating the basin environement including the modeling of the actuators, implemented in MARINTEK’s SIMA. It communicated in real time with the control system, providing a realistic and flexible simulation environment for design and verification. Among other applications, this emulated physical substructure was used to verify the numerical substructure implementation and settings.
The ReaTHM tests were performed by incremental order of complexity, ending with the Ocean Basin tests in October 2015. A post-analysis of the ReaTHM testing method revealed a generally satisfactory performance, through indicators such as the intrusiveness of the method on pure hydrodynamic tests, force tracking errors and effect of delayed inputs. The fully numerical environment provided by the emulated physical substructure played a central role in the post-analysis. Also, an integrated linear model was developed for uncertainty analysis. The linear model was able to accurately capture both structural dynamics, hydrodynamics, aerodynamics and wind turbine control, as well as to model the entire ReaTHM testing control system, with the possibility of adding extra uncertainty. It enabled a flexible and efficient use of the frequency domain. Such a tool can also be used for model-based control design if developed at an early development stage.
As expected when developing a novel method, some aspects of the design were not optimal. The most critical issues were linked to the mechanical design of the scale model and of the actuators. Process noise arising from structural vibrations of spurious eigenmodes showing an overly low natural frequency and/or too little damping turned out to have a dramatic effect on the performance of the force controller. This was worsened by the action of time delays. Efforts were made to study these undesired processes and illustrate them, leading to a list of design rules and fixes that should be followed/applied to facilitate the success of future ReaTHM testing campaigns.||nb_NO