Application of OMA to operational wind turbine

The presented study focuses on application of Output Only Modal Analysis to operational wind turbines. Issues like time varying nature of operational wind turbine, significant harmonic components due to rotor rotation and considerable aerodynamic damping make OMA of operational wind turbines a difficult task. The study presents the results of OMA applied to experimental data from ALSTOM Wind 3MW Eco-100 wind turbine. Issues like data handling, using clustering for modal identification and uncertainties analysis are presented and discussed. 2 IOMAC'11 – 4 th International Operational Modal Analysis Conference come this OMA limitation were suggested and validated via simulated experiment (Tcherniak et al. (2010b)). Based on the positive results of the simulated experiments, a field experiment was planned and performed. The overall goals of the project were to verify current FE and aeroelastic models but also to develop an experimental methodology of obtaining modal parameters of operational wind turbines. The subject of the experiment was a prototype of the of ALSTOM Wind Eco 100 turbine which is a newly developed 3 MW turbine. The detailed description of the goals, planning and performing of the experiment can be found in Chauhan et al. (2011). While the last mentioned study presented OMA on the wind turbine in a parked state, this study extends the analysis to the operational states. Resulting modal parameters for so-called RPM-regulated regime are presented as a function of rotor speed (Campbell diagram). Simple regression and uncertainty analysis is performed in order to understand the reliability of OMA results for different wind turbine modes. 2 MEASUREMENT CAMPAIGNS: DATA ACQUISITION AND HANDLING To accomplish the objectives of the project keeping the instrumentation costs as low as possible, it was decided to carry out the measurement campaigns in two stages; a long term campaign and a short term campaign. The main goal of the long term campaign was to experimentally determine the global modes of the entire wind turbine structure: i.e. towerand rotor-related modes. Some of these modes are expected to depend on wind load and rotor speed. According to the project objectives, all operational wind speeds should be covered, thus the probability of strong wind during the campaign period together with some logistic issues defined the campaign duration. The focus of the presented study is the global wind turbine modes, therefore only the data from the long measurement campaign is used. The detailed description of the goals of the short campaign is given in Chauhan et al. (2011). In total, 80 channels were recorded during long measurement campaign, among them 68 channels measuring acceleration; 10 control signals and 2 tachoprobe signals. The detailed description of the wind turbine instrumentation is given in Chauhan et al. (2011). Together with acceleration signals, a number of so called control signal were recorded. Most of these signals represent metadata, i.e. characterize the conditions during the measurements. Examples of such signals are wind speed, wind direction, rotor RPM, etc.; they are not directly used in any computations. In contrast, the yaw angle defines the mutual orientation of the nacelle and the tower; it plays important role in computation of global mode shapes. It is important to note that knowledge of blade dynamics plays important role in modal analysis. Firstly, mode identification is much more straightforward when blades accelerations are known. Secondly, one can be more confident in damping estimations provided by OMA algorithms. The abovementioned points reflect the observability issue and are especially important for rotor-related modes. At the same time, previously performed simulated experiments (Tcherniak et al. (2009)) showed that the rotor-related modes can be extracted even without measuring blade acceleration. It should also be noticed that measurement on the blades is a challenging task involving a number of technical issues: for example, obtaining data from the rotating part and synchronizing it with the static part is a difficult though feasible task. Additionally, blade instrumentation is a job which has to be performed by specially trained personal. Taken all these into consideration, it was decided to skip blade instrumentation and measurements. It is necessary to note that measuring of such a big amount of channels is not necessary for identification of global modes: fewer accelerometers and control signals can be used. Providing the minimal number of accelerometers and selecting their optimal location are among the project results; these issues are discussed in the conclusion. Long term measurement campaign started in July 2010 and ended in October 2010; in total the acquisition system was operational 3.5 months. The current study is based on the data collected during the first 41 days of the campaign. During this period about 4000 datasets each 15