Offshore Wind Turbine (OWT) structural design can be optimized to reduce structural costs, thereby lowering wind energy costs. Increasing turbine size and further structural cost reduction require more advanced insights into the dynamic system. One practice of analysing these vibrating structures is called Operational Modal Analysis (OMA). OMA estimates parameters of a structural dynamic model of which damping is considered especially important due
to the direct relation with the magnitude of the response. Although OMA is an established field, damping estimation of operational Offshore Wind Turbines is often difficult due to violation of the stationary white noise excitation condition posed by classical OMA algorithms. Rotor rotation causes a harmonic loading on the non-rotating part of the structure, which is problematic for conventional
methods. Harmonics-mitigating OMA methods were developed in recent years to tackle this issue. Four state-of-the-art OMA algorithms are selected to identify the damping of an operational OWT. Each algorithm is applied to a dataset obtained from a simple simulation test case and a wind turbine simulation before considering field measurements. Subsequently, algorithm
robustness is examined by comparing estimates from multiple measured datasets with corresponding simulations and estimates from different algorithms for multiple wind speeds. It was found that the first mode corresponds well with the model for wind speeds under 15 m/s, whereas the second mode frequency has a large offset from the modelled value, revealing modelling flaws. Moreover, the estimates of the Modified Least-squares Complex Exponential (LSCE), PolyMAX, and Kalman Filter-based Stochastic Subspace Identification (KF-SSI)
algorithms correspond well for both measured and simulated data. They are well-suited for identifying the damping of operational OWTs. Cepstrum editing is not recommended for damping identification as the editing affects the damping value. Enhanced Power Spectral Density Transmissibility (PSDT) is not recommended when few sensor levels are available. Novel enhancements were developed in this thesis. First, the results of the KF-SSI algorithm were improved by concatenating datasets in the identification steps and reducing variance.
Also, numerical issues were addressed by implementing the Square Root Covariance Filter (SRCF) instead of the conventional Kalman filter. Other developments include an algorithm for automated interpretation of stabilization charts and a harmonic localization method based on the rotor velocity.