| Summary: | With the increasing size of wind farms and wind turbines, and their deployment further offshore, addressing power losses and operational challenges becomes critical for minimizing the levelized cost of energy (LCOE). This is particularly important as reduced spacing between wind farms leads to increased wake losses due to their aerodynamic interactions. These cluster wakes have a substantial impact on energy yield, requiring accurate calibration of analytical models to improve power production estimates. Simultaneously, the remote locations of offshore wind farms lead to longer downtime when components fail, necessitating remaining useful life (RUL) models that predict component failure and further reduce the LCOE. SCADA (Supervisory Control and Data Acquisition) measurements, which capture real-time turbine performance metrics such as wind speed, power output, and wind direction, along with condition monitoring systems (CMS), provide important information to address these challenges. Accurate analysis of SCADA measurements from large offshore wind farms yields valuable insights into wake effects within and across wind farms. In addition, SCADA measurements combined with CMS data enable the development of physicsbased RUL models that can function with SCADA data as a reference. This thesis uses SCADA and CMS data to develop, calibrate, and validate models across wind farm, turbine, drivetrain, and component scales. A calibration framework for analytical wake models that incorporates SCADA data in time-series form is developed, addressing the limitations of traditional binned methods. The framework demonstrates scalability by enabling calibration for both individual and multiple wind farms, facilitating analysis of large-scale cluster wakes, and achieving a strong match between model predictions and observed data. The convergence of wake losses across different models is observed after calibration, with varying performance identified through quantitative metrics. Using a calibrated analytical wake model and a ...
|
|---|