Abstract: During the design phase of offshore wind turbines, statistical data from relevant offshore areas are typically utilized to determine wind speed conditions for analyzing the aerodynamic loads on the turbine. However, this approach often leads to inconsistencies between the design loads and actual loads, thereby increasing uncertainty in structural fatigue assessment and posing safety risks to wind turbine structures. To address these issues, this paper proposes a method for monitoring aerodynamic loads on offshore gravity-based wind turbine towers. This method involves monitoring strain responses on the tower structure and employing dynamic load identification techniques to infer the power spectral density function of aerodynamic load components. Nevertheless, the mathematical models for load identification are frequently afflicted by ill-conditioned problems. To mitigate or completely eliminate the ill-conditioned issues in the mathematical models, this paper introduces a semi-empirical method for sensor placement optimization. By optimizing the locations of structural measurement points, the condition number of the mathematical model is reduced, thereby decreasing the condition number of the frequency response function matrix at resonance frequencies from 1600 to 28.45. Subsequently, the optimized non-ill-conditioned mathematical model is utilized with a matching pursuit inverse pseudo-excitation method to identify aerodynamic loads. The original loads and the identified loads are then inputted separately to analyze fatigue damage at the bottom of the tower. The analysis results indicate that when the signal-to-noise ratio of the monitoring signal exceeds 40 dB, the relative error in damage does not exceed 2%.

Key words: Offshore gravity-based wind turbine, Wind load identification, Ill-posed problem in mathematical models, Illposed elimination, Semi-empirical method, Fatigue analysis