Probabilistic estimation of fatigue loads on monopile-based offshore wind turbines - Application to sensitivity assessment clustering optimization for support structure cost reduction

摘要

Offshore wind energy faces three important trends: (1) wind farms grow in size, (2) monopiles are installed in deeper water, and (3) cost reduction remains the most important challenge. With wind farm size, the importance of variations in environmental site conditions across the wind farm increases. These site variations, e.g. water depth and soil conditions, can lead to significant differences of loads on support structures. For monopiles in deeper water, design is dominated by wave-induced fatigue loads. Since full fatigue load calculations are computationally demanding, they can typically not be performed for each turbine within large wind farms. Therefore, turbines must be grouped into clusters in early project phases, making time-efficient approaches essential. Optimization of design clustering is necessary to reduce design conservatism and the cost of offshore wind energy.Hence, the goal of this thesis is to investigate load site variations and clustering. Therefore, a probabilistic fatigue load estimation method is developed and verified with aero-elastic simulations in the time domain. Subsequently, the developed method is applied for an exemplary wind farm of 150 turbines in 30-40m water depth to perform(1) sensitivity studies of loads to changes in MSL, soil stiffness, and wave parameters,(2) probabilistic assessments of data, statistical and model uncertainties, and(3) deterministic and probabilistic design clustering.The estimation method is based on frequency domain analysis to calculate wave-induced fatigue loads, a scaling approach for wind loads, combination of wind-wave loads with quadratic superposition, and Monte-Carlo simulations to assess uncertainties. Verification confirms an accuracy of 95% for lifetime equivalent fatigue loads compared to time domain simulations. The computational speed is in the order of 100 times faster than typical time domain tools. Sensitivity studies show a significant influence of water depth and wave period on EFLs. The influence of soil on EFLs is minor for high soil stiffness but can increase significant for soils with low stiffness. Normal distributed input parameters in a probabilistic assessment yield a positively skewed probability distribution of EFLs.Design clustering is optimized based on site-specific fatigue loads using brute-force and discrete optimization algorithms. Results for the exemplary wind farm show a design load reduction of up to 13% compared to standardized design. Probabilistic clustering proved to be only relevant at cluster borders leading to a difference in allocation for 12 out of 150 turbines.Project results show that it is essential to account for load differences in large wind farms due to varying site conditions. This study improves clustering and provides a basis for design optimization and uncertainty analysis in large wind farms. Further work is needed to extend tool verification and formulate design clustering for cost optimization.