Multi-scale analysis of offshore wind farm : fine-scale dynamic performance and coarse-scale cooperative control

Abstract

To contain the continuous temperature rise due to greenhouse gas emissions, seeking alternative renewable energy is imperative. Wind power has emerged as a promising and competitive alternative, among which offshore one occupies an essential growth section. Nevertheless, its mass deployment is still crumbled by price pressure. Thus, pursuing a considerable LCOE reduction becomes the critical target for the current development, focusing on an economical and safe design and operation of the wind farm. This research will conduct a multi-scale analysis of this point, namely, the dynamic structural responses of individual turbines at the fine scale and cooperative control of wind farms at the coarse scale. The former contributes significantly to safe and cost-effective turbine designs, thus reducing turbine manufacturing and replacement costs. The latter can mitigate the wake deficit effectively, thus improving wind farm power production and economic benefits.

The fine-scale research covers critical issues including foundation modeling and wave spectral variability. A high-fidelity FE model of the soil-pile system is developed and compared with the two simplified approaches, involving their sensitivity to the foundation stiffness. The discrepancies in the effect of the wave spectral variability induced by the inhomogeneous wave fields are investigated through different dominant roles of swell and wind-induced waves in the wave spectrum, accompanying the discussion on the rationality of the JONSWAP wave spectrum. Results show that based on the benchmark FE model, the distributed spring model exhibits a more pronounced advantage than the apparent fixity model. Marked disparities exist in their sensitivity to the foundation stiffness change resulting from different factors. The influence of wave spectral variability may exhibit an opposite trend depending on whether the wave spectrum is dominated by the swell or wind-induced wave for strong coupling. Accurate characterization of the foundation modeling and stochastic wave loading can improve the prediction accuracy of the dynamic structural responses of the individual turbine by up to 20%.

The coarse-scale research concentrates on crucial subjects, including yawed wake description, real-time optimal control framework, and intelligent control scheme. An innovative double-layer ML framework combining a novel ANN yawed wake mode and a Bayesian machine learning algorithm is developed to perform the real-time yaw control of the wind farm. Then a new row-based control scheme is further proposed to reduce the optimization parameters reasonably, thus improving the optimization rate. Results show that the proposed framework can realize a more accurate power prediction and remarkable power improvement compared with the analytical wake model, and this superiority is enhanced under relatively low inflow velocity and turbulence intensity. In addition, the row-based control scheme can improve the optimization rate remarkably at the expense of a slight decrease in optimal power production. The divergence of the wind distribution and staggered layout will weaken the superiority of the row-based scheme. The proposed framework combining with the smart control scheme can increase the AEP by nearly 5%.

This research also builds the foundation for future investigations on the long-term performance of offshore wind turbines and multi-objective and joint optimization of the wind farm.