Presentations

Abstract

Accurate wind speed knowledge at the height of the rotor swept area is crucial for wind resource assessments. For offshore applications ERA5 data combined with short-term measurements (most recently by floating lidars) through the "Measure, Correlate, Predict" (MCP) method is commonly used in this context. However, ERA5 poses limitations in capturing site-specific wind speed variability due to its low resolution. To address this, we used floating lidar data to develop random forest models extending near-surface wind speed up to 200 m, focusing on the Dutch part of the North Sea. Our methodology uses the wind parameters measured by the ultra-sonic anemometer and met station installed on the buoy as input and is trained on the wind profiles captured by the floating lidar. This model can predict the wind profile up to the maximum measurement height of the floating lidar. Our results show that a random forest model trained on site-specific wind profiles outperforms corrected ERA5 profiles, with a 39% improvement in RMSE and a 35% improvement in Bias. This suggests, as one possible application, significant improvements for gap filling during a floating lidar campaign, when the lidar fails due to various reasons (low aerosol concentration or system failures) but the near surface measurements are still available. In absence of rotor-height measurements, a model trained within a 200 km region handles vertical extension effectively. Our analysis reveals that the horizontal extension of the model primarily impacts bias, with a significant increase when the training location is 200 km away, reaching up to 0.37 m/s in absolute value (300%), accompanied by a relatively minor drop in accuracy. Our regionally trained random forest model exhibits superior accuracy in capturing wind speed variations and local effects, with an average deviation below 5% from measurements compared to corrected ERA5 with a 20% deviation. The random forest model adeptly captures the high frequency range of the power spectrum where ERA5 shows degradation, which can be attributed to its low resolution. Another notable finding of this study is that ERA5 encounters challenges in modeling winds coming from coastal areas. Our analysis reveals that the random forest model also exhibits the highest Mean Absolute Error in the coastal wind sector but is still 0.5 m/s more accurate than ERA5. In conditions where the wind aloft is decoupled from the surface, as observed during stable stratification, the prediction of wind speed poses increased challenges for the random forest model, primarily due to its reliance on information from the near-surface level. Our study highlights the potential enhancement in wind resource assessment by means of machine learning methods. Future research may explore extending the random forest methodology for higher heights, benefiting new generation of offshore wind turbines, and investigating cluster wakes in the North Sea through a multinational network of floating lidars, contingent on data availability.