Recently, wind power generation is growing rapidly around the globe to reduce carbon emissions from conventional power sources. But because of the intermittency and the irregularity of the wind speed, which easily causes instability in the wind power generation output. Hence, for getting a more stable and accurate wind power generation output target, an enhanced wind speed prediction should be used. In this study, an advanced prediction method based on LSTM and EMD is employed for the wind speed prediction for 60 minutes (1 hour) ahead. The wind power generation output is calculated by considering the yaw misalignment of the wind turbine. Finally, the simulation result indicated a daily fluctuation from the wind farm generation output as well as a high accuracy and followability in the predicted wind speed when compared with the measured wind speed data. Furthermore, the error between the generation target and the generation output is reduced by considering the yaw misalignment.

View full abstract

In this paper, the supercomputer version of RIAM-COMPACT is applied to an offshore wind farm. In calculations using a virtual offshore wind farm model, we found that when the wind enters from an oblique direction (15 degrees), there is no significant decrease in the wind speed entering the downstream wind turbines. Furthermore, in the case of a full-scale offshore wind farm, it was clearly shown that the trends in the measured data could be reproduced by periodically varying the wind direction within a horizontal cross section.

View full abstract

The wakes of a wind turbine with a fixed foundation are mostly understood. However, the wakes of a floating offshore wind turbine become complex due to the motion of the floating foundation, so further knowledge is needed. In this study, a device simulating a floating wind turbine was developed, and the wakes of the model wind turbine were measured when the Strouhal number (St) was 0.08, and the amplitude of the pitch oscillation was 5 degrees. By applying the phase-averaging method to the wake measurement data, the behavior of the wake during pitch motion was clearly expressed. It was found that the distribution of wakes shifted vertically with time and returned to the original distribution over one cycle.

View full abstract

[in Japanese]

View full abstract