Abstract
By combined use of a meso-scale meteorological prediction system, a Computational Fluid Dynamics (CFD) model and a statistical model, which are developed by CRIEPI, in addition to a wind turbine wake-model and a prediction model for operational status of wind-turbine, a prediction model for wind power generation of a wind farm have been constructed. Evaluating the uncertainty step-by-step in the model, it is shown that the performance of predictability is improved at each step of the model ; the meteorological prediction system show the possibility of rapid variations, the CFD model separates difference between wind-turbines, and the statistical model improve the bias and error. This analysis, furthermore, makes it clear that the error statistics is affected by the prediction of operational status based on observational data from a wind farm. The model is validated by the prediction simulation for 5 wind farms, where the wind fields have different properties, 3 sites in Tohoku Area, 1 site each in Kanto Area and Kyushu Area. The prediction error is less than 20% as for the next day prediction, where the lead-time is between 18 and 42 hours. The improvement rate defined by comparison with the continuous model, which predicts continuously the same value of the wind power generation as that at the prediction time, is greater than 30%.
