Search

To forecast wind speed at hub-height is a challenging task for wind energy application, especially in Japan where the terrain feature is very complex and large fluctuations are observed in surface wind field. In this study, an integrated system to predict the hub-height wind speed has been developed by combining data assimilation and Kalman filter with the high resolution Weather Research and Forecasting (WRF) model. Assimilating the nacelle wind data (quality-controlled) and the Kalman filter algorithm effectively improves accuracy of the WRF model forecast by optimizing initial condition and post-processing the model output, respectively. It is found that the WRF model forecasts can be markedly improved after assimilating the nacelle wind data through the Gridpoint Statistical Interpolation analysis system, with the relative improvements of 34.3%, 23.9%, and 8.8% in ME (mean error), RMSE (root mean square error), and IA (index of agreement), respectively. The implementation of the Kalman filter can significantly reduce ME and RMSE while increases the value of IA as well. Further improvement can be achieved if the Kalman filter and nacelle wind data assimilation are implemented simultaneously. It is observed that the role of the Kalman filter is more dominant for the wind band of rated out speeds, while data assimilation is effective in reducing the random errors and becomes more important in rare or extreme weather conditions. Both data assimilation and Kalman filter modules apply the nacelle wind data which is routinely available, so the system can be easily adopted in different wind farm sites for operational use. [ABSTRACT FROM AUTHOR]

The rapid development of wind energy in Japan and the associated high uncertainties and fluctuations in power generation present a big challenge for both wind power generators and electric grids. Accurate and reliable wind power predictions are necessary to optimize the integration of wind power into existing electrical systems. In this study, a hybrid forecasting system of wind power generation was developed by integrating the Kalman filter (KF) with the high resolution Weather Research and Forecasting (WRF) model as well as an empirical formula of wind power output (power curve). The system has been validated with observations including wind speed and power output over a six-month period for 15 turbine sites at a wind farm in Awaji-island, Japan. The results show that the tuned WRF model is able to provide hub-height wind speed prediction for the target area with reliability to some extent. The predicted wind field can be substantially improved by the Kalman filter as a post-processing procedure. The 15-turbine averaged improvements of mean error, root mean square error, and correlation coefficient are 97%, 22%, and 10%, respectively. Meanwhile, the Kalman filter also demonstrates a promising capability of reducing the uncertainties in the power curve model. Systematic validations regarding both wind speed and power output were carried out against the observations for the target wind farm, which show that the hybrid power forecasting system presented in this paper can be an effective and practical tool for short-term predictions of wind speed and power output in Japan area. [ABSTRACT FROM AUTHOR]