A multiscale modeling framework for short-term wind power forecasting

点击翻译按钮获取中文摘要

Wind energy industry greatly relies on accurate wind forecasts. Since high-resolution wind forecasts are generally not available for most locations or, even if available, the required accuracy at wind turbine hub heights is not always achievable due to the stochastic and uncontrollable nature of the wind. A need for seeking accurate wind forecast with high spatial resolution to support the rapid growth of wind power becomes even more pressing. Many approaches and methods have been developed to overcome the unavailability and inaccuracy of wind forecast. Physically-based numerical weather prediction (NWP) models are more promising to either predict wind power directly or to provide necessary meteorological information for other special methods such as artificial neural networks (ANNs). However, NWP models are in most cases not accurate in local wind prediction over terrains with complex topography due to its inability to capture micro-scale phenomena. Waterloo CFD Engineering Consulting Inc. (WATCFD) - the supporting company of this Engage proposal - has developed very advanced computational fluid dynamics (CFD) methods that resolve complex topography and air/surface interactions at very high-spatial resolution (applicable to urban scales). WATCFD is keen to deliver the most accurate and reliable wind power forecasting tool by coupling its microscale CFD code urbanSTREAM with the mesoscale Weather Research and Forecasting (WRF) model, and by using the outputs of the multiscale model to train ANN to forecast wind power for the next 0~72 hours.