基于支持向量机的多因子风速预测

Multi-factors wind speed forecasting based on support vector machine

针对目前数值模式预报的风速普遍存在小风预报偏大、大风预报偏小的问题,本文基于支持向量机的方法,结合WRF模式预报和自动观测站的实况数据资料,建立多因子SVM风速预测模型,对渔业避风港锚地风速进行预测修正.实验结果表明,新模型预测的风速和实际风速基本一致,相关性达到了99%,很好地表达了风速与WRF模式预报因子之间的非线性关系,验证了该模型能改进WRF模式输出的风速数据.与仅利用历史风速的非数值SVM和LSSVM风速预报的结果进行比较,进一步证实了多因子SVM风速预测的优越性.

The problem of deviation has long been identified in the wind speed forecasting with numerical model when wind speed turns too high or low. To tackle this problem, multi-factors SVM model to predict and update the wind speed of fishery harbor anchorage is established, which combines the data of WRF model forecast and those from automatic observation stations. The experimental results suggest that the forecasting wind speed and the actual one is basically in agreement with each other, and the correlation is found to reach up to 99%. The presented model also acts as a good expression for the nonlinear relationship between wind speed and WRF model forecasting factors, which verifies the model＇s ability for updating the WRF wind speed forecasting. By comparing forecast results obtained in this work with other forecast models such as SVN and LSSVM only using historical wind speed, it proves the advantage of the proposed model.