支持向量回归的机器学习方法在海浪预测中的应用

基于支持向量回归(SVR)方法,建立了渤海海域近岸海浪有效波高短期预测模型,并设计了多组风浪信息组合输入方案,开展了有效波高预测敏感性试验。研究发现:综合考虑当前风浪信息作为模型的输入,对3 h和6 h有效波高预测具有较高的预报技巧,但随着预测时效的延长其预测准确性迅速降低;若此时引入未来预测风速信息作为模型输入,则可极大提高对12 h和24 h有效波高的预测能力;此外,若输入信息与预测对象之间不存在显著相关,多个信息的输入对有效波高预测效果提高无显著作用。建立的机器学习模型对小样本数据集具有良好的适应能力,能够有效解决海浪预报中的非线性问题,可为近岸海浪有效波高短期预测提供合理的技术参考。

Application of support vector regression in significant wave height forecasting

A short-term SWH forecasting model for the Bohai Sea is established using the support vector regression(SVR)in this paper,and sensitive experiment is conducted with multiple input combination of wind and wave data.It is found that the forecasting model reveals high forecasting skill for 3-hour and 6-hour prediction forced by the current wind and wave information,while the forecasting skill decreases dramatically along with the extension of the forecast leading time.Nevertheless,the forecasting skill for 12-hour and 24-hour prediction increases significantly under forcing of the future wind speed.Moreover,inputs of multiple information won’t increase the forecasting accuracy if there is no significant correlation between the inputs and forecasting variables.In summary,the established SWH forecasting model reveals good adaptability for data of small samples and could effectively solve the nonlinear issues in wave forecasting,which provides technical reference for short-term forecasting of nearshore waves.