季节性支持向量机中长期径流预报模型

A Seasonal Support Vector Machine Mid-long Term Runoff Forecast Model

中长期径流预报对水资源利用工作具有重要的指导意义。针对月径流季节性特征提出了季节性支持向量机(SVM)中长期径流预报模型,以月份值嵌入样本的形式考虑了月径流的非平稳季节特性;引入了"参数网格化搜索耦合交叉验证"的参数率定方法,确保了模型参数选取的合理性。实例表明,季节性SVM中长期径流预报模型是有效的,相比标准SVM模型、分月SVM模型、BP模型具有明显的优势。

the mid-long term runoff forecasting plays an important role in water resource utilization. In view of the fact that monthly runoff has seasonal characteristics, a seasonal support vector machine （SVM） mid-long term runoff forecast model is proposed. In the model the month value is embedded in samples to consider the non--stationary seasonal characteristic of the monthly runoff, and the grid search coupled with cross validation method is introduced for the parameter calibration of the model which ensures the rationality of the parameters. The case study shows that the proposed method is effective and has obvious advantages over the standard SVM model the monthly SVM model and BP model.