论文部分内容阅读
降水通常是一个地区水资源的主要补给来源,其准确预测对于水资源量的预测等十分重要.为提高降水量的预测精度,以吉林省西部某气象站为例,采用奇异谱分析对月降水量数据进行预处理,提取出多个独立的子序列,再利用支持向量回归机对不同子序列单独建立预测模型,对不同子序列预测模型的输出值求和即可得到该耦合模型的预测值,并利用该耦合模型(SSA-SVR)与小波分析-支持向量回归机耦合模型(WA-SVR)以及在原始降水量数据基础上建立的支持回归机预测模型(SVR)对其月降水量进行步长为1个月、3个月以及6个月的预测.结果表明,三种模型中,SSA-SVR模型的预测值与实测值最为接近,预测精度更高.“,”Rainfall is generally the main source of replenishment of water resources in a region,therefore accurate rainfall forecasting is crucial for the prediction of water resources.A certain weather station in western Jilin Province was taken as an example.To improve prediction accuracy,singular spectrum analysis was used to preprocess the monthly rainfall data,and to decompose the data into several independent subseries.Afterwards,support vector regression was adopted to construct forecasting models based on these subseries.Finally,the final prediction of rainfall was obtained by assembling the outputs of these forecasting models.Based on raw rainfall data,this coupling model (SSA-SVR) along with wavelet analysis-support vector regression model (WA-SVR) and the support vector regression model (SVR) were used to predict monthly rainfall at 1,3 and 6 month-ahead horizons.The results show that among the three forecasting models,the prediction by the SSA-SVR is the closest to the original data,and it has the highest accuracy.