摘要
从环流指数和ENSO事件指数出发,考虑到大尺度环流系统和海温影响的长期性,使用滑动平均方法对预报因子做前期处理,计算了预报因子在2、3、6月尺度的滑动平均值。在此基础上采用BP神经网络建立了长江大通站汛期径流预测模型(BP-Pre),重点对ENSO事件影响下的径流极值进行预测。为验证模型可行性,采用均生函数模型和利用传统因子建立的BP模型(BP-Ori)作对照分析并进行精度评价。结果表明,部分因子在滑动平均处理下Spearman秩相关系数得到提高;历史平均上,大通站汛期径流与前年发生的ENSO事件关系密切;事件类别不同,对应径流也相对偏大或偏小;BP-Pre模型在率定期拟合效果不及均生函数模型和BP-Ori模型,但从多方法综合评价分析来看其检验期的预测精度更高,对径流极值预测更为精准。
Proceeding from the circulation and ENSO events index,the article considers the long-term influence of large scale circulation system and SST.Moving average method is used to conduct pre-treatment of forecasting factors and calculate the sliding average of forecasting factors in 2,3and 6month scale.On the basis of BP neural network,BP-Pre model is established to forecast the runoff of Datong station on the Yangtze River in flood season,focusing on extreme runoff prediction under the influence of ENSO events.To verify the feasibility of the model,mean generating function model and BP-Ori model based traditional factors are established for contrast analysis and accuracy evaluation.The results show that the Spearman rank correlation coefficients of partial factors were improved under the moving average processing;the flood season runoff of Datong Station has a close relationship with ENSO events in history.While ENSO event type is different,the corresponding runoff is relatively small or large;BP-Pre model is inferior to mean generating function model;during calibration period it has better prediction accuracy based on the evaluation of several methods,which makes more accurate prediction in extreme runoff during validation period.
出处
《水电能源科学》
北大核心
2016年第5期5-8,24,共5页
Water Resources and Power
关键词
ENSO事件
前期处理
大通站
汛期径流预测
ENSO events
pre-treatment
Datong station
flood season runoff prediction