摘要
本文提出了以几何形式表示集对的G-SPA模型,该模型以向量之间的夹角、相关系数、欧式距离以及向量的模为指标对径流集合建立对应的指标集合,将指标集合的相似度作为判断径流集合相似度的依据,其优点是不仅考虑了径流大小的相似性,而且考虑了径流变化趋势的相似性。将G-SPA预测模型应用于新疆黄水沟年径流量预测中,并分别与GRNN神经网络模型以及AR(2)模型的预测结果进行了对比。结果表明:G-SPA模型预测的平均相对误差为7.42%,预测结果优于GRNN模型和AR(2)模型。
In the paper,G-SPA model representing set with geometry form is proposed. Corresponding index set is established for runoff collection by the model with angle between vectors,correlation coefficient,Euclidean distance and vector model as indexes. Index set similarity is regarded as basis for judging runoff set similarity. It has advantage that rainfall size similarity is considered on one hand,rainfall change trend similarity is considered on the other. G-SPA model forecast is applied in Xinjiang Huangshuigou River annual runoff forecast. They are respectively compared with GRNN neural network model and AR( 2) model. The results show that average relative error of G-SPA model forecast is 7. 42%.The forecast result is better than that of the GRNN model and AR( 2) model.
出处
《水资源开发与管理》
2016年第1期65-67,72,共4页
Water Resources Development and Management
关键词
G-SPA模型
预测
年径流
黄水沟
G-SPA model
forecast
annual runoff
Huangshuigou River
