摘要
日径流预测是水资源优化调度的重要组成部分,日径流预测精度的高低直接影响水资源优化配置的程度。针对日径流序列的特性,研究提出一种改进的支持向量机回归模型,并应用于日径流预测。与基本支持向量机和BP神经网络对比分析的实验结果表明,基于改进支持向量机回归预测模型的日径流预测精度明显高于BP网络,尤其是对于变化剧烈的径流序列表现出较基本支持向量机回归模型更优越的预测性能,为日径流预测分析提供了一种可靠、有效的途径和方法。
The accuracy of daily runoff forecast has great influence on the optimal allocation of water resources.A forecasting model for daily runoff based on improved Support Vector Machine Regression was proposed according to the characteristic of the daily runoff series.With a comparison to BP neural network and basic Support Vector Machine Regression,the forecast accuracy is significantly higher than BP network,and the forecast performance is more excellent than basic Support Vector Machine Regression special for the runoff series with rapid change.
出处
《水力发电》
北大核心
2010年第3期12-15,共4页
Water Power
基金
国家973重点基础研究发展计划项目(2007CB714107)
国家科技支撑计划项目(2008BAB29B08)
水利部公益性行业科研专项(200701008)
关键词
支持向量机
BP神经网络
核函数
径流预测
support vector machine
BP neural network
kernel function
runoff forecast
