摘要
实时而准确的日径流量预报在防洪减灾、优化调度等方面起到了巨大作用。将遗传算法(GA)与支持向量回归(SVR)改进模型耦合,同时对SVR三个重要参数(C,σ,ε)进行动态寻优,构建了动态三参数优化GA-SVR日径流非线性预报模型(DGA-SVR)用于黑水河流域日径流预报,通过与BP神经网络和多元线性回归预测结果进行对比分析,DGA-SVR模型预测精度明显优于BP神经网络和多元回归模型。
Accurate runoff forecasting plays an important role in flood control, disaster prevention and optimal operation of reservoir system. This paper combines genetic algorithm with improved support vector regression model and optimizes parameters （C, σ and ε） of SVR dynamically. And then the nonlinear prediction model DGA-SVR is established to forecast the daily runoff in Heishuihe River Basin. Comparative analysis of the BP neural networks and multiple variables linear regression, the results show that the prediction accuracy of DGA-SVR model is obvious better than that of the BP neural networks and multiple variables linear regression model.
出处
《水电能源科学》
北大核心
2012年第8期23-25,共3页
Water Resources and Power
关键词
日径流
非线性
时间序列
遗传算法
支持向量回归
daily runoff
nonlinear
time series
genetic algorithm
support vector regression
