Volterra模型在月径流预测中的应用研究

Monthly Runoff Prediction Based on Volterra Model

水文过程的月均径流序列存在着较明显的低维混沌特性,利用Volterra模型可以较好的预测低维混沌序列。引入低维混沌动力系统相空间坐标重构的Volterra自适应预测模型,对多年月均径流序列采用二阶Volterra自适应滤波器进行预测。以大渡河石棉站33年的月径流量为例进行验证,预测相对误差<10%的天数为73.3%,相对误差<20%的天数为90.0%,与人工神经网络预测结果对比表明该方法具有较满意的准确率。

Hydrologic processes are of low-dimensioal chaotic nature, which can be predicted effectively by Volterra Model. The paper introduces the Volterra adaptive filter based on the state space delay-coordinate embedding reconstruction of dynamic system , and predicts average monthly runoff series by second order Volterra adaptive filter, compares the performance with ANN by monthly runoff in 33 years of Shimian Station on Dadu River. The results show that days with relative error less than 10% account to 73.3% and days with relative error less than 20% account to 90.0%, with satisfactory prediction precision.