论文部分内容阅读
采用相空间重构理论计算实测月径流的延迟时间、嵌入维数、G-P饱和关联维数和Laypunov指数,证明克鲁伦河月径流时间序列存在混沌现象。混沌时间序列的最小二乘支持向量机预测模型涉及参数较少,计算过程简便,训练速度快。RBF神经网络预测模型具有较快的训练速度和较强的非线性映射能力。同时,二者在建模中都引用了径向基函数,从而更加简化了非线性问题的求解。实例表明:将这两种模型应用在月径流时间序列预测上,其运算速度都很快,但在预测精度上,最小二乘支持向量机预测模型要优于径向基神经网络预测模型。
Using phase space reconstruction theory to calculate the measured monthly runoff delay time, embedding dimension, G-P saturation correlation dimension and Laypunov exponent, it is proved that there exists chaos in the monthly runoff time series of Krulun River. The least square support vector machine prediction model of chaotic time series involves fewer parameters, simple calculation process and fast training speed. RBF neural network prediction model has faster training speed and strong nonlinear mapping ability. At the same time, both of them refer to radial basis functions in modeling, thus simplifying the solution of nonlinear problems. The examples show that the two models are applied to the prediction of monthly runoff time series, and their computational speed is very fast. However, the least square support vector machine prediction model is better than the RBF neural network prediction model in prediction accuracy.