摘要
为了有效地揭示水资源系统复杂的非线性结构及变化规律,对具有混沌特性的水资源时间序列重构相空间,计算出相空间的饱和嵌入维数和最大Lyapunov指数,并以此为指导,提出一种适用于高精度逼近和泛化建模的混沌神经网络的学习算法,运用混沌方法构造训练样本及确定神经网络的网络结构,用神经网络拟合相空间相点演化的非线性关系,建立混沌神经网络预测模型。实例表明,该模型有较高的预报精度。
In order to effectively open out the complicated nonlinear structure and the movement law of the water-resource system, the phase space of the time series in the water-resource system, which has chaos idiosyncrasy, is reconstructed, and the systemic embeding dimension and maximal Lyapunov exponent of the phase space are calculated in this paper. On the basis of this, the learning arithmetic is put forward, which is suitable for high precision approach and abroad modeling. The training data construction and the network structure are determined by chaotic phase space, and the nonlinear relationship of phase points is esteblished with neural networks. The predicting model of chaos neural network is constructed. The example indicates that the model has a higher forecasting precision.
出处
《水利与建筑工程学报》
2005年第4期6-9,共4页
Journal of Water Resources and Architectural Engineering
基金
国家高技术研究发展计划(863计划)资助项目(2002AA2Z4291)
华北水利水电学院青年科研基金(HSQG2004005)
关键词
混沌
神经网络
地表水资源
预测模型
chaos
neural network
surface water resources
predicting model
