应用于水文预报的优化BP神经网络研究

利用广东省滨江流域的水文观测资料，建立了以前期降水量为预报因子、以水位为输出的BP人工神经网络水文预报模型。首先采用了合理的方法进行样本组织，进而利用最优子集回归技术进行输入因子的确定，然后进行了不同隐层节点数、不同转移函数、不同训练算法的组合试验，确定了应用于水文预报中的优化BP神经网络：网络结构为8-9-1；转移函数的组合方式为tansig-线性函数；训练算法为采用evenberg-Marquardt(Lm)算法。为便于精度分析，还采用了最优子集回归模型作了研究。结果表明，优化BP网络模型无论在拟合精度还是在预测精度上都高于最优子集模型。总的来说BP网络是一种精度较高的水文预测模型。

Research on the optimization neural network model for hydrologic forecast

Based on the hydrological observatory data of Bin Jiang basin in Guangdong province,the BP artificial neural network (ANNs) models were established with the rainfall within previous several periods as input variable and the water-level as output variable. Firstly samples were well organized, then the optimized subclass regression was undergone for the selection of input variables, different ANNs were tested with different layer knots numbers, transfer functions and train algorithms and finally the optimized Neural Network Model for Hydrologic forecast was established: the network structure is 8-9-1, the transfer functionsconfiguration is tansig-linear function, and the train algorithm is Lm. The result indicated that BP network model is a hydrologic forecast modelwith high precision.