灰色动态神经网络模型及其应用

介绍了灰色动态神经网络的基本原理，分析了传统误差反向传播(BP)算法所存在的缺陷，并提出了解决这些问题的主要途径。根据灰色系统理论，对大坝安全监控模型的因子集进行优化，选择有显著影响的因子建立模型，能明显减少计算工作量，提高工作效率。利用对隐含层结点数的动态调整，有效地解决了隐含层结点数难以确定的矛盾，同时，通过两组样本的相互校验，使训练后的网络预测精度更高。最后，以大坝安全监控模型为实例，对该模型的精度等进行了分析。

Gray dynamic neural network model and its application to dam safety monitoring

The elementary principle of gray dynamic neural network model is introduced. Some limitations, which are often occurred in the establishment of monitoring and controlling model for dam safety with BP model, are analyzed and discussed. The methods for solving these problems are suggested. The genes used to erect the dam safety monitoring and controlling models are optimized with gray system method. Since only two notable genes are necessary for erection of the models, the effeciency of calculation is enhanced. The difficulty of determining the crunode amount of implicit layer is solved by means of adjusting the crunode amount dynamically. The forecasting precision is greatly elevated by the training. An example is given for demonstration.