用模糊神经网络提高洪峰预报精度的研究

Using Fuzzy Artificial Neural Network to Improve the Accuracy of Flood Peak Forecasting

在大量研究的基础上,提出了基于模糊理论的神经网络改进算法,用来提高对洪峰的预报精度。该方法在网络训练时引入模糊理论来确定网络误差修改的程度。引入的算法增大了大值输出样本和期望输出的误差,使得网络向着提高洪峰拟合精度的方向修改权重。应用表明,改进的模糊BP神经网络能够较好的反映洪水演进机理,提高了神经网络洪水预报模型对洪峰的预报精度,保证了洪峰预报的可靠性。

On the basis of many research, a modified artificial neural network algorithm based on fuzzy theory is proposed to improve the forecasting precision of flood peak. The algorithm introduces the fuzzy theory into the process of network training to decide the degree of error of correcting. The algorithm expands the error between the big value stylebook and anticipant output, making the network modify the weight toward the direction of higher simulating precision to flood peak. The using of the algorithm indicates that the modified fuzzy ANN network can reflect the mechanism of flood, improved the forecasting accuracy of ANN forecasting modal, and ensured the reliability of flood peak forecasting.