泄洪雾化预测的人工神经网络方法探讨

为了预测高坝泄洪雾化引起的降雨强度分布，本文提出了一种基于人工神经网络的雾化预报模型。该模型将泄洪流量、入水流速、入水角度以及三维河谷地形坐标等作为输入变量，对相应河谷地形内的雾化降雨强度分布进行预测。研究中采用了径向基函数（RBF）网络建模，并且通过在其激发函数中引入Sign—d函数，构造一种混合RBF网络，以改善模型的稳定性和泛化能力。通过东江水电站雾化原型观测资料检验，证明该网络模型在求解泄洪雾化降雨的空间分布方面是适宜而有效的。

Atomization prediction based on artificial neural networks for flood releasing of high dams

A model based on artificial neural networked for predicting the atomization of flood releasing in high dams using flip bucket as energy dissipater is proposed. The released discharge, impinging velocity of jet, impinging angle of jet and the position of points to be studied in 3-D coordinate system are regarded as the input variables and the atomization precipitation at these points are defined as the output variable. The Radial Basis Function Network is adopted to establish the mathematical model but its activation function is setup by sigmoid function forming a kind of hybrid RBF network to promote the convergence of the model. The validity of the model is verified by the prototype observation data of rainfall distribution obtained from Dongjiang Hydro Project.