Schulze F H, Wolf H, Jansen H W, van der Veer P
Department of Stochastic methods, Witteveen + Bos consulting engineers, P.O. Box 233, 7400 AE, Deventer, The Netherlands.
Water Sci Technol. 2005;52(9):21-31.
An Artificial Neural Network (ANN) is nowadays recognized as a very promising tool for relating input data to output data. It is said that the possibilities of artificial neural networks are unlimited. Here we focus on the potential role of neural networks in integrated water management. An Artificial Neural Network (ANN) is a mathematical methodology which describes relations between cause (input data) and effects (output data) irrespective of the process laying behind and without the need for making assumptions considering the nature of the relations. The applications are widespread and vary from optimization of measuring networks, operational water management, prediction of drinking water consumption, on-line steering of wastewater treatment plants and sewage systems, up to more specific applications such as establishing a relationship between the observed erosion of groyne field sediments and the characteristics of passing vessels on the river Rhine. Especially where processes are complex, neural networks can open new possibilities for understanding and modelling these kinds of complex processes. Besides explaining the method of ANN this paper shows different applications. Three examples have been worked out in more detail. An intelligent monitoring system is shown for the on-line prediction of water consumption, ANN are successfully used for sludge cost monitoring and optimizing wastewater treatment and the usage of ANN is shown in optimizing and monitoring water quality measuring networks. An ANN appears to be a multiuse and powerful tool for modelling complex processes.
如今，人工神经网络（ANN）被认为是一种将输入数据与输出数据相关联的非常有前途的工具。据说人工神经网络的可能性是无限的。在这里，我们关注神经网络在水资源综合管理中的潜在作用。人工神经网络（ANN）是一种数学方法，它描述了原因（输入数据）和结果（输出数据）之间的关系，而不考虑其背后的过程，也无需对关系的性质做出假设。其应用广泛，涵盖测量网络优化、运营水资源管理、饮用水消耗预测、污水处理厂和污水系统的在线控制，甚至更具体的应用，如建立莱茵河丁坝场沉积物观测侵蚀与过往船只特征之间的关系。特别是在过程复杂的情况下，神经网络可以为理解和建模这类复杂过程开辟新的可能性。除了解释人工神经网络的方法，本文还展示了不同的应用。详细阐述了三个例子。展示了一个用于在线预测用水量的智能监测系统，人工神经网络成功用于污泥成本监测和优化污水处理，以及人工神经网络在优化和监测水质测量网络中的应用。人工神经网络似乎是一种用于建模复杂过程的多功能且强大的工具。
