丽江市黑龙潭泉群断流的人工神经网络模拟

近年来，云南省丽江市著名景点黑龙潭泉群断流频发，这将严重威胁丽江市旅游业的可持续发展。为了正确认识黑龙潭泉群断流的原因，并掌握其发生的规律，本文在对该泉群的水文地质条件、降水量和断流的关系进行分析的基础上，对该泉群的断流情况开展了人工神经网络模拟研究。本文发现黑龙潭泉群属于非全排型山前断裂溢流岩溶泉；年降水量不足与该泉群的断流具有一定的因果关系；构建了网络拓扑结构为6-13-3的 BP 人工神经网络模型对黑龙潭泉群的不同断流情况进行了模拟，该模型以前期降水量、温度与湿度作为输入向量参数，以1953-2002年的数据作为训练样本，以2003-2012年的数据作为模型检验样本，检验结果与实际情况吻合度约为90%，表明该模型可以较好地模拟黑龙潭泉群的断流情况。

Artificial neural network simulation to zero flow of the Heilongtan spring groups in Lijiang

Zero flow of the Heilongtan spring group that is famous scenery in Lijiang,Yunnan Province fre-quently occurs recently,which severely threatening the sustainable development of Lijiang tourism.In order to know the real reason for zero flow of the Heilongtan spring group and its occurrence regularity,hydrogeo-logical conditions and correlation between precipitation and zero flow of the spring group are analyzed sys-tematically,and a simulation based on artificial neural network model is made also.It is found that the Hei-longtan spring group is an incomplete-drainage overflow karst spring at the piedmont formed by fractures. There is causality between the annual precipitation deficit and the zero flow of the Heilongtan spring group. Finally,a BP artificial neural network model with 6 - 13 - 3 network topology of the Heilongtan spring group’s zero flow is established.The model uses antecedent precipitation,air temperature and humidity as input vector parameters to simulate different conditions of the Heilongtan spring group’s zero flow.Training samples come from data from 1 953 to 2002 ,and testing samples come from 2003 to 2012 in the model.At last,it is found that the testing results are coincide with real situation to great extent.