概率神经网络水质评价模型及其对三峡近坝水域的水质评价分析

Probabilistic neural network model and its application in evaluation of the water quality near the dam area of Three Gorges Reservoir

概率神经网络 (PNN)是一种结构简单、训练简捷、应用相当广泛的人工神经网络 ,在实际应用中 ,尤其是在解决分类问题的应用中 ,它的优势在于用线性学习算法来完成以往非线性学习算法所做的工作 ,同时又能保持非线性算法的高精度等特性。本文结合三峡近坝水域黄陵庙、太平溪、乐天溪和东岳庙等典型断面多年的水质监测数据 ,利用概率神经网络模型对坝区各监测断面 1997年至 2 0 0 2年的水质进行逐月评价 ,得到按月分布的水质评价结果 ,分析表明 :(1)坝区总体水质良好 ,各监测断面水质满足国家对长江干流所要求的地面水Ⅱ类标准 ;(2 )各监测断面水质均呈季节性变化 ,丰水期水质略次于平水期和枯水期 ,年际变化不显著 ;(3 )从不同分类因子的评价结果可以看出营养盐类因子水质评价结果较差 ,需要采取合理而有效的污染控制特别措施。

Probabilistic neural networks(PNN) model is a kind of artificial neural networks which is simple in structure,easy for training and wide used.Practically the linear algorithm can be used to complete the work that was done by the non-linear algorithm while the high accuracy of the non-linear algorithm can be kept.PNN model is applied to monitoring data of water quality at the representative sections in Three Gorges Reservoir from the year 1997 to 2002.The results are listed as follows:(1) Water quality in the reservoir is favorable,and is better than Class Ⅱ of Environmental Quality Standard for Water Quality(GB3838-2002).(2)Water quality at any monitoring section has seasonal changes in one year but unobvious varieties between many years.(3) Nutrients are the key factors that affect the water quality in the reservoir,which will potentially bring on water pollution near the dam of the reservoir.So,it is important to pay more attention to bring into effective measures for pollution control.