基于模糊BP神经网络的辽河口湿地水质评价

利用模糊BP神经网络法对辽河口湿地不同时期（汛期、非汛期和冰封期）不同区域（核心区、试验区和缓冲区）水环境质量进行水质评价。各样点监测结果表明，核心区全N（TN）和氨氮（NH₃-N）优于表水环境质量Ⅲ类标准，缓冲区和试验区中全N（TN）、全P（TP）和化学需氧量（COD）超Ⅳ类标准。建立以上述4个水质指标为输入变量、包含5个神经元节点的隐含层和1个水质类别输出结果所构成的BP人工神经网络，结合模糊数学综合评价法对输出结果进行隶属度分析。结果表明，核心区、缓冲区和试验区在汛期和冰封期评价结果相同，依次为Ⅲ类、Ⅳ类和Ⅳ类，而在非汛期评价结果依次为Ⅱ类、Ⅲ类和Ⅳ类，可见从试验区、缓冲区到核心区水质逐渐转好，说明湿地对污染物具有一定的净化能力。评价结果与实际监测数据基本吻合，说明模糊BP神经网络综合评价具有客观性和实用性。

Water Quality Evaluation of Liaohe Estuary Wetland Based on Back Propagation Artificial Neural Network

The water quality of Liaohe estuary wetland in different periods (flood,non-flood and freeze up seasons) was evaluated by using fuzzy BP (back propagation) neural network method.Monitoring results at sample plots showed that total nitrogen (TN) and ammonia nitrogen (NH₃-N) in the core area were better than those of class Ⅲ standard of surface water environmental quality.TN,total phosphorus (TP),and chemical oxygen demand (COD) in the buffer and experimental areas did not reach the standard of class Ⅳ.BP artificial neural network was established that included input variables (TN,NH₃-N,TP,and COD),output of five neurons of hidden layer and one water quality category.Combined with the fuzzy mathematics comprehensive evaluation method,the output results were subject to membership degree analysis.The evaluation results showed that the water quality levels of the core,buffer and experimental areas in freeze up period were similar to those in flood season,belonging to class level Ⅲ,Ⅳ and Ⅳ,respectively,and in non-flood season,they were in class level Ⅱ,Ⅲ and Ⅳ,respectively.These figures demonstrated that the water quality was improved gradually from the experimental and buffer areas to the core area,indicating certain purification capability of wetland for pollutants.The evaluation results were in consistence with the actual monitoring data,which indicated that the comprehensive evaluation of fuzzy BP neural network was objective and practical.