遗传神经网络在累积性环境风险评价中的应用  

Application of Accumulative Environmental Risks Assessment Based on Genetic Neural Network Model

作　　者：陈凯[1] 黄蕾[1] 方强[1] 

机构地区：[1]污染控制与资源化研究国家重点实验室南京大学环境学院,江苏南京210046

出　　处：《环境监控与预警》2012年第2期1-6,共6页Environmental Monitoring and Forewarning

基　　金：国家水体污染控制与治理科技重大专项项目(2009ZX07527-008;2009ZX07528-005);江苏省环境监测科研基金项目(916);中美国际合作项目(2010DFA91910)

摘　　要：以太湖流域常州段为研究对象,构建了累积性水环境风险评价指标体系,利用主成分分析法选取输入变量,并应用MATLAB建立遗传神经网络综合评价模型。运用遗传算法对BP神经网络的权值和阈值进行优化,将遗传算法全局搜索能力和BP算法局部搜索能力相结合,提高了收敛速度和精度。应用模型对2004—2009年常州市累积性水环境风险进行了综合评价,结果表明,2004—2009年风险综合指数总体上处在中级与高级之间,累积性水环境风险较大;2008—2009年风险综合指数不断增大,趋于低级;农业和畜禽养殖业等面源风险源、污水处理和风险管理投资等控制机制以及人口和环境敏感目标等风险受体是造成太湖流域常州段累积性水环境风险较大的主要原因。An assessment index system for accumulative water environmental risks was established according to the situation of Changzhou,part of Taihu lake basin.Principal component analysis was used to select input variables.Then,MATLAB was used to establish a comprehensive neural network model for evaluation.In this model,genetic algorithm was used to optimize the weight and threshold of network in order to improve the constringency rate and accuracy by taking advantage of both the overall searching ability of GA and the local search of BP.This model was applied to evaluate the index system for accumulative water environmental risks in Changzhou during 2004 to 2009.The results showed that during 2004 to 2009,Changzhou＇s composite index of accumulative water environmental risk was between intermediate level and severe state;the risk composite index was on the rise during 2008 to 2009,and tending toward lower level;non-point risk sources caused by agriculture and livestock farming,risk control mechanism such as sewage treatment and,investment of risk management and risk receptors including people and environment-sensitive targets were the main reasons which led to the high accumulative environmental risks in Changzhou.

关 键 词：遗传神经网络 累积性环境风险评价 太湖流域 

分 类 号：X824[环境科学与工程—环境工程]