摘要
利用遗传神经网络模型分析成都市各年人均生态足迹值与其主要影响因素之间的内在关系,并以此来预测未来生态足迹发展趋势。首先计算出成都市1985-2005年共21年的人均生态足迹值,结果表明,20余年来成都市人均生态足迹值缓慢上升,从1.344hm^2升至1.789hm^2。选取6个影响因素指标,建立了6输入、1输出、1个隐含层的三层BP人工神经网络,利用遗传算法的全局优化性能对BP网络进行初始权值优化,以克服BP网络易陷入局部极小点的缺陷。运用遗传神经网络模型对训练样本（1985-2002年人均生态足迹值）进行拟合的精度达99.70%;对检验样本（2003-2005年人均生态足迹值）进行仿真检验,仿真精度为99.10%,高于普通BP神经网络模型（97.89%）,体现出模型较强的应用价值。利用1985-2005年各影响因素的预测值作为网络输入,使用优化后的网络来预测成都市未来几年生态足迹值,最终预测得到2008-2010年成都市人均生态足迹的预测值分别为1.939、1.990、2.049 hm^2人^-1,人均生态赤字分别为1.629、1.688、1.749 hm^2人^-1。为减轻城市发展消耗所导致的生态环境压力,应有效控制成都市人口过快增长、引导消费观念和结构更新。
Genetic neural network model was used to analyze the intrinsic relationship between average ecological footprint and its influencing factors each year in Chengdu,and the developing trend of ecological footprint in the future was predicted.The average ecological footprint in Chengdu from 1985 to 2005 was increased slowly from 1.344hm2 to 1.789hm2.Then 6 influence factors were selected to establish BP artificial neural network,which was made up of one input layer of 6 inputs,one output layer and one hidden layer.The global optimization performance of Genetic Algorithm was used to carry out the initial weights optimization of BP artificial neural network,so as to eliminate the flaws which can easily lead BP artificial neural network into local minima.The fitting accuracy of genetic neural network model to learning samples（average ecological footprint from 1985 to 2002） reached 99.70%;the simulation accuracy to testing samples（average ecological footprint from 2003 to 2005） reached 99.10%,which was higher than ordinary BP neural network（97.89%）,and showed the high application value of the model.The predicted values of each influence factors from 1985 to 2005 were used as network input indexes,and the optimized network was used to predict the ecological footprint in the following years in Chengdu.The predicted values of average ecological footprint in Chengdu from 2008 to 2010 were 1.939 hm2,1.990 hm^2 and 2.049 hm^2,and average ecological deficit were 1.629 hm^2,1.688 hm^2 and 1.749 hm^2.In order to alleviate environment pressure caused by consumption of urban development,excessive population growth in Chengdu should be controlled effectively,and the concept and structure of consumption should be renovated.
出处
《生态学报》
CAS
CSCD
北大核心
2009年第1期359-367,共9页
Acta Ecologica Sinica
基金
教育部重点项目
四川省教育厅配套项目"成都平原土壤质量演变规律及其驱动因子研究"(3110)
关键词
遗传神经网络
成都市
人均生态足迹
genetic neural network model
Chengdu City
average ecological footprint
