摘要
通过BP神经网络模型对石油开采区土壤重金属的部分缺失试验数据进行补齐,并采用主成分分析法对石油开采区土壤中重金属进行源解析。结果表明,土壤中重金属的来源包括自然来源、农业来源、交通来源和燃煤来源。以不同来源重金属作为输入条件、土壤生物毒性作为输出条件构建不同来源重金属土壤发光菌生物毒性神经网络模型,模型验证结果表明,在0.05显著性水平下,25组验证样本模拟值和试验值之间的相关系数r为0.396,满足模型验证要求。结合相对灵敏度计算,获得不同来源重金属对土壤生物毒性的贡献率分别为自然来源26.68%、农业来源52.71%、交通来源4.67%、燃煤来源15.94%,即农业来源为石油开采区土壤中发光菌生物毒性的最主要来源。
The paper is entitled to provide a contribution analysis of the heavy metal contributive distributions in the soil from the different sources to the biological toxicity based on the BP neural network model. For the above mentioned purpose,we have chosen 6 oil wells in the oil exploration and transferring area in this paper by taking a sum of nearly 400 soil samples as our study subjects. At the same time,we have also investigated and examined 10 kinds of heavy metals( Pb,Cu,Zn,Ni,Fe,Cd,Cr,Co,Sb,Mn) and their toxic contamination soil luminescent bacteria. Thus,we have collected and analyzed the partially missing experimental data of such heavy metals in the soil under the input conditions of the Pb,Cu,Zn,Ni,Fe and the output conditions of Cd,Cr,Co,Sb,Mn by using the BP neural network model.As a result,it has been proved that the contaminant metal percentages of the average relative deviation of Cd,Cr,Co,Sb,Mn account for 19. 85%,10. 95%,7. 21%,7. 58%,and 11. 28%,respectively,with the correlation coefficient r between the modeled values and the measured ones being 0. 87,0. 89,0. 86,0. 92,and 0. 81,correspondingly. At the same time,the simulation efficiency coefficient has also been worked out at about 0. 86,0. 87,0. 85,0. 91,and 0. 81. Furthermore,we have also found out the source apportionment of the aforementioned heavy metals in the soil of the oilfield through the principal component analysis method. The results of the above study have shown that the sources of such heavy metal pollutants in the soil of the oilfield were made of natural origins,some agricultural results,the traffic and transportation pollutants and burning coal remnants. Moreover,we have established a neural network model of biological toxicity of heavy metals from different sources to luminous bacteria in the soil by taking such heavy metals from the different sources as the input condition,but the soil biological toxicity as the output condition,thus,confirming that the correlation coefficient r of the25 groups of verification samples should be equal to 0. 396,which can satisfy the verification requirement of the model. Besides,the relative sensitivity calculation has also helped us to determine the soil biological toxicity of such heavy metals from the respective sources at the following contribution rates of 26. 68% from natural source,52. 71% from the agricultural source,4. 67% from transportation source as well as 15. 94% from the coal combustion source. That is to say,in final analysis,the agricultural source remains the chief source of biological toxicity to the luminous bacteria in the soil of the oilfield under study.
出处
《安全与环境学报》
CAS
CSCD
北大核心
2016年第4期348-352,共5页
Journal of Safety and Environment
基金
国家“十一五”科技支撑计划项目(2008BAC43B01)
关键词
环境工程学
重金属
生物毒性
源解析
BP神经网络
灵敏度分析
environmental engineering
heavy metal
biological toxicity
source apportionment
BP neural network
sensitivity analysis