基于元学习的土壤-地下水有机污染物迁移转化小样本分析  

Meta-learning based small sample analysis of organic pollutant migration and transformation in soil-groundwater systems

作　　者：齐劭乾 张曌天 李建英 封梅青 王美丽 雒寒梦 QI Shaoqian;ZHANG Zhaotian;LI Jianying;FENG Meiqing;WANG Meili;LUO Hanmeng(Hebei Institute of Geological Environment Monitoring,Shijiazhuang 050022,China;Hebei Key Laboratory of Geological Resources and Environment Monitoring and Protection,Shijiazhuang 050022,China)

机构地区：[1]河北省地质环境监测院,河北石家庄050022 [2]河北省地质资源环境监测与保护重点实验室,河北石家庄050022

出　　处：《印染助剂》2025年第4期57-61,共5页Textile Auxiliaries

摘　　要：在土壤-地下水有机污染物迁移转化小样本分析中,由于数据不足,分析结果难以满足实际的精度要求。针对这一情况,提出基于元学习的土壤-地下水有机污染物迁移转化小样本分析。从多个污染场地收集了土壤和地下水样本,采用数据清洗及归一化处理方式整理小样本数据集,避免异常数据影响分析结果,在此基础上基于元学习设计一个能够利用少量样本进行高效学习的元学习框架,通过从多个参数中提取学习特征,完成小样本数据的训练与强化,筛选出与有机污染物迁移转化相关的关键变量,计算土壤-地下水有机污染物的溶质浓度。实验结果表明:基于元学习的小样本分析方法可以了解不同地下水埋深条件下有机污染物的迁移转化规律,该方法可以应用到类似的环境污染迁移转化研究中。In the small sample analysis of soil-groundwater organic pollutants migration and transformation,the analysis results are difficult to meet the actual accuracy requirements due to insufficient data.In view of this situation,a small sample analysis of soil-groundwater organic pollutants migration and transformation based on meta-learning was proposed.Soil and groundwater samples were collected from multiple contaminated sites,and data cleaning and normalization were adopted in the overall small sample data set to avoid abnormal data affecting the analysis results.Based on this,a meta-learning framework was designed based on meta-learning,which could effectively learn from a small number of samples,and complete the training and strengthening of small sample data by learning features from multiple parameters.The key variables related to the migration and transformation of organic pollutants were screened out and the solute concentrations of soil-groundwater organic pollutants were calculated.The experimental results showed that the proposed small sample analysis method based on meta-learning could understand the migration and transformation rule of organic pollutants under different groundwater depth conditions,and the method could be applied to similar environmental pollution migration and transformation research.

关 键 词：元学习 有机污染物 土壤-地下水 迁移转化 小样本分析 

分 类 号：TP39[自动化与计算机技术—计算机应用技术]