基于随机森林回归模型和高频数据的鄱阳湖子湖电导率预测  

Prediction of Electrical Conductivity in Sub-lake of Poyang Lake Based on Random Forest Regression Model and High-frequency Data

作　　者：刘丽贞[1] 黄琪[2] 迟殿委 方朝阳[2] 楚明航 LIU Li-zhen;HUANG Qi;CHI Dian-wei;FANG Chao-yang;CHU Ming-hang(Institute of Microbiology of Jiangxi Academy of Sciences,Jiangxi Academy of Sciences,Nanchang 330096,China;Key Laboratory of Poyang Lake Wetland and Watershed Research,Ministry of Education,Jiangxi Normal University,Nanchang 330022,China;College of Artificial Intelligence,Yantai Institute of Technology,Yantai 264005,China)

机构地区：[1]江西省科学院微生物研究所,江西南昌330096 [2]江西师范大学鄱阳湖湿地与流域研究教育部重点实验室,江西南昌330022 [3]烟台理工学院人工智能学院,山东烟台264005

出　　处：《水电能源科学》2023年第10期50-53,共4页Water Resources and Power

基　　金：江西省重点研发计划(20212BBG73014,20192ACB70014);江西省主要学科学术和技术带头人培养计划——青年人才项目(20212BCJ23034);江西省青年重点基金项目(20192ACBL21022);鄱阳湖湿地与流域研究教育部重点实验室开放基金项目(PK2019006);江西省科学院杰出青年人才培育计划(2021YSBG50004)。

摘　　要：电导率是衡量水质的重要参数,高频监测获取水体中电导率对水质管理具有重要作用。由于野外条件的变化复杂性引起设备故障导致数据缺失时有发生,为优化野外监测体系和插补缺失数据,基于高频监测获取的气象和水体物理指标,结合机器学习模型,预测水体中电导率值。结果表明,随机森林回归模型预测效果最优,其决定系数R~2可达0.996,均方根误差R_(RMSE)为1.31μS/cm,平均相对误差M_(MRE)为0.38%;pH值贡献率最大,是影响电导率的主导因素。研究结果利于优化野外高频监测系统平台,健全高频监测数据,为水质管理提供科学依据。Conductivity is an important parameter to measure water quality.High-frequency monitoring of water conductivity plays an important role in water quality management.Due to the complexity of field conditions,equipment failure often leads to data loss.In order to improve the high-frequency monitoring data,machine learning model was used to predict the conductivity content in water body based on the meteorological and physical indexes obtained from high-frequency monitoring.The results show that the random forest regression model has the best prediction effect,with its determination coefficient R~2 reaching 0.996,root mean square error(R_(RMSE)) 1.31 μS/cm,and mean relative error(M_(MRE)) 0.38%.The pH value contributed the most and was the dominant factor affecting the conductivity.The results are conducive to optimizing the field high-frequency monitoring system platform,improving the high-frequency monitoring data,which provides scientific basis for water quality management.

关 键 词：电导率 随机森林回归模型 高频监测数据 鄱阳湖 

分 类 号：X83[环境科学与工程—环境工程] TV11[水利工程—水文学及水资源]