摘要
快速精准预测低氧发生对维持水生生态系统的健康有着重要意义,利用皮尔逊相关性分析和最大信息系数两种方法,依据闽江上游水口水库典型渔业养殖区G1、G2和Z1点位2021年3月至2022年3月的数据,从多个水质、气象和水文参数中筛选出影响溶解氧的关键驱动因子。基于机器学习算法,构建了独立BP、皮尔逊相关性-BP、MIC-BP和MIC-SVR等溶解氧预测模型,对比分析了各模型的预测结果。结果表明:电导率、水温、pH、叶绿素a和水位是影响溶解氧的5个主要因素;经过相关性分析筛选后,构建的预测模型性能得到提升,其中最大信息系数(MIC)法的筛选效率优于皮尔逊相关性法的;MIC-SVR模型是最优的溶解氧预测模型,其R 2均大于0.98,RMSE均小于0.56,MAE均小于0.28,可以将溶解氧的预测误差控制在±0.30 mg/L以内。该研究成果可为湖库低氧预测预警提供借鉴。
Rapid and accurate prediction of hypoxia occurrence is of great significance for maintaining the health of aquatic ecosystems.Based on the data of G1,G2 and Z1 points in the typical fishery culture area of Shuikou Reservoir in the upper reaches of Minjiang River from March 2021 to March 2022,Pearson correlation analysis and Maximum Information Coefficient(MIC)were used to screen out the key driving factors affecting dissolved oxygen from multiple water quality,meteorological and hydrological parameters.Based on the machine learning algorithm,four dissolved oxygen prediction models of independent BP,Pearson correlation-BP,MIC-BP and MIC-SVR were constructed,and the prediction results were compared and analyzed.The results showed that conductivity,water temperature,pH,chlorophyll a and water level are the five main factors affecting dissolved oxygen.After screening through correlation analysis,the performance of the constructed prediction model is improved,and the screening efficiency of MIC technology is better than that of Pearson correlation method.The MIC-SVR model is the optimal dissolved oxygen prediction model,and R^(2) is greater than 0.98,RMSE is less than 0.56,and MAE is less than 0.28,which can control the prediction error of dissolved oxygen within±0.30 mg/L.The research results can provide references for the prediction and early warning of low oxygen in lakes and reservoirs.
作者
张鹏
梅书浩
石成春
卓越
李佳昊
宋刚福
ZHANG Peng;MEI Shuhao;SHI Chengchun;ZHUO Yue;LI Jiahao;SONG Gangfu(School of Environmental and Municipal Engineering,North China University of Water Resources and Electric Power,Zhengzhou 450046,China;Fujian Academy of Environmental Sciences,Fuzhou 350013,China)
出处
《华北水利水电大学学报(自然科学版)》
北大核心
2025年第1期87-95,共9页
Journal of North China University of Water Resources and Electric Power:Natural Science Edition
基金
河南省科技攻关项目(222102320023)
2022年科技协同创新专项(61352)
华北水利水电大学高层次人才科研启动项目(40768)。
关键词
水库溶解氧
相关性分析
最大信息系数
BP神经网络
支持向量回归
dissolved oxygen in reservoirs
correlation analysis
Maximum Information Coefficient
BP neural network
Support Vector Regression
作者简介
第一作者:张鹏(1988-),男,讲师,博士,从事水环境数学模型方面的研究。E-mail:zhangpeng2019@ncwu.edu.cn。
