基于机器学习的采煤沉陷区水体富营养化监测  

Monitoring of Water Eutrophication in Coal Mining Subsidence Areas Based on Machine Learning

作　　者：郎建 李小龙[1] 张世文[1] 徐勇 杨金香[1] 陈永胜 黄智慧 LANG Jian;LI Xiaolong;ZHANG Shiwen;XU Yong;YANG Jinxiang;CHEN Yongsheng;HUANG Zhihui(School of Earth and Environment,Anhui University of Science and Technology,Huainan Anhui 232001,China;Huahai Energy Health Industry Group Co.,Ltd.,Huainan AnHui 232001,China)

机构地区：[1]安徽理工大学地球与环境学院,安徽淮南232001 [2]淮河能源健康产业集团有限责任公司,安徽淮南232001

出　　处：《安徽理工大学学报(自然科学版)》2024年第6期99-108,共10页Journal of Anhui University of Science and Technology:Natural Science

基　　金：采煤地表沉陷区水面种植关键技术及模式研究与示范(HX2024012395);安徽省高潜水位矿区水土资源综合利用与生态保护工程实验室重点项目(2023-WSREPMA-01);安徽理工大学引进人才科研启动基金(2022yjrc01)。

摘　　要：目的水体富营养化及藻类大面积爆发是当前水环境治理中普遍存在的难题,严重威胁水生态系统的健康与安全。为提升对水体富营养化的监测能力,以淮南市顾桥矿采煤沉陷积水区“漂浮水稻”种植试验区为研究对象,提出一种基于机器学习的快速监测方法,为水体富营养化的监测提供技术支持。方法通过收集水温、pH值、溶解氧、浊度、电导率和氨氮等水质理化指标数据,构建了支持向量机(SVM)和随机森林(RF)模型,模型训练采用80%的数据用于建模,20%的数据用于测试,同时优化模型的特征选择和参数调整,以提升预测精度和泛化能力,实现快速监测水体富营养化状态。结果研究结果显示,SVM模型关于水体富营养化指数模拟的训练R和测试R分别为0.76和0.91,与RF模型的0.72和0.69相比,SVM模型在非线性数据处理方面展现出更高的准确性,能够为“漂浮水稻”种植区水体的富营养化监测提供更可靠的技术支持。结论成功构建了水体富营养化监测模型,其中SVM模型凭借更高的非线性处理能力,为“漂浮水稻”种植区水体的富营养化监测提供了可靠的技术支持。对推动水体富营养化实时预警和精准响应具有重要意义,并为水环境监测与治理提供了新的思路和方法。Objective Water eutrophication and large-scale algal blooms are common challenges in current water environment management,posing significant threats to the health and safety of aquatic ecosystems.To enhance the monitoring capability of water eutrophication,this study focuses on the“floating rice”cultivation test area in the coal mining subsidence water accumulation zone of the Guqiao Coal Mine in Huainan City.A machine-learning-based rapid monitoring method was proposed to provide technical support for water eutrophication monitoring.Methods The Support Vector Machine(SVM)and Random Forest(RF)models were constructed by collecting key physicochemical water quality indicators,including water temperature,pH,dissolved oxygen,turbidity,electrical conductivity and ammonia nitrogen.The dataset was divided into 80%for model training and 20%for testing.Feature selection and parameter optimization techniques were applied to improve the prediction accuracy and model generalization,realizing the rapid monitoring of water eutrophication status with enhanced precision and reliability.Results The study revealed that the SVM model achieved training and testing correlation coefficients(R)of 0.76 and 0.91 respectively,outperforming the RF model,which recorded R values of 0.72 and 0.69.The SVM model demonstrated superior accuracy in handling nonlinear data,making it a more reliable tool for monitoring eutrophication in the"floating rice"cultivation area.Conclusion The water eutrophication monitoring model SVM,exhibiting higher capabilities in processing nonlinear relationships,provides a robust technical support to the monitoring of eutrophication in"floating rice"cultivation areas.The research contributes significantly to the advancement of real-time warning systems and precise responses to water eutrophication,offering new approaches and methods for water environment monitoring and management.

关 键 词：机器学习 漂浮水稻 支持向量机 随机森林 水体富营养化 

分 类 号：X832[环境科学与工程—环境工程]