基于COD的水资源生态环境监测模型研究  

Research on COD-based water resource ecological environment monitoring models

作　　者：戴丽媛 张诚 田晓丹 杨丽雅 赵世雯 沈长松[5] DAI Liyuan;ZHANG Cheng;TIAN Xiaodan;YANG Liya;ZHAO Shiwen;SHEN Changsong(School of Hydraulic Engineering,Wanjiang University of Technology,Ma anshan 243031,Anhui,China;Ma anshan Water Engineering Health Diagnosis and Restoration Technology Research Center,Ma anshan 243031,Anhui,China;Ma anshan Key Laboratory of Landscape Ecological Restoration Research,Ma anshan 243031,Anhui,China;Kunshan Water Comprehensive Management Center High Tech Zone Management Station,Kunshan 215300,Jiangsu,China;College of Water Resources and Hydropower,Hohai University,Nanjing 210024,China)

机构地区：[1]皖江工学院水利工程学院,安徽马鞍山243031 [2]马鞍山市水工程健康诊断与修复技术研究中心,安徽马鞍山243031 [3]马鞍山市景观生态修复研究重点实验室,安徽马鞍山243031 [4]昆山市水事综合管理中心高新区管理站,江苏昆山215300 [5]河海大学水利水电学院,南京210024

出　　处：《安全与环境学报》2025年第4期1603-1608,共6页Journal of Safety and Environment

基　　金：安徽省教育厅2023年高校科学研究重点项目(2023AH052490);马鞍山市水工程健康诊断与修复技术研究中心2023年度开放基金项目(2023msgc002);马鞍山市丘陵地区水资源高效利用工程技术研究中心2023年度开放基金项目(WREU202302);马鞍山市景观生态修复研究重点实验室2024年度开放基金项目(YGYS24001);安徽省教育厅2024年中青年教师培养行动计划项目(JNFX2024098)。

摘　　要：水资源是人类生活和工业生产的重要基础,水质监测对保障水环境安全和促进水资源可持续利用至关重要。为了解决传统水资源监测数据受限于单一数据来源、更新和处理周期长、准确率不高等问题,研究提出了一种基于多源数据融合下的水资源生态环境监测模型。首先,采用紫外光谱和荧光光谱两种光谱源实时监测水资源生态环境,采用光谱法对水中的化学需氧量进行检测,结合光谱数据预处理减少噪声等因素的影响;其次,结合朗伯-比尔定律和高级数据融合方法整合不同光谱源的数据信息;最后,引入最小二乘支持向量机建立水资源生态环境监测模型并进行分析。结果显示,研究模型基本拟合在真实荧光强度附近,且对紫外波长有较高的拟合程度;对4种不同的水质进行测量,研究模型检测的最高准确率为91.2%,最低准确率为88.9%。结果表明,研究模型能够较为准确地对水资源进行检测,并为水资源管理提供可靠的数据。Water quality monitoring is a critical component of water resource management and environmental protection.To address challenges such as limited data sources,prolonged update and processing cycles,and low accuracy in traditional water resource monitoring methods,we propose a water resource ecological environment monitoring model that utilizes multi-source data fusion.This model integrates two spectral data sources,ultraviolet spectroscopy and fluorescence spectroscopy,and employs these spectral methods to measure Chemical Oxygen Demand(COD)in water,thereby assessing the model s performance.First,we preprocess the collected spectral data using baseline correction,normalization,and Savitzky-Golay smoothing techniques to minimize noise interference and enhance data quality.Next,for data processing,we perform quantitative analysis of the spectral data based on Lambert-Beer Law,employing advanced data fusion techniques to integrate information from different spectral sources,thereby improving the model s accuracy.Finally,the model incorporates a Least Squares Support Vector Machine(LSSVM),which integrates an error function into the original framework through its parameters.The Lagrange method is then employed to address the extremum problem of the LSSVM,enhancing its suitability for the high-dimensional nonlinear characteristics inherent in multi-source data.The experiment utilized a self-built spectral dataset that encompassed various water quality samples.Through ablation experiments,we compared the performance of different models.Among the four methods evaluated,the Support Vector Machine(SVM)exhibited the poorest predictive performance,followed closely by the LSSVM.After incorporating the High-Level Data Fusion(HLDF)module into both models,their predictive performance showed significant improvement.The results indicated that the detection accuracy of the High-Level Data Fusion-Least Squares Support Vector Machine(HLDF LSSVM)model across the four water quality samples ranged from 88.9%to 91.2%,demonstrating a parti

关 键 词：环境工程学 水资源 化学需氧量 最小二乘支持向量机 高级数据融合 

分 类 号：X703.1[环境科学与工程—环境工程]