Scholarlink

<p>The quality of the Tuul River has degraded due to population growth, urbanization, and industrialization, underscoring the urgent need for sophisticated water quality monitoring methodologies. The primary objectives are to analyze the physicochemical properties of the Tuul River and to assess the effectiveness of artificial neural network (ANN) models in predicting key water quality indicators. A dataset comprising 1260 measurements of 18 physicochemical parameters from 10 locations was analyzed. The methodology included water quality assessment, correlation analysis, and optimizing ANN neuron layers through a hybrid strategy combining rule-of-thumb and trial-and-error techniques. ANN models were trained using the Levenberg–Marquardt and Bayesian regularization learning algorithms. Results demonstrated the superior performance of Bayesian regularization-based models, particularly for chloride (CLBR 11–9-1) and biochemical oxygen demand (BODBR 11–12-1), with mean square errors of 3.34&#xa0;mg/l and 41.603&#xa0;mg/l and correlation coefficients of 0.992 and 0.92, respectively. This study not only analyze the physicochemical properties of the Tuul River but also presents an approach to optimizing ANN models, highlighting their potential for precise and efficient water quality prediction using reduced datasets.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Application of an ANN Model for Predicting Water Quality Parameters: A Case Study of the Tuul River, Mongolia

  • 期刊论文

  • 发表日期：

    2025年10月4日

查看全文
  • Bolor-Erdene Otgonbaigal,
  • Batsuren Dorjsuren,
  • Amarsanaa Badgaa,
  • Amartuvshin Renchin-Ochir,
  • Dariimaa Battulga,
  • Khureldavaa Otgonbayar,
  • Bilguun Tsogoo,
  • Sonomdagva Chonokhuu,
  • Denghua Yan,
  • Galbadrakh Batnasan,
  • Erdenechimeg Gongor,
  • Undrakh Enkhjargal

摘要

The quality of the Tuul River has degraded due to population growth, urbanization, and industrialization, underscoring the urgent need for sophisticated water quality monitoring methodologies. The primary objectives are to analyze the physicochemical properties of the Tuul River and to assess the effectiveness of artificial neural network (ANN) models in predicting key water quality indicators. A dataset comprising 1260 measurements of 18 physicochemical parameters from 10 locations was analyzed. The methodology included water quality assessment, correlation analysis, and optimizing ANN neuron layers through a hybrid strategy combining rule-of-thumb and trial-and-error techniques. ANN models were trained using the Levenberg–Marquardt and Bayesian regularization learning algorithms. Results demonstrated the superior performance of Bayesian regularization-based models, particularly for chloride (CLBR 11–9-1) and biochemical oxygen demand (BODBR 11–12-1), with mean square errors of 3.34 mg/l and 41.603 mg/l and correlation coefficients of 0.992 and 0.92, respectively. This study not only analyze the physicochemical properties of the Tuul River but also presents an approach to optimizing ANN models, highlighting their potential for precise and efficient water quality prediction using reduced datasets.