Scholarlink

<p>Ensuring access to clean drinking water is fundamental for public health and sustainable development, directly supporting several United Nations Sustainable Development Goals (SDGs). Desalination represents a sustainable, non-traditional water resource, particularly for potable use. Assessing water quality is essential to ensure the provision of safe water; however, traditional Water Quality Index (WQI) computations are time-consuming, error-prone, and often affected by manual rounding, subjective weighting, and data-handling inconsistencies. This study introduces a machine learning-based framework to predict and classify the drinking WQI score and water quality class (WQC) for eight full-scale seawater reverse osmosis (SWRO) desalination plants in Egypt. The study integrates real plant operational data with multiple supervised algorithms to enhance predictive accuracy and operational decision-making. Seven regression and seven classification models were developed and compared for forecasting WQI scores and WQC categories, with feature importance analysis conducted using SHapley Additive exPlanations (SHAP). Results revealed that most of permeate water quality parameters met World Health Organization (WHO) standards, except for pH, chloride, boron, and total bacteria. The weighted arithmetic WQI (WAWQI) ranged from 15.9 to 89.5, classifying samples as excellent (71%) and good (29%). Multiple linear regression (MLR) achieved the highest predictive accuracy (R² = 0.9992; RMSE = 0.338), while XGBoost outperformed other classifiers with 93.03% and 95.83% accuracy for validation and testing datasets, respectively. The SHAP analysis revealed that the key influencing parameters were total bacteria, total dissolved solids, sodium, chloride, and residual chlorine. The study demonstrates the potential of data-driven models as reliable and proactive tools for water quality management, supporting optimized operation, reduced monitoring effort, and enhanced decision-making in desalination plants.</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

Machine Learning-Based Assessment of Drinking Water Quality from SWRO Desalination Plants, Egypt

  • 期刊论文

  • 发表日期：

    2025年11月22日

查看全文
  • Ali Nada,
  • Mona G. Ibrahim,
  • Mohamed Elshemy,
  • Manabu Fujii,
  • Ahmed Makhlouf,
  • Mahmoud Sharaan

摘要

Ensuring access to clean drinking water is fundamental for public health and sustainable development, directly supporting several United Nations Sustainable Development Goals (SDGs). Desalination represents a sustainable, non-traditional water resource, particularly for potable use. Assessing water quality is essential to ensure the provision of safe water; however, traditional Water Quality Index (WQI) computations are time-consuming, error-prone, and often affected by manual rounding, subjective weighting, and data-handling inconsistencies. This study introduces a machine learning-based framework to predict and classify the drinking WQI score and water quality class (WQC) for eight full-scale seawater reverse osmosis (SWRO) desalination plants in Egypt. The study integrates real plant operational data with multiple supervised algorithms to enhance predictive accuracy and operational decision-making. Seven regression and seven classification models were developed and compared for forecasting WQI scores and WQC categories, with feature importance analysis conducted using SHapley Additive exPlanations (SHAP). Results revealed that most of permeate water quality parameters met World Health Organization (WHO) standards, except for pH, chloride, boron, and total bacteria. The weighted arithmetic WQI (WAWQI) ranged from 15.9 to 89.5, classifying samples as excellent (71%) and good (29%). Multiple linear regression (MLR) achieved the highest predictive accuracy (R² = 0.9992; RMSE = 0.338), while XGBoost outperformed other classifiers with 93.03% and 95.83% accuracy for validation and testing datasets, respectively. The SHAP analysis revealed that the key influencing parameters were total bacteria, total dissolved solids, sodium, chloride, and residual chlorine. The study demonstrates the potential of data-driven models as reliable and proactive tools for water quality management, supporting optimized operation, reduced monitoring effort, and enhanced decision-making in desalination plants.