Scholarlink

<p>To address the challenges of low prediction accuracy and limited generalization capability in forecasting complex water quality at refineries, this study proposes a novel hybrid neural network model (CBG). This model integrates convolutional neural networks, bidirectional long short-term memory networks, and grey wolf optimization algorithms. The CBG model demonstrates excellent accuracy in predicting key pollutants such as chemical oxygen demand (COD), oil, and ammonia nitrogen (NH<sub>3</sub>-N). Its correlation coefficients reach 0.95, 0.89, and 0.91 respectively, and the nash sutcliffe efficiency coefficients stand at 0.91, 0.79, and 0.83 respectively, which are significantly superior to those of other benchmark models. Additionally, the study innovatively developed a comprehensive warning water quality index (WWQI). This index, together with the CBG model, forms an integrated prediction and warning framework that triggers alerts when water quality indices exceed pre-set thresholds. This framework provides a valuable tool for the early detection and proactive intervention of risks within the water systems of integrated refining and petrochemical enterprises. This study holds significant practical implications for enhancing water resource utilization and maintaining the stability of production operations. By providing intelligent early warning and proactive risk management tools, this research contributes to improving the operational safety and resource efficiency of industrial water circulation systems. This comprehensive approach provides a clear, quantifiable method for forward-thinking, science-based decision-making in water systems management for integrated refining and petrochemical enterprises, ultimately helping to drive more sustainable industrial practices.</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

Optimizing water reuse in integrated refining and petrochemical enterprises: high-precision prediction of water quality enabling a novel proactive warning index

  • 期刊论文

  • 发表日期：

    2026年1月1日

查看全文
  • Jie Xu,
  • Shaoze Xiao,
  • Jiaying Ma,
  • Duanyang Shangguan,
  • Huaqiang Chu,
  • Jiacai Xie,
  • Xuefei Zhou,
  • Yalei Zhang

摘要

To address the challenges of low prediction accuracy and limited generalization capability in forecasting complex water quality at refineries, this study proposes a novel hybrid neural network model (CBG). This model integrates convolutional neural networks, bidirectional long short-term memory networks, and grey wolf optimization algorithms. The CBG model demonstrates excellent accuracy in predicting key pollutants such as chemical oxygen demand (COD), oil, and ammonia nitrogen (NH3-N). Its correlation coefficients reach 0.95, 0.89, and 0.91 respectively, and the nash sutcliffe efficiency coefficients stand at 0.91, 0.79, and 0.83 respectively, which are significantly superior to those of other benchmark models. Additionally, the study innovatively developed a comprehensive warning water quality index (WWQI). This index, together with the CBG model, forms an integrated prediction and warning framework that triggers alerts when water quality indices exceed pre-set thresholds. This framework provides a valuable tool for the early detection and proactive intervention of risks within the water systems of integrated refining and petrochemical enterprises. This study holds significant practical implications for enhancing water resource utilization and maintaining the stability of production operations. By providing intelligent early warning and proactive risk management tools, this research contributes to improving the operational safety and resource efficiency of industrial water circulation systems. This comprehensive approach provides a clear, quantifiable method for forward-thinking, science-based decision-making in water systems management for integrated refining and petrochemical enterprises, ultimately helping to drive more sustainable industrial practices.