Scholarlink

<p>Accurate forecasting of wastewater quality is critical for the efficient operation of wastewater-treatment plants (WWTPs), yet the strongly nonlinear and dynamic behaviour of WWTPs processes make prediction challenging. This study develops a hybrid modelling framework that couples signal decomposition with deep learning to enhance prediction of ammonia-nitrogen (NH₃-N), a key water-quality indicator. First, the whale optimization algorithm (WOA) adaptively tunes the parameters of variational mode decomposition (VMD), enabling data-driven identification of optimal decomposition settings that suppress noise in the raw NH₃-N time series and improve prediction stability. The resulting intrinsic mode functions are then fed into an Informer network, whose individual forecasts are integrated to yield the final NH₃-N estimate. Experiments based on field datasets from monitoring stations at two different WWTPs demonstrate that the proposed WOA-VMD-Informer model achieves coefficients of determination (R²) of 0.88 and 0.96, respectively. Compared to the best-performing single-model benchmarks, the accuracy is improved by 4.22% and 4.65%, respectively. These results verify the model’s suitability for wastewater-quality prediction and provide a technical foundation for intelligent wastewater management and sustainable urban development.</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

Hybrid ammonia-nitrogen forecasting model for wastewater treatment plants based on the WOA-VMD-informer algorithm

  • 期刊论文

  • 发表日期：

    2026年5月14日

查看全文
  • Kejie Xue,
  • Ying Tian,
  • Zeqiu Li,
  • Dongxiang Fu

摘要

Accurate forecasting of wastewater quality is critical for the efficient operation of wastewater-treatment plants (WWTPs), yet the strongly nonlinear and dynamic behaviour of WWTPs processes make prediction challenging. This study develops a hybrid modelling framework that couples signal decomposition with deep learning to enhance prediction of ammonia-nitrogen (NH₃-N), a key water-quality indicator. First, the whale optimization algorithm (WOA) adaptively tunes the parameters of variational mode decomposition (VMD), enabling data-driven identification of optimal decomposition settings that suppress noise in the raw NH₃-N time series and improve prediction stability. The resulting intrinsic mode functions are then fed into an Informer network, whose individual forecasts are integrated to yield the final NH₃-N estimate. Experiments based on field datasets from monitoring stations at two different WWTPs demonstrate that the proposed WOA-VMD-Informer model achieves coefficients of determination (R²) of 0.88 and 0.96, respectively. Compared to the best-performing single-model benchmarks, the accuracy is improved by 4.22% and 4.65%, respectively. These results verify the model’s suitability for wastewater-quality prediction and provide a technical foundation for intelligent wastewater management and sustainable urban development.