Page 83 - AJWEP-22-5

Hybrid optimization for LSTM DO prediction

The precise DO prediction results can further global water quality risks, it offers a practical solution
guide the optimization of specific water treatment for constructing sustainable and intelligent systems for
interventions, such as adjusting aeration intensity, water environment management.
applying chemical oxidants, or implementing From a research significance perspective, this
bioaugmentation measures. This optimization can study extends the application of artificial intelligence
improve treatment efficiency and enhance system optimization algorithms in environmental science,
energy efficiency in managing metals, inorganic strengthening the link between algorithm development
pollutants, and organic contaminants. As data- and practical needs in environmental engineering. It
driven approaches become increasingly prevalent in provides both a methodological basis and a practical
environmental engineering, this hybrid optimization direction for future interdisciplinary research.
framework can also be extended to the modeling and Despite its promising outcomes, this research
predicting of other key pollutants in aquatic systems. It has certain limitations. Oscillation phenomena was
thus provides an algorithmic foundation and decision- observed during the local iteration phase of model
making support for developing efficient, intelligent, training. In addition, extensive verification under
and sustainable pollution control and ecological cross-regional, multi-pollutant, and multi- objective
protection systems. scenarios has not yet been conducted. Future research
will aim to enhance training stability and explore the
7. Conclusion integration of this approach with the Internet of Things
sensing systems to establish an end-to-end, closed-loop
This research presents an innovative hybrid optimization framework for water quality monitoring and decision-
approach that integrates the Nadam optimizer with making feedback.
the DE algorithm. For the first time, this combination In summary, this paper not only proposes an original
is applied to the training of DO prediction models, hybrid optimization framework at the methodological
effectively enhancing the accuracy and stability of level but also explores valuable engineering applications
deep learning models in water quality time-series of environmental artificial intelligence. It thus holds
modeling. Through the dynamic coordination of local significant scientific value, engineering significance,
and global search strategies, this method overcame and social influence.
the issue of local convergence common in traditional
approaches to nonlinear modeling and demonstrates Acknowledgments
excellent performance in analyzing highly sequential
environmental data. The authors would like to thank Prof. Azman and
Compared with existing studies, this study Dr. Suhaili for their valuable guidance and suggestions
makes two methodological breakthroughs. First, during this research. We also acknowledge the use of
it introduces, for the first time in the field of water open-access datasets.
quality prediction, a dual-optimizer collaborative
framework based on an information interaction Funding
mechanism, thus addressing the current lack of
engineering applications of intelligent optimization None.
in environmental prediction. Second, it validates
the robustness and applicability of this method in Conflict of interest
water body prediction tasks, highlighting its strong
algorithmic novelty and adaptability. The authors declare no conflicts of interest.
The social contributions of this research are reflected
in three key areas. First, the proposed prediction model Author contributions
contributes to the development of a more efficient early
warning system for water environments, improving Conceptualization: All authors
the proactive monitoring and response capacity for Data curation: Tu Jun
pollution events. Second, the technical framework Formal analysis: Tu Jun
provides data support and modeling foundation for Writing – original draft: Tu Jun
intelligent water resource allocation and governance Writing – review & editing: Azman Yasin, Nur Suhaili
decision-making. Third, in the context of increasing Mansor

Volume 22 Issue 5 (2025) 77 doi: 10.36922/AJWEP025210165