Smart water: how AI is clearing the waters in urban rivers

Researchers have developed a new machine learning system to improve the accuracy and efficiency of sewer-river system models. This innovative approach, detailed in their latest publication, promises to significantly reduce parameter calibration time and enhance model precision in predicting urban water pollution.

The complexity of integrating sewer systems and urban rivers into a comprehensive model has long posed challenges due to extensive computational demands and limited monitoring data. Traditional calibration methods fall short in addressing these challenges effectively.

A recent study (https://doi.org/10.1016/j.ese.2023.100320) published in Environmental Science and Ecotechnology (Volume 18, 2024) introduces an advanced machine learning system designed to improve the accuracy and efficiency of sewer-river system modeling. This innovative technique significantly reduces the time required for parameter calibration and enhances the precision of predictions regarding urban water pollution.

At the heart of this breakthrough research is the ingenious combination of two advanced technologies: Ant Colony Optimization (ACO) and Long Short-Term Memory (LSTM) networks, integrated into a machine learning parallel system (MLPS). ACO is inspired by the foraging behavior of ants to find the most efficient paths, applied here to navigate through the complex parameter space of water models. Meanwhile, LSTM networks, a type of recurrent neural network, excel in recognizing patterns in sequential data, making them ideal for understanding the temporal dynamics of pollutants in sewer-river systems. By marrying these technologies, the researchers have crafted an MLPS capable of performing rapid and precise calibrations of sewer-river models. Traditional methods, often cumbersome and time-consuming, can't match the efficiency or the accuracy of this new approach. Specifically, the MLPS drastically reduces calibration times from potentially months to just a few days, without sacrificing the model's ability to predict pollution levels accurately.

Highlights

• A model calibration method is built with model surrogation and algorithm optimization.
• The process-based models and machine learning interact in a unique way.
• The optimization time of the integrated process-based model could be saved by 89.94%.
• The accuracy of complex models can be improved based on limited data.

Dr. Yu Tian, lead author of the study, states, "The integration of Ant Colony Optimization and Long Short-Term Memory algorithms into our machine learning parallel system represents a significant leap forward in environmental management. It allows for rapid, accurate model calibration with limited data, opening new avenues for urban water system planning and pollution control."

MLPS offers a robust solution for the accurate simulation of urban water quality, essential for effective environmental management. Its ability to quickly adapt to new data and scenarios makes it a valuable tool for urban planners and environmental scientists, facilitating the development of targeted pollution control strategies and sustainable water management practices.

###

References

DOI

10.1016/j.ese.2023.100320

Original Source URL

https://doi.org/10.1016/j.ese.2023.100320

Funding information

This study was supported by the National Key R&D Program of China (2019YFD1100300) and the Fellowship of China Postdoctoral Science Foundation (2020M681105). The authors also acknowledge the State Key Laboratory of Urban Water Resource and Environment, Harbin Institute of Technology (No. 2021TS23).

About Environmental Science and Ecotechnology

Environmental Science and Ecotechnology (ISSN 2666-4984) is an international, peer-reviewed, and open-access journal published by Elsevier. The journal publishes significant views and research across the full spectrum of ecology and environmental sciences, such as climate change, sustainability, biodiversity conservation, environment & health, green catalysis/processing for pollution control, and AI-driven environmental engineering. The latest impact factor of ESE is 12.6, according to the Journal Citation Report^TM 2022.