Enhancing hydrogen sulfide control in urban sewer systems using machine learning models: Development of a new...

Area of Science:

• Environmental Engineering
• Wastewater Management
• Computational Science

Background:

• Sewer networks are vital urban infrastructure, but hydrogen sulfide (H2S) generation causes odor and corrosion.
• Conventional models struggle with non-linear data in H2S prediction.
• Machine learning offers advanced capabilities for forecasting complex environmental data.

Purpose of the Study:

• To develop a novel machine learning approach for predicting hydrogen sulfide formation in sewer systems.
• To compare the performance of boosting and traditional machine learning algorithms for H2S simulation.
• To identify optimal sewer operational parameters for minimizing H2S generation.

Main Methods:

• Employed 11 machine learning models, including 4 boosting and 7 traditional algorithms.
• Utilized over 700 datasets analyzing sewer operational parameters (pH, DO, temperature, weather, sulfate, ammonia).
• Evaluated model performance using R-squared and Root Mean Square Error (RMSE).

Main Results:

• eXtreme Gradient Boosting (XGBoost) demonstrated superior prediction efficiency (R=0.97, RMSE=0.177 mg/L).
• The XGBoost model successfully predicted H2S formation across various sewer networks, validated against literature data (R>0.9).
• Optimal conditions for minimizing total sulfide generation were identified using the XGBoost model.

Conclusions:

• Boosting machine learning, particularly XGBoost, is highly effective for simulating non-linear H2S concentrations in sewer systems.
• The developed model provides a robust tool for predicting and controlling H2S formation.
• Findings can enhance the operational control and management of sewer systems to mitigate H2S issues.