Predicting combined sewer overflows chamber depth using artificial neural networks with rainfall radar data.

Area of Science:

• Environmental Engineering
• Water Resource Management
• Computational Fluid Dynamics

Background:

• Combined sewer overflows (CSOs) are prevalent in urban drainage, discharging excess water during storms.
• UK water industry utilizes extensive monitoring systems to gather data on CSO performance.
• Traditional hydraulic models are computationally intensive for real-time CSO management.

Purpose of the Study:

• To investigate the efficacy of artificial neural networks (ANN) for predicting CSO hydraulic performance.
• To explore ANN as a viable, data-driven alternative to conventional hydraulic modeling.
• To enhance predictive accuracy by incorporating rainfall intensity data from radar devices.

Main Methods:

• Development and training of an artificial neural network (ANN) model using historical CSO depth and rainfall data.
• Utilizing time series analysis for input data processing.
• Employing the pseudo-inverse rule for ANN model training and validation.

Main Results:

• The trained ANN model demonstrated high accuracy in predicting CSO chamber depth.
• Achieved prediction errors of less than 5% for data points over one hour ahead.
• Validated performance using real-world CSO data from a North England catchment.

Conclusions:

• ANN models provide a robust and accurate method for predicting CSO hydraulic performance.
• This predictive capability is crucial for proactive management of combined sewer systems.
• Data-driven approaches like ANN offer significant advantages over traditional hydraulic models for real-time applications.