Predictive models for wastewater flow forecasting based on time series analysis and artificial neural network

Area of Science:

• Environmental Engineering
• Water Resource Management
• Data Science

Background:

• Wastewater flow forecasting is essential for effective wastewater treatment plant (WWTP) operation.
• Challenges include complex precipitation-runoff dynamics, aging infrastructure, and data quality issues.
• Accurate inflow prediction is vital for optimizing WWTP management and resource allocation.

Purpose of the Study:

• To develop and evaluate time series models for predicting wastewater inflow.
• To compare the performance of Autoregressive Integrated Moving Average (ARIMA) and Multilayer Perceptron Neural Network (MLPNN) models.
• To provide decision support tools for WWTP management.

Main Methods:

• Employed ARIMA and MLPNN models for wastewater inflow prediction.
• Utilized 15-minute flow data, resampled to daily data, from Barrie Wastewater Treatment Facility.
• Validated models using Root Mean Square Error, Mean Absolute Percentage Error, Coefficient of Determination, and Nash-Sutcliffe Efficiency.

Main Results:

• Both ARIMA and MLPNN models demonstrated reliable wastewater flow forecasting capabilities.
• ARIMA model exhibited slightly superior performance compared to the MLPNN model in this case study.
• The developed models offer valuable insights for WWTP operational optimization.

Conclusions:

• Time series analysis and artificial neural networks are effective for wastewater inflow prediction.
• ARIMA models present a robust option for wastewater flow forecasting in WWTPs.
• These predictive models enhance decision-making for efficient WWTP management.