Urban Water Demand Prediction using ANN

Effective management of municipal water systems is vital for the sustainability of urban areas and ensuring water security for communities. Estimating urban water demand has always posed challenges for water utility managers and policymakers. This study introduces an innovative approach utilizing data preprocessing and an Artificial Neural Network (ANN) optimized through the Backtracking Search Algorithm (BSA-ANN) to forecast monthly water demand based on past consumption patterns. Historical data from monthly water usage in Gauteng Province, South Africa, spanning 2007 to 2016, were utilized to develop and assess the methodology. Data preprocessing techniques significantly enhanced data quality prior to model creation. The BSA-ANN model demonstrated superior performance, achieving a root mean square error of 0.0099 mega liters and a coefficient of efficiency of 0.979. Compared to the Crow Search Algorithm (CSA-ANN), the BSA-ANN model exhibited greater efficiency and reliability based on error metrics. This study introduces a novel application of the hybrid BSA-ANN model, showcasing its potential to accurately predict water demand in urban areas facing the challenges of climate change and population growth.