Application of SCADA system historical data for developing machine learning models in detection and reduction of water loss

Water loss management in water supply systems is crucial for improving distribution efficiency and reducing costs. This paper investigates the application of SCADA system historical data and the use of machine learning in model development, with a focus on detection, analysis, and reduction of losses. By using data on pressure, flow, and key parameters, a model based on the Random Forest algorithm has been developed to predict losses and identify high-risk moments. The analysis of residuals between predicted and actual values enables quick detection of irregularities, such as leaks or water theft, allowing proactive responses to potential issues.