Pressure Reduction Forecasting in Urban Water Distribution Systems Using EPANET and Machine Learning Models

Ageing phenomena are inevitable in urban water distribution systems (WDSs). One of the most popular techniques to reduce the consequences of water losses caused by ageing is the management of hydraulic parameters such as pressure reduction in the water mains. In this study, aiming to investigate the effect of pressure reduction on leakage, EPANET 2.2 software is used to simulate an urban water distribution network. The application of Machine Learning (ML) models such as ANFIS (Adaptive Neuro-Fuzzy Inference System)-Genetic Algorithm (GA), ANFIS-Particle Swarm Algorithm (PSO), and Extreme Learning Machine (ELM) is evaluated to reduce damage due to high operating pressure in a WDS while considering the measured values of head loss and velocity data through hydraulic simulation caused by diurnal demand patterns. In order to investigate the difference between the historical and estimated values, the Root Mean Square Error (RMSE), Mean Absolute Error (MAE), Akaike Information Criterion (AIC), and R are used. A real-world case study is selected to apply the proposed models. After the application of Machine Learning, the obtained results indicate that the ELM technique provides an appropriate tool for predicting pressure in the WDS with minimum error and high desired accuracy. This means that the implementation of the results of the proposed ML model in a real urban WDS is feasible and plays a key role in reducing water losses.