Improved EPANET Hydraulic Model with Optimized Roughness Coefficient using Genetic Algorithm

Abstract

The process of calibrating hydraulic models for water distribution systems (WDS) is crucial during the model-building process, particularly when determining the roughness coefficients of pipes. However, using a single roughness coefficient based solely on pipe material can lead to significant variations in frictional head losses. To address this issue and enhance computational efficiency, this study proposes a single-objective procedure that utilizes Genetic Algorithm (GA) for optimizing roughness coefficients in the EPANET hydraulic model. EPANET-GA incorporates an automated calibration process and a User Graphic Interface (GUI) to analyze the water head pressures of WDS nodes. Notably, the proposed method not only optimizes roughness coefficients based on pipe material but also spatial characteristics of pipes. To demonstrate the effectiveness of this method, the study builds a hydraulic analysis model for the Zhonghe and Yonghe district of the Taipei Water Department, integrating graph theory’s connectivity and the GIS database. The model was optimized with 34,783 node items, 30,940 pipes, and 140 field measurements. Results show that the optimized roughness coefficient produces a high correlation coefficient (0.9) with the measured data in a certain time slot. Furthermore, a low standard error (8.93%) was acheived compared to 24-hour monitoring data. The proposed method was further compared to WaterGEMs, and the study concludes that the proposed model provides a reliable reference for the design and routing scenario of WDS.