Estimação da rugosidade em sistemas de abastecimento de água utilizando algoritmo genético e gêmeo digital

Abstract

A calibration of water supply systems is necessary at certain times due to the change in roughness, which is a parameter that changes over time or network conditions. The increase in roughness over time changes the hydraulic conditions of a pipe, as it reduces the pressure and diameter, which can bring blockages and overloads to the system, making a new calibration necessary to predict the model. In this work we present two projects of calibration methodology in water through genetic algorithms and digital twins in water supply networks using roughness, pressure drops and pressures as variables for optimization. The first method consists of creating a digital twin of an experimental bench that simulates the water supply in the Laboratory of Energy and Hydraulic Efficiency in Sanitation (LENHS) to estimate the roughness of the pipe. The digital twin was modeled using the data from the system components catalogs as parameters. The methodology consists in the use of an optimization algorithm (Trust Region Reflective) to change the roughness values and local losses of the pipe, seeking to minimize the error of the pressure and flow sensors. The second method was the implementation of a genetic optimization algorithm that was used in conjunction with EPANET, where it is possible to change the roughness in order to minimize the error between real and simulated pressure and flow. Scenarios were simulated in digital twin and EPANET under the same conditions for an algorithm performance comparison. The simulated results compared to the experimental ones obtained errors smaller than 5%, attesting that the methodologies used were able to calibrate the system. A comparison was also made between the methods and an error of 5% was obtained between them. Noting that the algorithm used also has good performance for this type of application.