Multilayer Artificial Neural Networks Associated With Genetic Algorithms for Demand-Based Operation and Forecasting of Water Distribution Networks Aiming to Increase Hydroenergetic Efficiency

The limited availability of natural resources, notably water and energy, poses significant challenges to maintaining a reliable water supply for populations, thereby necessitating the efficient management of water distribution networks. Employing optimization methods such as genetic algorithms enables the effective operation of these networks; however, identifying optimal solutions alone is insufficient due to pressure fluctuations resulting from variable demand. To enhance system adaptability, perceptron-type Artificial Neural Networks (ANNs) with seven hidden layers were implemented to forecast operational parameters over the preceding 72 hours. For both distribution networks studied, predictions largely conformed to the requirements of the NBR 12218/2017 Standard, with loss reductions reaching 27.3% at the least favorable time (67) and 37.7% at the most favorable time (69). The integration of Artificial Neural Networks with Genetic Algorithms demonstrated robust predictive capabilities for network operations in complex settings, outperforming standalone optimization algorithms and indicating substantial potential for broader applications, including implementation in real-world networks.