Modelagem de Um Sistema de Abastecimento de Água Baseado em Redes Neurais Artificiais

Abstract

The use of control systems is often associated with efficient and effective models. Currently, many water supply systems, which are essential to society, lack adequate control systems, resulting in losses in the process and energy inefficiency. In control system design, the modeling of the system becomes highly important, as it is nonlinear and complex in water supply systems, requiring the development of black-box models using Artificial Intelligence methodologies. Thus, this study aims to develop a model based on artificial neural networks for a water supply system. To achieve results, a pilot plant from the Laboratory of Energy Efficiency in Hydraulics and Sanitation (LENHS/UFPB) was utilized, equipped with a fully instrumented and automated water distribution system. To develop the model, a methodology consisting of four steps was followed: data acquisition, data processing, neural network modeling, and implementation in the supervisory system. The maximum value obtained for the validation metric, mean absolute percentage error, was 1.532% for pressure and 1.7239% for flow rate. These results solidify the effectiveness of the method and its applicability in water distribution systems.