Pumping systems play a crucial role in the management of urban wastewater and stormwater, particularly under adverse hydraulic conditions and variable rainfall and climate regimes, which lead to highly nonlinear and dynamic behavior in these networks. Within this framework, the present study proposes the implementation of cascaded neural networks for the characterization and prediction of the hydraulic behavior of pumping systems in sewerage networks.
The proposed cascaded configuration allows the predictive process to be organized in such a way that the information generated by one model can be used in subsequent modeling stages. This approach aims to improve the representation of temporal and functional relationships among the different components of the system. Accordingly, the methodological framework is based on the use of artificial intelligence (AI) techniques for time-series analysis, integrating relevant hydraulic variables such as tank or wet-well levels, pumping flow rates, and rainfall records.
In this context, the main objective is to evaluate the ability of this approach to predict flow behavior and support decision-making in the operation of key infrastructure. This is intended to contribute to the optimization of energy and hydraulic management in pumping stations within sewerage networks. In this way, the study supports the advancement of digitalization in urban water systems.