Desenvolvimento de modelo computacional com aplicação do Kalman Filter para detecção de vazamentos em redes de distribuição de água

Abstract

In water supply systems, one of the types of loss are the Real Losses, mainly caused by leaks in water distribution networks and, therefore, constitute one of the most complex and urgent problems to enable a water resources management system for the urban environment that is effective, both from an environmental and financial perspective. Among the various approaches for combating and controlling leaks in water distribution networks, there is the use of probabilistic methods that can translate with reasonable accuracy the functioning of this type of dynamic system and among these methods is the Kalman Filter, largely used in the scientific field as a state prediction tool from the combination of data modeled with errors and data obtained through sensors, whose measurement has distortions that affect its accuracy. In this work, the objective was to review the main concepts of this application and analyze the viability of the method as an operational instrument that can be incorporated in a viable way in the daily life of operators of water supply systems. For this, the analysis parameters were established for a real water supply system in the city of São José do Rio Preto, São Paulo. Hydraulic modeling was carried out based on data from the water supply service with the EPANET software and, later, real measurements were carried out at specific points of this network that served as a basis for the application of the Kalman Filter to determine leaks. From the collected data, a python code was developed that allows the determination of residual values of the iterative calculation of the Kalman Filter and that, therefore, demonstrated the relationship between these values and the occurrence of leaks in the distribution network. Despite the positive result, the study was able to determine that the method has limitations, such as the need for diffuse pressure measurements along the distribution network and the existence of a reliable system record, which in a large part of Brazil can be important obstacles.