Detection, Localization and Quantification of Leaks in Pipeline Networks Using a Parameter Estimation Approach

This paper presents a pipe network monitoring system based on a parameter estimation approach. The initial step involves the calibration of a dynamic pipe network simulator to the analyzed network, using a parameter estimation procedure. After the calibration, the network can be monitored, through an online simulation employing updated input data. If a properly statistical criterion indicates a discrepancy between the predicted and measured output variables, a possible leak is detected. Then, automatically, a routine of leak localization and quantification starts. This routine consists of an estimation procedure where the network model includes two additional parameters: the leak diameter and location in a pipeline section. This procedure is applied for each pipeline section selected by the user to be investigated. The section with the best match between predicted and measured data corresponds to the leak information that must be checked by the operation staff.