متدولوژی برای جداسازی نشت با استفاده از تجزیه و تحلیل حساسیت فشار در شبکه های توزیع آب

Leaks are present to some extent in all water-distribution systems. This paper proposes a leakage localisation method based on the pressure measurements and pressure sensitivity analysis of nodes in a network. The sensitivity analysis using analytical tools is not a trivial job in a real network because of the huge non-explicit non-linear systems of equations that describe its dynamics. Simulations of the network in the presence and the absence of leakage may provide an approximation of this sensitivity. This matrix is binarised using a threshold independent of the node. The binary matrix is assumed as a signature matrix for leakages. However, there is a trade-off between the resolution of the leakage isolation procedure and the number of available pressure sensors. In order to maximise the isolability with a reasonable number of sensors, an optimal sensor placement methodology, based on genetic algorithms, is also proposed. These methodologies have been applied to the Barcelona Network using PICCOLO simulator. The sensor placement and the leakage detection and localisation methodologies are applied to several district management areas (DMA) in simulation and in reality.

Water loss in distribution system networks is an issue of great concern for water utilities, strongly linked with operational costs and water resource savings. Continuous improvements in water loss management are applied and new technologies are developed to achieve higher levels of efficiency. Usually a leakage detection method in a District Metered Area (DMA) starts by analysing input flow data, such as minimum night flows and consumer metering data (Lambert, 1994 and MacDonald, 2005). Once the water distribution district is identified to have a leakage, various techniques are used to locate the leakage for pipe replacement or repair. Methods for locating leaks range from ground-penetrating radar to acoustic listening devices or physical inspection (Colombo et al., 2009 and Farley and Trow, 2003). Some of these techniques require isolating and shutting down part of the system. The whole process could take weeks or months with a significant volume of water wasted. Techniques based on locating leaks from pressure monitoring devices allow a more effective and less costly search in situ. This paper presents a model-based methodology to detect and localise leaks. It has been developed within a project carried out by Aguas Barcelona, Water Technological Centre CETaqua, and the Technical University of Catalonia (UPC). The objective of this project is to develop and apply an efficient system to detect and locate leaks in a water distribution network. It integrates methods and technologies available and in use by water companies, including DMA and flow/pressure sensor data, in conjunction with mathematical hydraulic models. The method is based on the analysis of pressure variations produced by leakage in the water distribution network (Pudar & Ligget, 1992). This technique differs from others in the literature, such as the reflection method (LRM) or the inverse transient analysis (ITA), since it is not based on the transient analysis of pressure waves (Ferrante and Brunone, 2003a, Ferrante and Brunone, 2003b, Misiunas et al., 2005 and Verde et al., 2007). Alternatively, the leakage detection procedure is performed by comparing real pressure and flow data with their estimation using the simulation of the mathematical network model. Simulation of the network in presence and absence of leakage provides an approximation of pressure sensitivity of nodes in a network when a leak is present in a node. The approximation is used to generate a sensitivity matrix that is binarised using a threshold independent of the node. In order to successfully apply this methodology, the characterisation of district metered areas and consumers, considered a critical issue for a correct model calibration, should be also addressed but is not described in this paper (see, e.g. Perez, de las Heras, Aguilar, Pascual, & Peralta, 2009a, for further details). Another critical point is the data validation of DMA sensors that can be addressed as it is described for flowmeters in Quevedo et al. (2010). The paper also proposes a methodology for placing pressure sensors within a DMA that optimises leakage detection using a minimum number of sensors based on the approach proposed in Pérez et al. (2009b). Finally, the leakage detection methodology proposed will be tested with sensors installed in a DMA used as case study. Section 2 reviews water distribution network modelling and presents the case study used to illustrate the proposed methodologies. Model-based fault detection and isolation techniques described in Section 3 are used for the leakage detection and location. Section 4 presents how the leak signature matrix is obtained from the pressure sensitivity matrix. Since the sensor placement is a critical issue for maximising discriminability, an algorithm is presented in Section 5. The signature matrix is generated for the set of sensors selected. This matrix has to be compared with the signature obtained comparing the model and the real measurements. From this comparison, the leakage is located in a set of possible nodes. This methodology is presented in Section 6 and is illustrated by simulation and real results. Finally, Section 7 summarises the conclusions.

A leakage localisation method based on the pressure measurements and sensitivity analysis of nodes in a network has been proposed. The leakage localisation methodology is founded in standard model-based fault diagnosis well established theory. In order to maximise the isolability with a reasonable number of sensors, an optimal sensor placement methodology based on genetic algorithms is also proposed. The objective function in the minimisation process was the size of the maximum group discriminated. The confidence of the information provided by pressure sensors about leakage could be studied using the Fisher Information Matrix generated using the sensitivity matrix. This new approach is studied as a possible way to define the sensor placement avoiding the optimisation process. To assess the validity of the proposed approach, it has been applied to a DMA of Barcelona network in real and simulated leak scenarios. Models and information were provided by the water company. For these sectors (DMA), the sensor placement and the leakage detection and localisation methodologies have been applied with successful results even in presence of demand uncertainty in simulation. In real test where sensors used where already installed results were poorer. Two main causes are suggested. First the non-optimal distribution of the sensors thus the methodology proposed in Section 4 is currently being applied in an on-going project in order to improve such results. On the other hand, the estimation of demands should be improved and an evaluation of the influence of the misfit of demand model on the methodology has been studied. First results have been published (Pérez et al., 2011). An issue in the process is to recalculate the sensitivity matrix for each boundary condition using the simulation model because of the high dependence of it to global consumption. This approach is being currently developed using linear parameter varying (LPV) models that consider the consumption as a scheduling variable (Vento & Puig, 2009). Finally, a new approach is being studied that avoids the binarisation of the sensitivity matrix and it is based on correlation of model pressures with leakage and the measurements (Quevedo et al., 2011)