Abstract

The design of efficient and robust utility networks, such as water distribution systems, is an essential task in modern society. In fact, there is an increasing demand for these systems to be highly reliable in their operations. In this context, the level to which the network is able to provide the required service needs to be assessed by taking into account uncertain conditions [4]. These conditions include future demands and supply as well as network operational capacity [3]. In this work, the hydraulic performance assessment of large scale water distribution networks in presence of uncertainty is considered. In this context, the network performance assessment is referred to its reliability, sensitivity and leak regions detection. The estimation of the network reliability and sensitivity is performed by an efficient Markov chain Monte Carlo method [1]. Nodal demands, nodal heads and pipe coefficients are model as uncertain parameters with different levels of variability. Failure is assumed to occur when the head of some node is lower than a minimum allowable value. On the other hand, a Bayesian system identification methodology is proposed for leakage detection. In particular, a class of sequential particle filter methods called the transitional Markov chain Monte Carlo method is adopted in the present work [2]. A model class selection is used to identify leak locations and intensity. The methodology properly handles the unavoidable uncertainties in measurement and modeling errors. The effectiveness of the proposed stochastic framework is demonstrated with the analysis of a real high dimensional water distribution system consisting of a large number of nodes and pipes (of the order of thousands). The methodology gives an important insight into the performance, reliability and sensitivity of complex utility networks.