Performance assessment of water distribution systems

Water distribution systems are often susceptible to failure events, mainly due to component malfunctions, increase in demand and pollution events. However, levels of service to the consumers cannot be compromised. Therefore, to understand the behaviour of distribution systems, performance assessment is important. In this thesis, problem of failure events in water distributions system is discussed and the causes of failure are described. Component failures are selected to simulate the extreme situations in the distribution systems. Random nature of the component failures are simulated by way of employing a Monte Carlo technique based on the failure probabilities of the components. The methodology was illustrated with an example application. Appropriateness of existing network analysis methods to simulate failure events is analysed and their shortcomings identified. To demonstrate the impact of component failures, they are simulated with the hydraulic network analysis model. The traditional demand driven network analysis approach is not sensitive to pressure variations in the system. Therefore, simulating failures with demand driven analysis methods produces inaccurate flows at the nodes. The pressure dependent demand analysis on the other hand, is capable of accommodating the flow redistributions in the water distribution network, caused by failure events. The pressure dependent functions used in the analysis are meant to predict the flows that are consumed by the secondary networks (tree network supplied from primary node). However, representing the secondary network behaviour by using only a few coefficients (as in the PDD functions) do not always results in correct predictions. An alternative method that is based on micro level models (secondary networks) is proposed. Micro level models try to simulate the exact network conditions, taking into account of the consumers piping arrangements. Applying micro level models to a large real network will be a tedious process, as the size of the network will increase by many folds. To avoid the difficulties in the micro level modeling, a method based on artificial neural networks (ANN) is introduced. The ANNs mimic the behaviour of secondary networks in the micro level model. Therefore, instead of physically attaching the secondary networks, ANNs are incorporated with the analysis. The ANN based network analysis model predicts the pressure dependent demand outflows at the nodes. The behaviour of water distribution system is evaluated using performance measures. Existing performance indicators are reviewed and their shortcomings identified. New measures are proposed that give better insights into the behaviour of the system and also the failure experience of the consumers. The improved performance assessment method is applied to a case study network and results were explained.