Comparison of Methods for the Solution of Pressure-Deficient Networks using Artificial Elements

Abstract

When the available pressure at the nodes becomes less than the minimum pressure, the network is said to be in pressure-deficient condition, and under such a condition the demand at the nodes cannot be satisfied. In the pressure-deficient condition, as such, the demand-driven solver does not work and the solution requires that the node flow equation should be embedded either internally or externally in the demand-driven solver. In the last decade, researchers have focused considerable attention on the solution of pressure-deficient conditions using a demand-driven solver. The objective of the study is to compare the methods for the solution of pressure-deficient networks using either artificial reservoirs or emitter features of the demand-driven solver. Both the approaches are applied to three different benchmark water distribution networks, data for which are adopted from literature. In any real-world water distribution network, the field observed data for validation of results of simulation are normally not available because of different demand conditions in real time, condition of the existing pipe, the variation of hydraulic pressure, etc. Thus, this comparative study would be useful for water distribution network practitioners and field engineers to validate the results obtained by one method. It is found that both methods have some common structures that can be utilized in the validation of the results. The comparative study further shows that a simple modification in the structure of one method of solution yields the solution by another method. The results show that the pressure-deficient network can be analyzed successfully using either the artificial reservoir or emitter feature of the graphical user interface of EPANET 2.0. Under the condition of a very small difference between the required and minimum head values at a node, it is found that both the approaches provide a similar result for the water distribution network considered. However, the approach based on the emitter is found to be more flexible in handling the nodal head–flow relation, provided appropriate emitter coefficient and emitter exponent are used. Moreover, it has been observed that in a network, the pressure-deficient nodes are not fixed and change during the simulation period depending on the demand pattern, reservoir level, etc., and hence cannot be identified at the beginning of the simulation.