Improving numerical efficiency of water networks models reduction algorithm
Date
Advisors
Journal Title
Journal ISSN
ISSN
DOI
Volume Title
Publisher
Type
Peer reviewed
Abstract
Nowadays, it is common that water distribution network (WDN) models contain thousands of elements to accurately replicate hydraulic behaviour and topographical layout of real systems. Such models are appropriate for simulation purposes, however optimisation tasks are much more computationally demanding, hence simplified models are required. Variables elimination is a mathematical method for the reduction of such large-scale models described by non-linear algebraic equations. The approach benefits of preserving the non-linearity of the original WDN model and approximates the original model at wide range of operating conditions. However its compute-intensive nature demanded that its implementation should take into account the development in programming languages and the recently released libraries allowing an optimisation of the executable program for multi-core machines. This will ensure that model reduction application will be able to cope with complex topologies of large size networks. In this paper the process of design and development of the research software is described with focus put on the emerged computational research aspects. It is demonstrated that utilisation of parallel programming techniques and sparse matrices ordering algorithms drastically decrease computational time of the model simplification.