Imperial College London


Faculty of EngineeringDepartment of Civil and Environmental Engineering

Senior Lecturer



+44 (0)20 7594 6035ivan.stoianov Website




Miss Judith Barritt +44 (0)20 7594 5967




408Skempton BuildingSouth Kensington Campus






BibTex format

author = {Abraham, E and stoianov, I},
doi = {10.1109/TCNS.2016.2548418},
journal = {IEEE Transactions on Control of Network Systems},
pages = {610--619},
title = {Constraint preconditioned inexact Newton method for hydraulic simulation of large-scale water distribution networks},
url = {},
volume = {4},
year = {2016}

RIS format (EndNote, RefMan)

AB - Many sequential mathematical optimization methodsand simulation-based heuristics for optimal control anddesign of water distribution networks rely on a large numberof hydraulic simulations. In this paper, we propose an efficientinexact subspace Newton method for hydraulic analysis of waterdistribution networks. By using sparse and well-conditionedfundamental null space bases, we solve the nonlinear systemof hydraulic equations in a lower-dimensional kernel space ofthe network incidence matrix. In the inexact framework, theNewton steps are determined by solving the Newton equationsonly approximately using an iterative linear solver. Since largewater network models are inherently badly scaled, a Jacobianregularization is employed to improve the condition number ofthese linear systems and guarantee positive definiteness. Afterpresenting a convergence analysis of the regularised inexactNewton method, we use the conjugate gradient (CG) methodto solve the sparse reduced Newton linear systems. Since CGis not effective without good preconditioners, we propose tailoredconstraint preconditioners that are computationally cheapbecause they are based only on invariant properties of the nullspace linear systems and do not change with flows and pressures.The preconditioners are shown to improve the distribution ofeigenvalues of the linear systems and so enable a more efficientuse of the CG solver. Since contiguous Newton iterates can havesimilar solutions, each CG call is warm-started with the solutionfor a previous Newton iterate to accelerate its convergence rate.Operational network models are used to show the efficacy ofthe proposed preconditioners and the warm-starting strategy inreducing computational effort.
AU - Abraham,E
AU - stoianov,I
DO - 10.1109/TCNS.2016.2548418
EP - 619
PY - 2016///
SN - 2325-5870
SP - 610
TI - Constraint preconditioned inexact Newton method for hydraulic simulation of large-scale water distribution networks
T2 - IEEE Transactions on Control of Network Systems
UR -
UR -
VL - 4
ER -