Pasanisi, Alberto (2004) Aide à la décision dans la gestion des parcs de compteurs d'eau potable. PhD thesis sciences de l'eau option statistique, ENGREF.

Full text available as: - These_Pasanisi.pdf ( 2311 Kb )

Licence: Copyright

Abstract

Water meters' metrology become more and more inaccurate during their operating life, which originates, in particular for meters actually used in France, the under-estimation of accounted water. That is a source of problems for water distribution companies: it gives rise to significant financial losses and to an unequal billing policy. Furthermore, a more and more severe national standard will soon oblige water companies to keep the rate of inaccurate devices below a fixed value. Thus, the planning of meters' renewal is a complicate problem, which needs the determination of an optimal strategy.
Every management method needs firstly the preliminary knowledge of meters' metrology in real operating conditions. The goal of this PhD thesis is to give indications useful to apply optimal management rules, currently used by "Générale des Eaux" water distribution company. Meters' degradation is studied by a (markovian) dynamic model, based on four discrete states, each one of which characterises a more and more inaccurate metrology. Inference calculations are made in a Bayesian framework by MCMC (Markov Chain Monte Carlo) techniques. This estimation method is a very attractive alternative to usual procedures based on the constrained maximisation of likelihood.
Finally we show as the statistical model can give forecasts which can be directly use by decision-makers: the estimation of unaccounted-for water and non-conformity probability as a function of age, aggressiveness, and annual consumption.

Statistiques de consultation

Repository Staff Only: edit this item