Skip to Main content Skip to Navigation
Theses

Modélisation dynamique, classification et détection de changement dans les panels catégoriels issus d'un réseau d'eau intelligent

Abstract : Nowadays, we observe a growing concern raised by the environmental issues and those related to management of the resources as electricity and water. As part of a collaborative project with Veolia Eau d'Île-de-France and le syndicat des eaux d'Île-de-France, this PhD research addresses initially the clustering of water consumers based on their consumption behavior dynamics over time. These dynamics, in each cluster, depend on a number of exogenous factors. To model this joint density, non-homogeneous Markov models are investigated as the components of a mixture model. Hence, the estimation of the parameters in each cluster allows to predict the future consumption behaviors independently. Afterwards, the problem of online structural change detection in a set of consumption behavior sequences is addressed. To this end, a sequential hypothesis testing of generalized likelihood ratio, based on a non-homogeneous Markov model is proposed. An adaptive threshold is also used which can be adjusted throughout the various types of changes and may reduce the number of false alarms. The results on a real dataset which is issued from a water network allow to highlight the effectiveness of the proposed methods both in terms of clustering and change detection. Finally, the analysis of the estimated parameters of both models allows to study the influence of exogenous factors on clustering and detected changes
Document type :
Theses
Complete list of metadatas

Cited literature [141 references]  Display  Hide  Download

https://tel.archives-ouvertes.fr/tel-02520488
Contributor : Abes Star :  Contact
Submitted on : Friday, March 27, 2020 - 2:05:16 PM
Last modification on : Friday, July 17, 2020 - 5:08:34 PM
Long-term archiving on: : Sunday, June 28, 2020 - 2:39:06 PM

File

TH2019PESC2003.pdf
Version validated by the jury (STAR)

Identifiers

  • HAL Id : tel-02520488, version 2

Collections

Citation

Milad Leyli Abadi. Modélisation dynamique, classification et détection de changement dans les panels catégoriels issus d'un réseau d'eau intelligent. Ingénierie assistée par ordinateur. Université Paris-Est, 2019. Français. ⟨NNT : 2019PESC2003⟩. ⟨tel-02520488v2⟩

Share

Metrics

Record views

122

Files downloads

74