Algorithmes de graphes pour la découverte de la topologie d'un réseau énergétique par la connaissance de ses flots

Abstract : In energy network, the knowledge of equipments, their locations and their functions are theimportant information for the distributor service operator. In fact, each operator has a networkplan often named synoptic schema. That schema shows the interconnexion between equipments inthe network. From this schema, some management decisions have taken for ensuring an optimalperformance of a network.Sometimes, a synoptic schema has some mistakes because the maintenance operations, such aschanged the connexion between equipments or replaced equipments, have not been updated orhave been written with errors. And these mistakes increase exploitation cost in the energy network.We consider an electric network of a datacenter. This network consists of physical topologymodelised by a DAG without circuit and measurements are on the edges of a DAG. The mainpoint of the network is that measurements are some mistakes and the topology is unknown i.ewe know edges but the nodes of edges are unknown. When measurements are correct then thecorrelations between pairwise edges provide the adjacency matrix of the linegraph of undirectedgraph of the DAG. A linegraph is a graph in which each node and the neighbor are partitionnedby one or deux cliques. However, with the mistakes in measurements, the obtained graph is nota linegraph because it contains more or less edges. If the obtained graph is a linegraph then it isa linegraph of the other DAG. Our problem is to discovery the topology of the DAG with somemistakes in measurements.We start by the state of art in the measurement correlations in order to choose the good methodfor our problem. Then, we propose two algorithms to resolve our problem. The first algorithmis the cover algorithm and it returns the set of cliques in the graph. The second algorithm is acorrection algorithm which adds or deletes edges in the graph for getting a nearest linegraph ofthe DAG. In the last, we evaluate the performances of the algorithms by checking the number ofedges corrected and the ability to return a nearest linegraph of the DAG.
Document type :
Theses
Complete list of metadatas

https://tel.archives-ouvertes.fr/tel-01951756
Contributor : Abes Star <>
Submitted on : Tuesday, December 11, 2018 - 4:03:07 PM
Last modification on : Tuesday, May 14, 2019 - 4:52:22 AM
Long-term archiving on : Tuesday, March 12, 2019 - 3:14:54 PM

File

72152_EHOUNOU_2018_archivage.p...
Version validated by the jury (STAR)

Identifiers

  • HAL Id : tel-01951756, version 1

Collections

Citation

Joseph Ehounou. Algorithmes de graphes pour la découverte de la topologie d'un réseau énergétique par la connaissance de ses flots. Modélisation et simulation. Université Paris-Saclay, 2018. Français. ⟨NNT : 2018SACLV056⟩. ⟨tel-01951756⟩

Share

Metrics

Record views

111

Files downloads

71