Find a copy in the library
Finding libraries that hold this item...
|All Authors / Contributors:||
Bertrand Michel; Pascal Massart; Thomas Duquesne; Université de Paris-Sud. Faculté des Sciences d'Orsay (Essonne).
|Notes:||Thèse soutenue en co-tutelle -|
|Description:||1 vol. (210 p.) : ill. ; 30 cm.|
|Responsibility:||Bertrand Michel ; [sous la direction de] [Pascal Massart [et de] Thomas Duquesne.|
This thesis proposes a modelling of the oil production in a hydrocarbon basin. The model is built on a probabilistic description of reserves, of the exploration process and of the launching process for the discovered fields. The use of the Levy-Paréto distribution to model field sizes is justified first by a probabilistic modelling of the reserves creation during the evolution of the geologic time and second by the invariance properties of the Poisson Dirichlet distribution under coalescence and fragmentation operations, within the Bolthausen Sznitman model framework. Two main statistical problems of model selection in the density estimation framework are identified. The first topic is about the estimation of the oil exploration model and the second is a production curve study which is carried out with a clustering and a variable selection obtained by the selection of a Gaussian mixture model. In both cases, a maximum likelihood criteria is defined in order to achieve an oracle inequality. The complete model for oil production in a basin allows to specify the shape of basin production profile. Its also allows to propose production scenarios in the future for producing basins.