URE 2014-18 Integration of Seismic, Well and Production data constrained by Physical/Geological Models

Popular science presentation: Hydrocarbon accumulations in Norway are offshore and located on a couple of thousand meters depth. The reservoir rocks are heterogenuous with sand and shale layers, and various hydrocarbon source rock types. The pore volume is saturated with brine (salt water), oil or gas. In order to efficiently deplete the oil and gas, a reliable mapping of the rock types and pore fillings need to be available. Focus of this project is on this type of reservoir characterization. This mapping must be based on geological interpretations, relatively reliable observations in a low number of wells and abundance of seismic data with highly varying information content. In the project, advanced probabilistic modelling and statistical methodology, socalled Bayesian spatial inversion, is utilized to integrate the available reservoir information. Moreover we quantify the uncertainty in the reservoir characterization. The dynamic changes in the oil and gass content are also modelled by combining reservoir production data and mathematical models for flow in porous media. The research activity is performed by Phd-students under supervision of professors in mathematical sciences and petroleum engineering - in close cooperation with petroleum related industry. The most important contributions so far are seismic full wave form inversion With uncertainty quantification, seismic Bayesian inversion With multi-modal prior models and basin modelling conditioned on seismic data.

Virkninger og effekter av Prosjektet: - Norsk petroleums sektor har fått tilført høy-kvalitets arbeidskraft med matematisk-statistisk kompetanse - gjennom åtte Phd-kandidater samt mer enn 15 MSc-kandidater - langt de fleste av dem norske statsborgere. - URE-sponsor næringsliv er blitt utfordret på måten reservoar-beskrivelse utføres - ny metodikk samt prosedyrer for usikkerhets-kvantifisering er utviklet. - Metodikk innen romlig statistikk inspirert av 'sub-surface' anvendelser er utviklet - dette kan komme til nytte i andre anvendelser - samt at det utvider det statistiske metode-tilfang. - Norsk U&H-sektor synliggjør høy statistisk kompetanse utenlands - dette letter internasjonalt samarbeid på mange områder. - URE-prosjektet har bidratt til heving av mye undervisning på Institutt for matematiske fag, NTNU - som rekrutterer mange av Norges aller beste studenter - uvurderlig viktig !

Project summary: Reservoir characteristics should be represented by a two-level model: top-level as a categorical lithology/fracture/fluid representation and bottom-level as continuous variations within the lithology/fluid classes. The reservoir model sh ould be stochastic in order to represent heterogeneities and to provide predictions with associated uncertainties. The primary reservoir specific informant is seismic data calibrated to well observations - whenever available. The seismic data has good s patial coverage. It is, however, unprecise since it is indirectly collected at the earth's surface, and this uncertainty has large impact on the reliability of the reservoir model. Assessment of this uncertainty in seismic data remain a challenge. Modell ing of spatial categorical variables like lithology/fluids is complicated because of vertical orderings and large-scale lateral patterns, and due to convolution of the seismic data. Moreover, the impact of a high-contrast thin, lateral lithology unit on h ydrocarbon depletion is dependent on structural features like fracturing. Many challenges remain in assessing lithology/fracture/fluid characteristics from seismic and well data. The hydrocarbon depletion is associated with large uncertainties caused by lack of knowledge about both initial state and dynamic properties of the reservoir. Time-lapse seismic data combined with dynamic data from wells provide important information about the depletion process. Many challenges related to assimilation of dynamic seismic and well data and assessment of depletion uncertainties remain. The stochastic reservoir modelling will be cast in a Bayesian inversion setting which includes uncertainty assessment. The research will focus on: . Topic A: Assessment of seismic velocities from seismic data . Topic B: Lithology/fracture/fluid inversion from well and seismic data . Topic C: Fluid monitoring from time-lapse seismic and production data