Objective The aim of the project is to implement and test in proprietary case studies a ready-to-use software system that is used by the seismic interpreter on a 3D seismic interpretation workstation. Application of this software will give the interpreter more accurate and more detailed insight in the subsurface, thus reducing the overall drilling activities and increasing chances of oil discoveries.Software version 1.0 has been released to the sponsors in March 95. The proprietary case study reports have been delivered in April 95.Seismic reservoir characterization techniques have a requirement to integrate data and knowledge from other sources and disciplines. Integration of such data and information is not trivial because of the varying datatypes, scales and accuracies involved.In this project relevant data, from a seismic reservoir characterization perspective, is described in terms of a common subsurface model: the integration framework. Factual wells, i.e. one-dimensional (1D) stratigraphic profiles with attached physical properties, and simulated wells, described in terms of the integration framework, are commensurable. Datasets consisting of wells with corresponding seismic responses can then be compiled from factual and/or simulated data. The objective is to arrive at a dataset that is representative of the target zone in a particular study area. The factual well data is combined with the surface seismic traces at the well locations. This part of the problem space is defined as real space. Simulated wells and corresponding synthetic seismograms are part of the problem space that is defined as model space. The combination of real space and model space is defined as total space. The seismic reservoir characterization technique uses the representative dataset to establish relations between seismic response and underlying salient well properties. The technique is referred to as total space inversion. The "GeoProbe" software which was developed in the Probe project, supports Total Space Inversion concept. In GeoProbe, factual and simulated well data and corresponding factual and simulated well data and correpsonding factual and synthetic seismic responses can be combined into a dataset, representative of the target interval. The system offers a very powerfull well data simulation algorithm in which stochastic input can be combined with geological reasoning to yield realistic one-dimensional geological profiles with attached physical properties. Within the Total Space Inversion concept, GeoProbe offers two options for inverting the seismic data: direct inversion and segmentation.In direct inversion, the representative dataset is tested for relations between seismic response and salient reservoir properties. Two types of artificial neural networks are supported by GeoProbe to establish these relations : Multi-Layer Perceptrons and Radial Basis Functions. Trained networks are subsequently applied to a 3D seismic horizon slice to yield lateral prediction output grids.In segmentation a subset of the factual seismic data is used to train an Unsupervised Vector Quantiser (UVQ) network to classify the seismic response into a number of classes. The trained UVQ is applied to the entire 3D seismic horizon slice to obtain the spatial distribution of the different seismic classes. In a subsequent step a representative dataset consisting of factual and/or simulated well data with corresponding seismic responses is offered to the trained UVQ. Programme(s) ENG-THERMIE 1 - Programme (EEC) for the promotion of energy technology in Europe (THERMIE), 1990-1994 Topic(s) 5.4 - GEOPHYSICS AND PROSPECTING Call for proposal Data not available Funding Scheme DEM - Demonstration contracts Coordinator TNO Institute of Applied Geoscience EU contribution No data Address Schoemakerstraat 97 2600 JA Delft Netherlands See on map Total cost No data
