AI

Using existing logs, petrophysical parameters, and core data, machine learning models can be trained to predict reservoir presence, properties and heterogeneities near and far from selected well datasets. Training datasets commonly use wells that have drilled known reservoirs, through Q&A and multiple parameter (log) correlation such as GR, density, porosity, acoustic impedance, etc.

3D propagation and prediction of reservoir presence and properties can go both ways: from a set of wells building out to the 3D using available seismic and attributes; or from seismic data being used to target future wells where no reservoir information is presently available. An iterative loop between seismic data and wells can also be established for cross-checking and quality control purposes. Synthetic wells and ties can be created to maximise communication between wells and seismic data and to ensure appropriate labelling of parameters.
The workflow that we currently have under development aims to assist in the automated identification of geological bodies on seismic volumes, such as sand bodies, channels, reefs, etc., using a combination of seismic attributes and input coming from wells.