Saturation and Pressure Inversion - From 4D Seismic to Reservoir Model Updating

We develop a fast petro-elastic inversion using the artificial neural network (ANN) approach. ANN is robust and fault tolerant, capable of handling model non-linearity, and can efficiently deal with large amounts of data. The inversion workflow consists of two main steps; network training on a carefully chosen part of the reservoir model to bring in the engineering insight, followed by the inversion using the trained network. We invert simultaneously the most informative 4D elastic attributes, such as 4D VP, for changes in reservoir state parameters, such as saturations and pressure.

The workflow is demonstrated on a North Sea field, which exhibits a complex 4D signal related to a mix of saturation and pressure changes. We use the petro-elastic inversion results further for conditioning the reservoir model, and show that seismic data may contribute not only to a better understanding of a producing reservoir, but also to reservoir model updating for the purpose of enhancing the predictability of future performance.