Abstract:

The 3D reservoir modeling technology based on seismic inversion constraints effectively combines seismic data and logging data, which not only reflects the high vertical resolution of logging data, but also considers the lateral variation characteristics of reservoirs reflected by inversion data. It is currently the mainstream method of reservoir analysis. However, in the traditional combination of seismic data and logging data, the weight of the constraints on the final lithofacies is difficult to control. And the traditional full independence or conditional independence hypotheses lead to seismic data and logging data combination algorithms are rough, they may result in inconsistencies such as probability values greater than 1 if the each conditional probabilities are valued independently one from another. An alternative combination algorithm, model τ, is proposed which is not only simple, but also can ensure consistency of results in the presence of complex data interdependencies. At the same time, the algorithm is applied to the lithofacies modeling of D oilfield in Niger, and the average error of sand body probability is reduced by 30 %, which effectively improves the prediction accuracy of the sand body.

Key words: seismic inversion, reservoir modeling, data integration, conditional probability, conditional independence