Research The Uncertainty In Reservoir Geological Modeling

According to the Cares Jef s proposition in the book of Modeling uncertainty in the earth science. It is commonly convenient to categorize uncertainty into "accidental" and "epistemic" uncertainty. Additional data or knowledge would not change the accidental uncertainty. Epistemic uncertainty can be reduced with additional data or a more complete understanding of the physical process or linkage of processes. The modeling uncertainty in this paper is mainly the epistemic uncertainty.Uncertainty is a core problem of reservoir modeling study.In the case of imperfect existing data and different research methods, Reservoir modeling results exist great uncertainty, and can not reflect the reservoir geological condition comprehensively.This paper utilizes two modeling methods which are the logging data modeling and the seismic data constraint modeling and analysis the uncertainty of the simulated results.This paper uses the multi-point geo-statistical algorithm simulating the sedimentary micro facies and analysis the uncertainty of the simulated results. When using the logging data modeling, the simulated results of two training image can be compared. When using the seismic data constraint modeling, probability curve can be used to constrain the modeling, and the uncertainty of simulated results can be analyzed. Property modeling can estimate the uncertainty of the modeling results.We find that if only use the logging data, because of the limitation of the logging data, the zone between the different wells will out of control. The uncertainty is very high. But in the modeling based on seismic data, different random seeds and the control level of seismic data can not result make the simulated results have a big difference. This phenomenon proves that seismic data constraint in the modeling can reduce the uncertainty tremendously. The property modeling also prove that the results are reasonable and the level of uncertainty is small.