In the petroleum industry, characterization of reservoir properties such as porosity and permeability is definitely essential for a design of reliable development plans and optimization of proved reserves. Geostatistical modeling is one of well-known methods to estimate the reservoir parameters from various data sources. Generally, reservoir property distributions are estimated based mainly on directly sampled data (hard data) like well data, yet the number of hard data is not enough to make the highly accurate reservoir model. On the other hand, indirectly sampled data (soft data) such as seismic data is less reliable but is distributed extensively.
In this study, the main objects are to improve the quality of geological modeling by integration of multiple soft data of different scale and to verify that soft data are useful for the refinement of estimation. Traditional geostatistical simulation methods cannot incorporate multiple soft data of different scale. To solve this problem, we developed the geostatistical program that can estimate reservoir properties by applying both well data and soft data of different scale such as seismic data and pressure transient test analysis results. This program enables the estimate not only by conventional methods such as Sequential Gaussian Simulation (SGS) but also by the new ones incorporating up to two soft data: specialized Markov-Bayes Simulation, Trend Modeling (TM), the combined method of SGS and block kriging (SGSBK), and SGS with Bayesian Updating (SGSBU).
To verify the function of the program thus developed, we carried out two types of estimates using hypothetical data:
permeability distribution modeling in 2-dimensional field and
porosity distribution modeling in 3-dimensional field.
The verification includes the comparison between simple methods using only hard data and special methods described above. The results reveal that the integration of soft data to the estimate can improve the accuracy of modeling. In addition, this study suggests that the incorporation of multiple soft data could be effective, even if the scale of the soft data is different from that of the hard data, to estimate properties in specific areas with the insufficient number of hard data.
This study concludes that soft data are useful to increase the accuracy for modeling reservoir property distributions, and that applying multiple soft data fairly improves the reliability of estimation for the regions where hard data do not exist sufficiently.