2019 AAPG Annual Convention and Exhibition:

Datapages, Inc.Print this page

Reservoir Prediction Based on Geostatistical Inversion by the Facies Trend Model: A Case Study From the Gudian Block, Northeast China


Abstract Accurate reservoir prediction in the early stage of gas field exploration generally is very difficult because few seismic data and well logs data are typically available. The uncertainty throughout the reservoir prediction process causes the illusions of the geological interpretation. Using more advanced techniques of the reservoir prediction to improve the credibility of the prediction results is necessary. In order to reduce the uncertainty in the reservoir prediction, we analyzed seismic data and well logs data from tight reservoirs in the Gudian block of the northeast China, and evaluated the coincidence rate between new and conventional methods. The Shahezi Formation in the Gudian Block of the northeast China is one of the major reservoirs. A slope-type fan deltaic system which shows obvious chaotic seismic reflections on the seismic section is deposited in the west of the Shahezi Formation, whereas a braided deltaic system which shows progradational seismic reflections is deposited in the east of the Shahezi Formation. The conventional model-based seismic inversion methods are difficult to describe the boundaries of the fan while ensuring the high precision of the braided river prediction. So we added a constraint called the facies trend model to the geostatistical inversion. The facies trend model is a model that shows the probability of the sedimentary facies on the basis of the stratigraphic framework. First, through core data, conventional well logs data, FMI data, and post-stack seismic data, we characterized the boundaries of the fan delta. Secondly, we used well logs data to interpret lithology. Meanwhile, we used probability density function (PDF) and variogram to determine the overall trend of lithology data distribution in the study area. Then, we use the gravel content and the boundaries of the fan delta to establish a facies trend model. Finally, combining the low-frequency stratigraphic framework model, we used the Markov chain Monte Carlo (MCMC) methods to perform the seismic inversion under the constraints of the facies trend model. The inversion results are output when the artificial synthetic seismic trace and the original seismic trace have the smallest residual. This study implies that geostatistical inversion by facies trend model is better than conventional seismic inversion methods in predicting complex reservoirs which are similar to the Shahezi Formation in the Gudian Block of the northeast China, and reduce the uncertainty of geostatistical inversion. At the same time, the accuracy of the facies trend model in the process of the geostatistical inversion directly determines the credibility of reservoir prediction results.