Reservoir Prediction Based On2D Seismic Data

C8reservoir is characterized by various changes in sedimentary facies and thesand in LMY area of Ordos Basin. The traditional methods based on wells often meettrouble in reservoir prediction for reservoir evaluation and development. Recentdrilling operation shows that higher risk appears in the results from traditionalmethods. This paper aims at the applicability of2D seismic data in reservoirprediction and its application methods in order to improve the reliability of reservoirprediction.There are significant differences among2D seismic lines. Standardizedprocessing of seismic attributes is the key step before the reservoir prediction.Standardized methods for2D seismic attributes in the measured line intersectionconstraints optimize seismic attributes processing parameters and reduce thedifference between the seismic lines at the intersection of seismic attributes bysimulated annealing algorithm, and achieve the purpose of improve the seismicattributes comparability between surveying lines. It’s an effective2D seismicattributes standardized method.Seismic attribute analysis shows that seismic attributes which response is moresensitive of reservoir change in the region include RMS amplitude, average energy,reflection strength, instantaneous frequency, the arc length and the effectivebandwidth. The RMS amplitude, average energy, reflection strength and the reservoirthickness is negatively correlated, while the effective frequency band and thereservoir thickness is a linear position correlation. The relationship between the length,frequency and the reservoir is relatively complex. Thick sandstone and thickmudstone seismic attributes often characterized by low-frequency and long arc length.The interbed (sand and shale), sand with medium thickness are characterized byhigh-frequency and short arc length.Seismic facies obtained through the unsupervised self-organizing neuralnetwork can reveal the distribution profile characteristic of the river, but thecharacterization and prediction accuracy of facies is lower(<55%); the pattern recognition method can build a reliable a relationship between the facies and seismicinformation, as a result in the higher prediction accuracy. Fuzzy neural network has agood knowledge representation and logical reasoning ability, also better for theseismic information fusion effect. This method can get higher accuracy in predictingthe results of sedimentary facies (83%).Through the comparison between the traditional interpolation method based onthe wells, geostatistic method with seismic attribute trend and fuzzy neural network, itshows that the error of the sand thickness from the traditional method is the largest(>5m), while geostatistic method reduces prediction error (4-5m). Fuzzy neuralnetwork method with well and seismic data gets the best prediction result and theminimum error (about3m).