The Research Of Reservoir Prediction Method Based On Multiple Attributes Analysis

Hidden reservoir has three features: complex distribution, difficult exploration, andtechnical demanding. Therefore its exploration and development need the high demand ofreservoir prediction accuracy, and the original technical means can’t meet people’s needs.The seismic attributes contain a lot of underground strata information, which can betterreflect the lithology, physical properties and oil and gas characteristics. Using a single seismicattribute often makes reservoir prediction have multiple solutions and uncertainty. Thedifferent combinations of seismic attributes make the results of reservoir prediction different.Seismic attribute analysis is critical to the use of seismic attribute in the course of the study ofthe hidden reservoir.In order to improve the effectiveness and accuracy of the reservoir prediction, this paperputs forward a combination of expertise and mathematical theory. It is preferable to select themost favorable and optimal seismic attributes to reservoir prediction. Through the K-Ltransform, we can reduce the dimension mapping to the optimized seismic attributes in orderto eliminate the unrelated and redundant attributes. And we can reduce the space dimensionthat attributes need to build the reservoir prediction and improve the speed and accuracy ofreservoir prediction. Then we analyze the validity using the sensitivity of seismic attributesand reservoir parameters. We use the improved BP neural network to create a goodrelationship between a variety of selected attributes and the reservoir parameters to predictreservoir, greatly improving the accuracy of reservoir prediction. On the basis of thetheoretical studies, the article also describes the software implementation process of areservoir prediction method based on the seismic multi-attribute analysis. We carry out the thickness prediction of reservoir to the biological limestone and dolomite lawyer in yangxinarea and achieve a good result.