Practical internal substratum automatic interpretation method

Abstract

The invention provides a practical internal substratum automatic interpretation method. The practical internal substratum automatic interpretation method comprises the first step of conducting single-well fine sand group division, the second step of conducting fine synthetic record calibration, the third step of calculating an index bed controlled residual value through the fine synthetic record calibration, the fourth step of conducting nonlinear trend surface fitting on the residual value of each single well through a support vector machine method to obtain the residual position of sand groups and the fifth step of conducting automatic interpretation of the sand groups. According to the practical internal substratum automatic interpretation method, the nonlinear trend surface fitting of a support vector machine controlled by an index bed is used for automatically tracking a substratum, by means of the substratum interpreted through the method, the result is more objective and accurate, the working efficiency is effectively improved, and the interpretation error is reduced.

Description

technical field [0001] The invention relates to the technical field of oilfield development, in particular to a practical automatic interpretation method for insider sublayers. Background technique [0002] At present, exploration has entered the stage of high exploration. For areas with high exploration degree, there are many subdivided sand groups in a single well. Whether the reservoir prediction results are reliable or not requires accurate sand layer positions. Therefore, the control layers of sand groups The location is the key to reservoir prediction. It is a huge project to interpret the horizon of the sand layer group. The interpretation of the sand layer group is becoming more and more important. There are usually three methods for the interpretation of the sand layer group. (1) Manual interpretation: due to the sand layer There are many formations, the continuity of the seismic axis is poor, the tracking and interpretation is difficult, the efficiency is low, and ...

AI Technical Summary

Problems solved by technology

Both methods can get a good regression effect when there are many learning samples, but the generalization ability of the support vector machine is better than that of the neural network. With the reduction of the number of samples, in the case of small samples, the shortcomings of the neural network regression show. It is very obvious that the generalization ability of the neural network is very poor, while the support vector machine still shows a good generalization ability. It can be seen that because the neural network is committed to minimizing the sample error, it is easy to produce overfitting during regression. The combination phenomenon affects the generalization ability

Images