A superimposed velocity spectrum picking method and processing terminal based on deep reinforcement learning

Abstract

The invention relates to a stacking velocity spectrum pickup method based on deep reinforcement learning and a processing terminal. The method comprises the following steps of: S1, acquiring originalcommon midpoint gathers data including seismic reflection waves, and calculating a stacking velocity spectrum composed of optimal scanning speeds at various moments; S2: inputting the stacking velocity spectrum into an auto-encoding network to obtain coded high-order energy group features; S3: inputting the codes of the high-order energy group features into a policy network, picking up the optimalscanning speeds at various moments, and outputting a speed sequence; S4: evaluating the speed sequence and outputting reward values; S5: training the policy network according to the reward values; and S6: executing theS3 to the S5 iteratively until the set maximum reward value is acquired in the S4, and then outputting the optimal speed sequence. According to the stacking velocity spectrum pickupmethod based on deep reinforcement learning and the processing terminal, intelligent velocity spectrum pickup is realized without manual intervention, the interference of multiple waves can be eliminated during the pickup process, and the obtained stacking velocity curve is more accurate.

Description

technical field [0001] The invention relates to the technical field of seismic data processing, in particular to a stacking velocity spectrum picking method based on deep reinforcement learning and a processing terminal. Background technique [0002] Seismic wave velocity is one of the important parameters in seismic data processing and seismic imaging, especially multiple wave suppression based on velocity difference and pre-stack time (or depth) migration based on wave equation theory are more reasonable for the velocity model The accuracy directly affects the processing results and the quality of seismic imaging. Therefore, in the process of seismic data processing, it is necessary to obtain the velocity model that is closest to the actual situation as much as possible. [0003] At present, the method of manual velocity spectrum picking is mainly used. The artificial method is not only affected by human factors, but the result of velocity spectrum picking depends to a lar...

AI Technical Summary

Problems solved by technology

[0003] At present, the method of manual velocity spectrum picking is mainly used. The artificial method is not only affected by human factors, but the result of velocity spectrum picking depends to a large extent on uncontrollable factors such as human experience and knowledge, and the manual method leads to low efficiency; and Using other methods, such as the two patents with publication numbers "CN105445788A" and "CN105572733A", the operation is complicated, the interference wave is greatly affected, the picking accuracy is not high, and because a large amount of manual intervention is required in the application process , did not realize the intelligentization of velocity spectrum picking, and its efficiency has not been improved qualitatively. It is only used as an auxiliary tool in actual work

[0004] In the velocity spectrum of actual seismic data, the focus of the effective velocity spectrum energy group is poor, and the existing velocity spectrum picking methods are based on the high recognition of the velocity spectrum energy group, so the existing velocity spectrum picking method is difficult Meet the requirements of practical applications;

[0005] The current velocity spectrum picking method more or less still needs manual intervention, and the operation is complicated, and it is not intelligent;

[0006] For marine seismic data, it is seriously interfered by multiple waves. Multiple waves are a kind of coherent interference received by geophones after multiple reflections between the sea surface and the geological interface. Multiple waves seriously interfere or even shield effective reflections, making effective wave energy clusters in the velocity spectrum unfocused and difficult to identify and track. The existing velocity spectrum picking methods are almost powerless to the impact of multiple waves

Images