Search

CN-121995456-A - Free surface multiple wave pressing method and storage device

CN121995456ACN 121995456 ACN121995456 ACN 121995456ACN-121995456-A

Abstract

The invention provides a free surface multiple suppression method and a storage device, and belongs to the field of seismic exploration. The method comprises the steps of 1, extracting header words of seismic data, carrying out statistical calculation, 2, obtaining global data by using a distributed file and converting the global data into an elastic distributed data set RDD, 3, carrying out slicing and parallel operation on the seismic data in the elastic distributed data set RDD, and 4, merging operation results of all the slicing again. The invention plays the advantages of Spark in the aspects of resource management, fault tolerance and the like, supports more flexible and convenient parallel granularity, and enables the practicability of the free surface multiple technology to be more controllable, stable and efficient.

Inventors

  • WANG CHAOYANG
  • YANG XIANGSEN
  • XIE JINE
  • GUO QINGHUA
  • WANG MINGQIU

Assignees

  • 中国石油化工股份有限公司
  • 中石化石油物探技术研究院有限公司

Dates

Publication Date
20260508
Application Date
20241107

Claims (10)

  1. 1. A free surface multiple pressing method is characterized by comprising the following steps: step 1, extracting a header word of seismic data, and carrying out statistical calculation; step 2, acquiring global data by using a distributed file and converting the global data into an elastic distributed data set RDD; Step 3, slicing and parallel operation are carried out on the seismic data in the elastic distributed data set RDD; and 4, merging the operation results of all the fragments again.
  2. 2. The method of claim 1, wherein the heading words include position information of seismic data acquisition, namely shot point, detection point coordinates and offset information.
  3. 3. The method for pressing multiple waves on the free surface according to claim 1, wherein the statistics of the step 1 are calculated as statistics of the properties of the road head, including the minimum number cdp, the maximum number cdp, the minimum number and the maximum number of the input data, and the properties of the road head are stored in a specified one-dimensional array after the statistics.
  4. 4. The method of pressing multiple on a free surface according to claim 1, wherein the step 2 comprises: Step 21, loading the seismic data to be tested into Spark, and storing the seismic data in a distributed file system (HDFS) in a distributed mode; Step 22, converting the global data into a resilient distributed data set RDD of Spark, depending on the API in Spark.
  5. 5. The method of claim 1 or 4, wherein the global data is raw seismic data for the entire work area.
  6. 6. The method of claim 1, wherein the step 3 is performed in parallel according to the trace set, and comprises: step 31, slicing the seismic data of the multiple to be predicted according to the trace set; step 32, independently calculating the inside of each partition according to the channel; And 33, carrying out convolution and superposition operation on the calculation results of all the fragments.
  7. 7. The method of claim 6, wherein the step 32 searches contribution gathers in each gather according to the header attribute counted in the step 1, calculates multi-contribution gathers of each gather, and reads in the multi-contribution gathers of each gather for each gather of the multi-gather to be predicted after slicing in the step 31.
  8. 8. The method of claim 6, wherein the step 33 is performed for pairs of the multiple-contributing gathers, wherein the two gathers included in each pair are convolved, and wherein the results of the convolutions of each pair are all added.
  9. 9. The method of claim 1, wherein the step 4 uses a reduce operation to aggregate the results of all the slices.
  10. 10. A computer-readable storage medium, characterized in that the computer-readable storage medium stores at least one program executable by a computer, which when executed by the computer, causes the computer to perform the steps of the free surface multiple pressing method according to any one of claims 1-9.

Description

Free surface multiple wave pressing method and storage device Technical Field The invention belongs to the field of seismic exploration, and particularly relates to a free surface multiple suppression method and a storage device. Background The free surface multiple suppression algorithm often needs to search multiple contribution gathers in the process of predicting multiple, load data and finally calculate multiple. At present, MPI parallel mode is used to process different data through different processes, and a plurality of threads are used for calculation in each process. The reading and circulation of data depends on the disk file. The existing free surface multiple pressing technology depends on the parallel mode of MPI, and although MPI plays a great role in high-performance calculation, as the calculation amount of data increases, how to perform necessary resource management, operation monitoring, fault tolerance and the like in the execution process of the free surface multiple program becomes a bottleneck problem which is difficult to solve. Disclosure of Invention Aiming at solving the problems in the prior art, the invention provides a free surface multiple pressing method, which is based on Spark parallel frames, uses elastic distributed data set RDD to conduct slicing parallel on data, enables each slice to conduct multiple model prediction of each path in the partition independently, and then merges the multiple models of each slice so as to generate multiple prediction models of data to be predicted. The invention is realized by the following technical scheme: In a first aspect of the present invention, there is provided a free surface multiple pressing method, the method comprising: step 1, extracting a header word of seismic data, and carrying out statistical calculation; step 2, acquiring global data by using a distributed file and converting the global data into an elastic distributed data set RDD; Step 3, slicing and parallel operation are carried out on the seismic data in the elastic distributed data set RDD; and 4, merging the operation results of all the fragments again. Further, the header words comprise position information of seismic data acquisition, namely shot points, detection point coordinates and offset information. Further, the statistics in the step 1 are calculated as statistics of the track head attributes, wherein the statistics comprises the minimum number cdp and the maximum number cdp of input data, the minimum number and the maximum number of cannons, and the track head attributes are stored in a specified one-dimensional array after statistics. Further, the step 2 includes: Step 21, loading the seismic data to be tested into Spark, and storing the seismic data in a distributed file system (HDFS) in a distributed mode; Step 22, converting the global data into a resilient distributed data set RDD of Spark, depending on the API in Spark. Further, the global data is raw seismic data for the entire work area. Further, the step 3 is executed in parallel according to the gather, and includes: step 31, slicing the seismic data of the multiple to be predicted according to the trace set; step 32, independently calculating the inside of each partition according to the channel; And 33, carrying out convolution and superposition operation on the calculation results of all the fragments. Further, in the step 32, contribution gathers are searched in each gather according to the header attribute counted in the step 1, multiple contribution gathers of each gather are calculated, and for each gather of the multiple gathers to be predicted after the slicing in the step 31, the multiple contribution gather of each gather is read in. Further, the step 33 is to generate multiple contribution trace integration pairs, to convolve the two trace sets contained in each pair, and to add the results of the convolution of each pair of multiple contribution trace sets. Further, the step 4 uses a reduce operation to aggregate the calculation results of all the fragments. In a second aspect of the invention a computer readable storage medium is provided, characterized in that the computer readable storage medium stores at least one program executable by a computer, which at least one program, when executed by the computer, causes the computer to perform the steps of the free surface multiple pressing method according to any one of claims 1-9. Compared with the prior art, the method has the beneficial effects that the free surface multiple pressing method based on the Spark parallel frame is provided, the advantages of Spark in the aspects of resource management, fault tolerance and the like are exerted, and the parallel granularity which is more flexible and convenient is supported by the method by combining the method, so that the practicability of the free surface multiple technology is more controllable, stable and efficient. Drawings Figure 1 is a schematic diagram of the data flow of the me