Seismic lithology identification method based on integrated deep learning

Abstract

The invention discloses a seismic lithology prediction method based on integrated deep learning, and belongs to the technical field of geophysical exploration for petroleum, and the method comprises the steps: obtaining lithology data and well bypass seismic data corresponding to a target horizon according to the calibrated target horizon; based on the well-side seismic channel data and the ground lithology data, taking the well-side seismic channel data as input data, taking the ground lithology data as target data, and establishing a seismic lithology identification integrated deep learning model; and inputting actual seismic data of a to-be-predicted area based on the seismic lithology identification integrated deep learning model to obtain predicted lithology. According to the method, the weak seismic response characteristics between the seismic data and the lithology data are extracted by establishing the seismic lithology identification integrated deep learning model, and the lithology data volume of reservoir distribution can be determined more simply and efficiently. According to the scheme, the problem of inter-well reservoir prediction is solved, and a reference basis is provided for oil-gas exploration and development.

Description

technical field [0001] The invention relates to the technical field of petroleum geophysical exploration, in particular to a seismic lithology identification method based on integrated deep learning. Background technique [0002] Lithofacies information can reflect reservoir lithology and fluid characteristics, and plays an important role in seismic reservoir prediction. Conventional methods mainly use elastic parameters that are closely related to lithofacies information to qualitatively or quantitatively interpret lithofacies information. In practical applications, the acquisition of elastic parameters is mainly based on the pre-stack seismic inversion technology, but the pre-stack seismic inversion method is difficult to avoid multi-solution and instability. In recent years, the lithofacies inversion method based on Bayesian theory has also begun to be used in lithofacies prediction as an automatic interpretation method based on pre-stack elastic parameters. The inversi...

AI Technical Summary

Problems solved by technology

In order to fully excavate the effective information contained in a large amount of seismic data, the present invention innovatively uses The combination of convolutional neural network (CNN) and long short-term memory (LSTM) neural network is introduced into the field of geological exploration, and a seismic lithology identification method based on the deep fusion network of CNN and LSTM network is proposed, which can effectively solve the problem of lithology identification and classification and improve recognition accuracy

Images