Research On Intelligent-Sensing-Based Distributed Seismic Data Acquisition Method

Seismic exploration has become the most important method for oil and gas exploration because of its deep detection capability.As the demands of oil and gas have increased in recent years,large-scale seismic data acquisition has also become a hot issue in both academic and industrial research.Due to the complex application environment and huge cost of seismic data acquisition,it is difficult to rely on hardware to drive seismic data acquisition to a larger scale.Therefore,improving the efficiency of existing hardware resources to meet the demands of larger-scale seismic data acquisition has become a new research idea.Due to the diversity of nonlinear characteristics of seismic data,current methods for efficient acquisition of seismic data have shortcomings in terms of adaptability and feature extraction.Aiming at these shortcomings of the existing methods,this thesis proposes an intelligent-sensing acquisition method combined with compressed sensing,sparse representation and machine learning.According to common seismic data acquisition scenarios(seismic exploration and microseismic monitoring),two different distributed intelligent-sensing acquisition methods are designed,which are seismic intelligent-sensing acquisition method based on multi-hop network and intelligentsensing acquisition method based on microseismic event detection.The main research contexts of the thesis are detailed as follows:(1)The relevant theories of seismic data efficient acquisition methods,such as compressed sensing,sparse representation and machine learning,are studied.Then,an intelligent-sensing acquisition method is proposed via combining compressed sensing,sparse representation and machine learning.Specifically,the intelligent-sensing acquisition method employs compressed sensing theory as the basic framework.Sparse representation and machine learning algorithms are introduced to the method,for enhancing the ability to extract and fit nonlinear features based on a data-driven way under the sparsity constraint.The proposed method not only ensures strong adaptability,but also explores sparseness characteristics in sparse land and nonlinear features of input data.After proposing the basic idea,the theoretical support of the intelligent sensing acquisition method is discussed with the combination of three theories,which provides a theoretical basis for the study of specific application scenarios.(2)Focusing on the characteristics of seismic data and network structure of acquisition systems,the overall scheme of intelligent-sensing-based distributed seismic data acquisition method is designed.By analyzing the sparse land of seismic data from seismic wave equations and information entropy,seismic data is certified to satisfy the requirements of efficient acquisition methods.Based on two main application scenarios: seismic exploration and microseismic monitoring,the structures of two acquisition systems are analyzed.According to the basic idea of the intelligent-sensing acquisition,characteristics of data and the network structure of acquisition system,two different kinds methods,namely seismic intelligent-sensing scheme based on multi-hop network and intelligent-sensing scheme based on microseismic event detection,are proposed.(3)According to seismic intelligent-sensing scheme based on multi-hop network and the characteristics of conventional seismic exploration data acquisition,seismic intelligent-sensing acquisition method based on multi-hop network is proposed.Aiming at the unbalanced problem of multi-hop data transmission,an encoding framework based on multi-hop network is designed.This key technology combines the multi-hop network structure with the encoding process of compressed sensing,which realizes a balanced transmission and reduces total transmission volume.Aiming at the problem of measurement matrix optimization and data reconstruction,a compressed sensing algorithm based on generative adversarial networks is proposed.This key technology uses the compressed sensing framework as the constraint condition of generative adversarial networks,as well as the adversarial mechanism and convolutional neural network are used to extract detail features for seismic data reconstruction.Based on the numerical test results,the seismic intelligent-sensing acquisition method based on multi-hop network can increase the channel capacity by 16 times with only little noise,whose power is 0.001 of the signal's power.Moreover,the energy consumption of data transmission in the entire line is reduced to 0.125 of the original.(4)According to intelligent-sensing scheme based on microseismic event detection and the characteristics of microseismic monitoring,intelligent-sensing acquisition method based on microseismic event detection is proposed.Aiming at the phenomenon that a large amount of irrelevant data is transmitted to data center during microseismic monitoring,a compressed sampling technique based on microseismic event detection is proposed.This key technology combines machine-learning-based classification algorithms with compressed sampling,which can detect microseismic events with the under-sampling rate and only transmits data including microseismic events.Aiming at the problem of sparse representation of microseismic data,a clustering dictionary learning algorithm based on singular value decomposition is proposed.This key technology combines sparse representation with machine learning to search the sparse land of microseismic data in an adaptive way.Then,SPGL1 algorithm is used to recover the original microseismic signal via using the undersampling data and learned sparse land.According to the numerical test results,the intelligent-sensing acquisition method based on microseismic event detection can reduce the energy consumption of data recording by 30% when the power of introducing noise is the 0.01 of original signals.It can significantly reduce the irrelevant transmission data(usually more than 90%)and achieve a high accuracy of microseismic event detection in low SNR cases(accuracy rate of 96.83% in the case of-15dB).In summary,the intelligent-sensing-based distributed seismic data acquisition method proposed by this thesis can greatly improve the bandwidth and energy efficiency of the acquisition system.In other words,it is possible to accommodate more acquisition nodes under same hardware conditions,which means a larger seismic data acquisition.Besides,the proposed method provides a new idea for the development of large-scale seismic data acquisition system,as well as provides a theoretical basis for the research on intelligence of deep oil and gas exploration equipment.