Research On Seismic Prediction Method Of Total Organic Carbon Content In Shale Reservoir Based On Deep Neural Network

The total organic carbon(TOC)content of shale reservoir is an important parameter for evaluating the organic matter and hydrocarbon generation potential of shale reservoir,which is of great significance for the evaluation of oil and gas bearing capacity,reserve prediction,and development plan design of shale work area.Currently,the prediction research of TOC content of shale reservoir is mainly based on logging data.With the characteristics of continuous and high resolution in the vertical direction of logging data,the TOC content of reservoir in the vertical direction at the borehole location can be predicted.Utilizing the characteristics of wide lateral coverage,high resolution and spatial geometry of seismic data,the seismic data and logging data are combined can realize the spatial distribution prediction of total TOC content of shale reservoir,thereby extending the prediction result of TOC content of shale reservoir from borehole to adjacent formation space and realizing the detailed description of hydrocarbon generation potential of shale reservoir.With the rapid development of artificial intelligence technology,deep learning algorithms are widely used in the field of shale reservoir parameter prediction.Therefore,a vertical prediction model of TOC content in shale reservoir based on depth neural network logging data and a prediction model of spatial distribution of TOC content in shale reservoir based on depth neural network combining logging data and seismic data are proposed.The specific research content mainly includes the following two aspects:(1)Based on logging data,the depth neural network algorithms are used to establish the TOC content vertical distribution prediction model of shale reservoir.Firstly,the correlation analysis of logging curves and TOC content of shale reservoir is carried out to determine the TOC content sensitive logging curves of shale reservoir.Using the sensitive logging curves as a feature and the measured TOC content of shale reservoir as a label,the TOC content dataset of shale reservoir is constructed;Secondly,the TOC content prediction models for shale reservoir is constructed using Fully Connected Neural Network,Convolutional Neural Network and Long Short-Term Memory Network respectively;Finally,the performance of three types of prediction model for TOC content of shale reservoir are compared and analyzed.(2)Constrained by logging data,seismic data is used to invert the TOC content sensitive parameter data volumes of shale reservoir,and deep neural network algorithms are combined to achieve spatial distribution prediction of TOC content in shale reservoir.Firstly,based on the logging data,extract the well bypass seismic trace,compared the synthetic seismic record with the well bypass seismic trace to conduct time-depth conversion,make the seismic data in the time domain correspond to the logging data in the depth domain.Then,intercept the threedimensional seismic trace data of the target layer constrained by the borehole,and construct the seismic trace data volume for the TOC content of shale reservoir.Secondly,the seismic trace data volume of the target layer is used to inverse the TOC content sensitive parameters of the shale reservoir analyzed by logging data to obtain the acoustic transit time,density,p-wave velocity and wave impedance spatial data volume;Finally,based on the inversion of spatial data volume of sensitive parameters,the spatial data volume of TOC content in shale reservoir is calculated by using the Fully Connected Neural Network prediction model,Convolution Neural Network prediction model and Long Short-Term Memory Network prediction model trained with logging data in the combination of well and seismic data as the constraint.The spatial distribution of TOC content in shale reservoir bounded by borehole is obtained,and model comparison analysis is conducted.The research result shows that the Fully Connected Neural Network prediction model not only perform well in predicting the vertical distribution of TOC content in shale reservoir,but also is the optimal model in the prediction of the spatial distribution of TOC content in shale reservoir,and the prediction result matches the actual shale reservoir TOC content well,providing a more reliable evaluation of TOC content in reservoir in shale work area.