Research On Seismic Facies Identification System Based On Convolution Neural Network

At present,fossil fuels such as oil and natural gas occupy a dominant position in the global energy structure,and seismic exploration technology can be used to explore oil and gas resources.With the continuous progress of seismic exploration technology,seismic data can more and more reflect a large number of underground geological information,and bring into play an significant role in the exploration and exploitation of petroleum,seismic interpretation is particularly important in seismic exploration.Seismic facies analysis is an important component of seismic data interpretation,which plays a crucial role in understanding the distribution of underground sedimentary facies in the work area and can improve the accuracy of oil and gas exploration.However,with the rapid development of seismic data acquisition and processing technology,the traditional "phase plane method" seismic phase analysis has been unable to accurately identify the rich information in seismic data,and there are teaser of inefficiency and low precision.Therefore,it is necessary to rely on the excellent computing ability of the computer to process and interpret seismic data.With the continuous improvement of computer performance in recent years,the artificial intelligence algorithm based on convolution neural network has been developed rapidly and has been widely used in seismic interpretation fields such as seismic facies analysis.This article delves into the study of seismic facies recognition based on intelligent methods.The main task includes the following three ways.First,a convolution neural network seismic facies recognition method based on efficient attention mechanism is proposed.In order to identify seismic facies at pixel level on seismic section,An intelligent seismic facies recognition model based on "encoding-decoding" network structure was studied.Although many improved semantic segmentation models can improve the recognition accuracy,they also increase a lot of computing time costs.In view of this,by combining the Effificient Channel Attention(ECA)mechanism with the Unet++network model and increasing the use of the original Unet++ network for different channel feature information,the recognition effect of the network can be effectively improved while adding a small number of model parameters.The model is trained by using the weighted composite loss function of the multivariate cross-entropy loss function and the logarithmic Dice loss function,which improves the problem that the model prediction results are not ideal due to the imbalance of data samples.On the Dutch F3 seismic facies data set,the comparison experiment with the U-Net,U-Net+PPM and Unet++network models,the ECA-Unet++model proposed in this paper increases the training time cost and recognition accuracy effectively.Second,carry out seismic facies identification based on seismic attributes.In order to gain a more comprehensive understanding of the sedimentary facies distribution in the work area and provide a more comprehensive theoretical basis for the subsequent implementation of seismic facies identification systems,seismic facies identification work based on seismic attributes has been carried out.Multi-seismic attributes have redundant information and increase unnecessary computing resource consumption during analysis.Therefore,the data feature dimensionality reduction method of principal component analysis(PCA)and independent component analysis(ICA)is used to optimize the seismic attributes,and the clustering analysis is executed.The outcome make known that the clustering analysis effect of the attribute optimization method based on ICA is better on the Dutch F3 seismic data.The application of attribute optimization can improve the efficiency of seismic facies analysis and further understand the lithology distribution and reservoir distribution of the study area.Thirdly,the design and implementation of an intelligent seismic phase recognition system.The system designed and developed can perform corresponding seismic attribute processing and seismic phase recognition functions,and the application of this seismic phase recognition system is practical.