Research And Application On Improving Seismic Data Resolution Based On Generative Adversarial Network

As the potential targets of oil and gas exploration become more and more complex,the search for complex geological anomalies(such as thin beds,thininter beds,etc.)has become an important task of exploration and development of oil and gas resources.The accuracy of oil and gas exploration is closely related to the resolution of seismic data.The higher the resolution of seismic data,the more benefit to understand the geological information of the area,which can accurately guide the exploration and development of oil and gas,reduce exploration risk,and produce significant social and economic benefits.However,with the deepening of seismic exploration,the information loss in the high frequency band of seismic waves is increasingly aggravated,which makes the resolution of seismic data received by geophone worse and worse,and it is difficult to meet the needs of exploration and development of oil and gas resources.To sum up,improving the resolution of seismic data collected has become an important task in the field of seismic exploration.After describing the relevant background and research status of improving the resolution of seismic data,aiming at the characteristics of seismic data,this paper proposes a method of improving the resolution of seismic data based on conditional deep convolution generation Adversarial network(C-DCGAN).The model of C-DCGAN adds convolutional layer and conditional information on the basis of GAN.The main work and achievements of this paper are as follows:(1)In order to improve the resolution of seismic data,conditional deep convolution generation adversarial network is used in this paper.The overall architecture of the network is as follows: deconvolution neural network is applied in the generative model of conditional deep convolution generating adversarial network,and convolutional neural network is applied in the discriminant model of conditional deep convolution generating adversarial network.The binary cross entropy Loss function(BCE Loss)is used to conditional deep convolution generation adversarial network.(2)In order to prove the effectiveness of the conditional deep convolution generation adversarial network used in this paper in improving the resolution of seismic data,the network used in this paper is applied to the construction model with different complexity.It can be seen from the results that the proposed method can improve the resolution of seismic data,but there are some short comings in the testing process,such as discontinuity of phase axis,energy imbalance,false in-phase axis,etc.(3)Experiment on high resolution processing of actual seismic data.The conditional deep convolution generation adversarial network model is used to process the actual seismic data at high resolution.On the whole,the network used in this paper can effectively recover the high frequency information of the seismic data and improve the resolution,but there is a problem of the loss of low frequency components.