Inversion Of Underground Anomalous Bodies Based On Deep Learning

The purpose of this paper is to study the inversion method of underground anomalous bodies based on deep learning,and take the identification of hydraulic fracturing fractures as the specific application scenario.Hydraulic fracturing is a common unconventional natural gas extraction technology,but its operation process may affect the underground geologicaal environment,such as the formation and propagation of fractures,resulting in the formation of underground abnormal bodies.Therefore,the accurate identification of hydraulic fracturing fractures is of great significance to the evaluation and management of underground geological environment.However,the traditional method of crack identification is to study the length and dip Angle of cracks by simulation.However,simulation is very time-consuming and cannot be accurately realized in large-scale application scenarios.In view of the current research status of fractures and the superior performance of deep learning in different fields,this paper proposes an intelligent inversion solution based on deep learning,adopts theoretical analysis,numerical calculation and other methods to carry out research,and designs a detection system based on B/S architecture to provide an efficient and accurate solution for the identification of hydraulic fracturing fractures.At the same time,it provides some reference for related research fields.In this paper,the finite element method is used for numerical simulation,and the modeling method based on the open source framework FEni CS is used to construct the single fracture model,the double fracture model and the corresponding apparent resistivity profile sample library.In order to improve the performance of the deep learning model,an improved ResAttention Unet network is proposed in this paper,which combines the Res Block residual module and the Attention mechanism in Res Net to extract more information features of underground abnormal bodies,enhance the robustness of the model and prevent the occurrence of underfitting.In the coding and decoding network of the traditional U-Net network,Res Block module is introduced to dig the primary features more deeply,and the Attention module is used to select and connect the primary features and advanced features more specifically,thus effectively improving the inversion accuracy and stability.After training,the qualitative analysis and quantitative comparison show that the convolutional neural network based on deep learning can accurately invert the position and physical properties of the two-dimensional model anomalous body.On the verification set,all the segmentation indexes tend to be good,and the accuracy can reach 99.7%,and the intersection ratio is greater than or equal to 0.5.In this paper,based on the full convolutional neural network,the recognition,detection and segmentation methods of hydraulic fracturing fracture images are effective,and the hydraulic fracturing fracture detection system is constructed.The research in this paper helps users realize more efficient,accurate and safe hydraulic fracture detection,which is of great significance for improving the economic benefits of natural gas resources in oil Wells.