Study On Characterization Method Of Complex Fracture Network Of Deep Shale After Fracturing Based On Microseismic Monitoring

Influenced by bedding planes,pre-existing natural fractures,and complex stress conditions in deep shale reservoirs,the fracture network created by multi-stage fracturing is highly complex.Precise characterization of the fracture network after fracturing is the key to fracturing effects evaluation and the basis for fracturing optimization.Based on the computer vision characteristics of the microseismic events cloud,this paper introduces a comprehensive characterization model of the complex fracture network for shale fracturing by integrating the complex fracture network reconstruction,stimulated reservoir volume(SRV)calculation,and stimulated reservoir area(SRA)calculation.Firstly,aiming at solving the problem of insufficient anti-noise ability of current fracture reconstruction methods,this paper has put forward a 3D complex fracture reconstruction method with high anti-noise ability combined with the pattern recognition algorithms in computer vision.Based on the geometric depiction of the fracture network in laboratory true triaxial physical simulation fracturing experiments,the random convex polygon is used to build the geometric model of the single fracture of the fracture network.After using the RANSAC method for fracture occurrence recognition,the Alpha-shape method for fracture shape detection,and the DBSCAN method for fracture denoising,the3 D fracture reconstruction model has been successfully established based on microseismic events.Besides,the study also proposed a synthetic events generation method and a model quantitative verification process by utilizing the Monte Carlo simulation and the Gauss mixture model.Through a large number of simulations,it is found that this model could effectively reconstruct the fracture network with arbitrary shape and arbitrary occurrence for microseismic events with an arbitrary distribution.The reconstruction error can be stable below 20% under 50% noise,which proves its good anti-noise ability and excellent reconstruction performance.Secondly,aiming at solving the problem that the current SRV envelopes contain numerous error regions,this paper has established the SRV calculation model based on the voxelized grid of the microseismic events cloud.Found on the in-depth analysis of the SRV envelope construction mechanism,the existing SRV calculation models are effectively classified.Enlighted by the point cloud voxelized methods in computer vision,new fine SRV computing models have been proposed by using uniform voxel grid and non-uniform voxel grid.Through the simulated SRV calculations,it is indicated that the uniform grid SRV model and non-uniform grid SRV model could effectively eliminate the blank area and the low-density area within the grid structures.They obviously outperform the conventional SRV envelope mode in accuracy,adaptability,and antinoise ability for complex fracture networks with different configurations.Besides,by further comparing the uniform grid SRV model and the non-uniform grid model,it is found that the non-uniform grid model has greater advantages than the uniform grid SRV model at the same grid resolution.Finally,aiming at solving the problem that the calculation error is too large due to the indirect calculation of SRA after fracture reconstruction,a direct SRA calculation method using microseismic events is proposed based on the octree decomposition algorithm in computer vision.After extracting the 3D structure of the microseismic point cloud through a multi-level octree grid structure,the equivalent area of the complex fracture network is obtained by calculating the characteristic area of the equivalent grid structures.Thus,the accurate SRA of the complex fracture network can be successfully achieved without the reconstruction of the fracture network in conventional methods.By a large number of simulations,the average calculation error for the middle-scale fracture network can be less than 20%.It is found that the SRA direct estimation model could effectively adapt the area estimation of the complex fracture network of different configurations with high accuracy and good robustness,which demonstrates its high application value for actual multi-stage fracturing.