Study On Discrete Fracture Network Modeling Method Based On Random Polygon Model

Three-dimensional geological model of fractures is an important geological basis for tight reservoir developments.To solve eight problems existing in current fracture modeling methods for tight reservoirs of areas without 3D seismic data,studies about constraint conditions determination method for 3D geological modeling,discrete fracture network modeling method based on random polygon model and validity testing method for fracture network models are carried out.Chang 6 reservoir of Huaqing area in Ordos Basin is taken as an example to test the proposed methods.The 3D fracture network model of the study area is established.The constraints for fracture modeling can be obtained by combining hard controls with high vertical resolution and soft controls with high lateral resolution.To handle the problems of few cores and imaging logs and ambiguity of fractures using conventional logs,a fracture identification method(Lap SVM)of semi-supervised learning based on kernel function is proposed.The kernel function maps the original classification problem of low-dimensional inseparability into a high-dimensional nonlinear space,and then a prediction model for fractures is established by semi-supervised learning.This method can effectively utilize a large amount of unlabeled logging data and improve the fracture identification accuracy using conventional logs.According to distribution features and patterns of fractures in tight reservoirs,a 3D fracture network modeling method(Fnm RPF)for multi-scale fracture system based on random polygon fusion is proposed for areas without 3D seismic data.It realizes the organic fusion and hierarchical modeling of multi-scale fractures based on the growing idea from small to large.By the improved methods of point process and mark process,fracture elements can be generated orderly and efficiently under constraint conditions,and the computational speed can be increased by more than 30 times,which makes the simulated 3D fracture network characteristics more consistent with the actual situation.Furthermore,the fracture network model is optimized by the iterative inversion of seed hard data and revised hard data.After the fracture modeling,the validation of the fracture network using dynamic data suffers from the problems of low efficiency and difficulty in quantification.To solve these problems,a fracture network model validation(FME)method based on the fracture intersection detection method(BBSL)and fracture network connectivity index(FCI)is proposed.The method applies bounding box and sweeping line to detect fracture intersections,which makes connected fracture clusters in the fracture network model quickly detected and changes the detection time from exponential growth to linear growth.The detection efficiency increases by order of magnitude.By calculating the fracture connectivity index of the fracture network based on dynamic data,the effectiveness of fracture network is quantitatively evaluated.In the three-dimensional fracture model of Chang 6 tight reservoir in Huaqing area,the coincidence rate of logging fracture interpretation increased by more than 10%,and the coincidence rate of connectivity test of 3D fracture network model was high,indicating a better application.