Seepage Characteristics And Laws Of Rock Fracture Network Based On Data Mining

China has abundant shale oil and gas resources,but it is difficult to exploit because of its deep burial and low reservoir permeability.As an important method for shale oil and gas production,reservoir fracturing technology can improve the distribution of reservoir fracture network and enhance the seepage ability of reservoir fracture network.Therefore,the study of the relationship between the characteristic parameters of rock fracture network and the seepage coefficient is helpful to guide the fracturing construction design of horizontal Wells for shale oil and gas and improve the oil and gas production efficiency.This thesis is based on the shale oil and gas reservoir percolation network problems,analysing the relation between seepage and fractured rock mass characterization,specific process is as follows: first,design the fractured rock mass seepage experiment on simulator,precast block 120 pieces of different fracture scale,different crack simulation and the complexity of seepage in fractured rock mass,the data foundation for the study of fracture seepage law.Secondly,electrochemical detection method was used to analyze the relationship between electrochemical characterization and seepage flow of fractured rock mass at the micro level.Thirdly,combined with image processing,neural network and digital-analog transformation algorithm,the digital-analog transformation software of fracture network was designed to excavate the characteristic parameters of fracture network in fractured rock mass.Finally,according to the characteristic parameters of fracture network and the characteristics of fracture network image,a seepage prediction model of fractured rock mass was designed and trained,and the relationship between the characteristic parameters of fracture network and the seepage coefficient was further analyzed by the neural network model.The research results are as follows :(1)Through the laboratory simulation experiment,it is found that the larger variation range of electrochemical impedance value,the smaller seepage coefficient,and the mean value of total impedance in electrochemical high frequency band is positively correlated with the seepage coefficient.(2)The split network feature extraction method integrates the machine vision processing algorithm and the neural network recognition algorithm,changes the threshold value of the split network feature extraction method from the whole to the local one,which effectively improves the problems of the split network image misidentification,crack line breakage,many noise points and so on.(3)The image processing algorithm,YOLO V4 neural network model,crack parameter calculation method and other algorithms were integrated to form the crack network digital analog conversion software suitable for this thesis.(4)By analyzing the relationship between the characteristic parameters of the fracture network and the seepage coefficient after the digital model transformation of the fracture network image,it is found that the seepage coefficient and the maximum fracture opening have an obvious rule,showing a positive correlation.When the seepage coefficient of the test block is small and the maximum fracture opening is small,the average fracture opening is also positively correlated with the seepage coefficient.(5)through independent design training of fractured rock mass seepage prediction model parameters and analyzing the characteristic of the fissure network seepage,found the biggest opening fractured rock mass crack network on its seepage coefficient of weight,the largest crack influence on seepage coefficient of weight minimum net area,with the fissure network seepage coefficient of image fractal dimension increases.(6)Through training the mixed input neural network prediction model of fractured rock mass,it is found that the neural network with fracture network image of fractured rock mass has a better seepage prediction effect.The research results in this thesis can provide reference for the evaluation and prediction of rock hydraulic fracturing effect,and provide basis for the construction design of horizontal Wells with hydraulic fracturing.