Machine Learning Based On Well Logging Image To Recognize Fracture And Predict Fracture Porosity

With the rapid development of the world economy,the demand for energy is also increasing.Conventional oil and gas fields tend to be exhausted,and the oil exploration industry must march towards complex and tight oil and gas reservoirs.As an important part of domestic petroleum resources,igneous rock reservoir has become a hot research topic because of its complex composition and structure and difficult evaluation.The energy storage of igneous reservoir is closely related to fractures,which can be used as a fluid channel or as a reservoir space for oil and gas.Accurate identification of fractures and prediction of fracture porosity are the key to the evaluation of fractured reservoirs in igneous rocks.In order to solve the problem of evaluating fractured reservoir of igneous rock,this paper takes the volcanic rock of Nanpu 5 structure in Huanghua depression of Eastern Hebei Province as the research object,studies the automatic recognition of fracture and the prediction of fracture porosity by computer.The fracture porosity of volcanic reservoir is closely related to lithology,fracture density and opening,so it is necessary to identify lithology first.In this paper,K-nearest neighbor(KNN)method is used to divide the lithology of volcanic rocks.Compared with BP neural network and support vector machine(SVM)method,KNN method has high accuracy in dividing the lithology of volcanic rocks,and the accuracy of 97 samples is 90%.In the aspect of automatic fracture identification,the Formation MicroScanner Image(FMI)data is taken as the breakthrough point.The high vertical resolution of FMI data contains more abundant formation information,but the traditional human-computer interaction identification of fractures has a large workload and low efficiency.In order to solve this problem,the Conditional generative adversarial nets(CGAN)method is proposed to identify the fractures in the electrical imaging logging image.In the process of fractures recognition,CGAN has the advantages of fast recognition speed and strong anti-interference ability.The accuracy of identifying horizontal and low angle fractures in the image of Nanpu 5 structural volcanic rock section is 90%.In the aspect of fracture porosity prediction,the convolution neural network(CNN)is used for regression prediction of fracture porosity.It is predicted that the average relative error of fracture porosity is 1.247%,among which the minimum is 0.025%.This method can automatically identify the fractures in the FMI image by computer,which greatly saves the time of human-computer interaction to extract the fracture information in FMI.Quantitative evaluation of fracture porosity paves the way for quantitative evaluation of energy storage and productivity of fractured volcanic reservoir.In the future,the treatment method in this paper can be generalized and improved and applied to the evaluation of other fracture type complex oil and gas reservoirs.