Research And Application Of Quantitative Fracture Prediction Based On AVAZ Frequency Variation Characteristics

Fractured reservoirs are an important part of the global hydrocarbon resource base.This reservoir type has a complex geometry and is characterized by a fracture network that has a significant impact on its fluid flow behavior.Therefore,accurately predicting the properties of fractured reservoirs and quantitatively assessing their hydrocarbon production potential are a challenging task that requires the integration of multiple data sources and techniques.Geophysical methods,particularly seismic imaging techniques,are powerful tools for characterizing fractured reservoirs and have been widely used for this purpose.However,the reliability and accuracy of these methods depend on the properties of the reservoir and the quality of the seismic data.Low resolution and high noise interference can lead to seismic wave attenuation and absorption,which makes it difficult to distinguish between different rock types and fluid phases,further affecting the identification ability of fractured reservoir models to identify them.Moreover,the nonlinear relationship between fracture parameters and seismic data may lead to the diversity and complexity of prediction,and the applicability or accuracy of fracture-based reservoir prediction methods are limited for different types and scales of reservoirs.Therefore,this paper focuses on fracture detection and attribute extraction,carries out numerical simulation of anisotropic waveform response,researches on machine learning methods related to fracture-type classification and research on high-resolution fracture refined portrayal of azimuthal seismic attributes,and achieves quantitative analysis with comprehensive evaluation of fractured reservoirs by introducing fuzzy logic algorithms which target multi-attribute fusion for prediction results of each task.The seismic response of anisotropic media is a complex phenomenon that depends on the distribution,geometry and physical properties of the fractures.Fractures have nonlinear effects on seismic wave propagation,and their presence leads to complex interference patterns and amplitude,phase,and frequency variations in seismic data.To consider anisotropic effects,this paper first investigates the variation of Thomsen anisotropy parameters and fracture parameters as a function of incidence angle,porosity,water content saturation,and net-to-gross(NG)ratio.By analyzing the sensitivity of different parameter variations to each anisotropy effect through petrophysical modeling,the source and effect of anisotropy in seismic data can be understood more deeply,and the pattern of amplitude variation with Offset and Azimuth(AVOAz)is visually compared,which is essential for further understanding and interpretation of anomalies and signal inhomogeneity in seismic data.Seismic reflection waveforms play an important role in oil exploration,which contain a large amount of geological information,such as lithology,reservoir and geological formations.In the field of machine learning,the role of label information and sample scales are very important as they provide true category information about the samples and provide vital support for model training and prediction.In this paper,we combine principal component analysis(PCA)and linear discriminant analysis(LDA)to construct a classification model with high accuracy for classification and prediction of tectonically fractured reservoirs through label learning and recognition of multi-scale,multi-label 3D seismic reflection waveforms.Moreover,in order to solve the inefficient manual sample labeling and small sample overfitting phenomena,a complete database establishment method is proposed in this paper.In this workflow,rich well trajectory information and logging interpretation information are used as the sources of adaptive tracing acquisition and labeling information,and then the sample scale is finely divided based on the Boyer-Moore majority vote algorithm for sample segmentation(BMMV-SS)method,which effectively enriches the diversity of the sample.The advantage of this method is that no manual labeling is required,and the target samples are obtained by simply inputting logging information into the workflow.However,the number of samples of various types in the initial sample database is often unbalanced,leading to an imbalance in the sample prior information in the model training set,which results in significant bias in the actual prediction.An effective method to solve the sample imbalance problem is the sample boosting technique,which aims to expand the scale of the training set,improve the generalization ability of the model,and alleviate the problems such as overfitting,but needs to be selected according to the specific task and data set.In this paper,the small sample imbalance problem is solved by a novel label shuffling balanced(NLSB)method using sample boosting methods in the spatial and frequency domains by incorporating the frequency variation characteristics of fracture construction.And the effectiveness of the method is verified from data simulation and real applications.Fractures are an important part of geological formations and have significant implications for reservoir evaluation,fracture prediction and reservoir description in oil and gas exploration and development.Therefore,fracture quantitative inscription has become one of the research hotspots in this field.In this paper,we discuss the quantitative fracture characterization method based on the orientation seismic attribute,which combines the orientation seismic attribute and the edge detection method by optimal surface voting(OSV)technique,and can achieve the accurate fracture characterization in3 D volume.This study adds a new idea in the field of fracture quantitative description,which enables a more accurate description of fracture parameters such as location,orientation and morphology,and has a more refined and accurate effect on reservoir evaluation and hydrocarbon reservoir description.Based on the analysis of fracture attributes,quantitative description and reservoir inversion results,it is necessary to further determine the influence degree of fractures in the reservoir and the reserve potential.By constructing the Interval Type-2 Fuzzy Logic System(IT2 FLS)and Geological Strength Index(GSI)model,multiple attributes such as fracture parameters,lithological inversion parameters and seismic attributes are integrated to quantitatively evaluate the coal structure.Compared with the traditional reservoir evaluation method,this study proposes a new method of multi-attribute fusion,which could more comprehensively consider the weight relationship and influence between different attributes,thus improving the accuracy and reliability of reservoir evaluation.In addition,the method can be applied to other types of fractured reservoir evaluation,which has certain generality and universality,and has a certain promotion effect on the development and research of fractured reservoir evaluation.