Prediction Of Pore Structure In Carbonate Rocks Based On Acoustic Properties

Existing methods of well-logging interpretation often contain errors in the exploration and evaluation of carbonate reservoirs due to the complex pore structure.The differences in frequency ranges and measurement methods deviated between the acoustic well-logging and indoor ultrasonic test cause inconsistent results.The imaging or nuclear logging still have considerable limitations,such as high financial cost and strongly empirical problems.In this paper,the effects of pore structure on acoustic well logging are studied by experimental and numerical simulation methods.Based on the qualitative identification and judgment of secondary pores using well logging data,combined with artificial intelligence methods,the prediction of pore structure in carbonate rocks is developed.In the experiments,we establish a new method of devising a borehole model with controlled pore structure and distribution.The carbonate rocks were simulated by a mixture of carbonate cuttings and cement slurry systems.Quantitative control of secondary pore structure in a large-scale model was achieved by different shapes and quantities of silicone disks and threading straightening method.A self-build acoustic logging system was used to measure critically refracted waves.The effects of pore structure(porosity,aspect ratio-AR,and size)on the acoustic logging signals in time and frequency domains were studied by the Hilbert-Huang transform(HHT)method.The results show that the velocity of the Stoneley wave is less sensitive to pore structure than P and S waves,while strong effects are observed on the attenuation of the Stoneley wave.The best-fitting power and logarithmic functions are used to quantify the relationships between the properties(velocity and attenuation)and pore structure.Predictions of the conventional rock physics models used on high frequencies always overestimate the velocities at the well-logging scale.Furthermore,pore AR and size have similar mechanisms of scattering attenuation,which differ from the absorbing attenuation of porosity,and the effect of size is much stronger.Based on the elastic wave equation and the principle of control variable method,a 2-D axisymmetric borehole model with complex pore structure was developed and the numerical simulation method for acoustic logging was constructed.The modeling results show that the power function can well discribe the effects of pore structure on the acoustic waves,while the velocity of the Stoneley wave is not sensitive to the pore structure.Crack-like pores with AR<0.1 greatly affect the velocities of P and S waves,while ‘spherical' pores have less effect.The model with larger pore sizes have high velocities of P and S waves.The velocities calculated by the equivalent medium theory are always higher than the numerical simulation results.The velocity deviation caused by the difference in frequency is much smaller than the pore structure.The fractal dimension increases with the pore AR or size when the porosity is constant,which can be described by a simple power function.The established borehole model obtained the influence of the pore structure in the lower frequency range than the conventional petrophysical model.The fracture development formation was identified by using the depth points and waveform information of the logging data.The results show that the method based on equivalent medium theory maintain high stability and accuracy in reflecting the fractures in the case of poor borehole situation.The fractal dimension of the acoustic time difference log and the double lateral resistivity anomaly log become larger when the structure becomes more complex.The dual-tree complex wavelet transform was selected to perform multi-scale processing on the acoustic logging data.The energy information in high-frequency of the acoustic logging data exhibits a distinctly high value in the fractured formation.Principal component analysis(PCA)method was used to integrate the multiple methods of fracture identification.The fracture development formation was characterized by calculating the factor comprehensive score,which can avoid the errors of a few methods on the stability of the overall results.Shear wave splitting method was used to obtain the spatiotemporal variation of the anisotropy around the hydraulic fracturing well,and then the large-scale fractures are identified and analyzed.The fault damage zone near the fracture surface will destroy the fracture formed by hydraulic fracturing.The time-domain characteristic parameters extracted from array acoustic logging data were integrated by PCA.A new prediction model of genetic algorithm-support vector machine method based on HHT of acoustic logging data was reported to predict the fracture density,which verified by the core analysis and imaging logging data.The results show that the fracture density has a great effect on the attenuation of high-frequency components.The energy of the Stoneley wave and S wave has higher sensitivity.Compared with the time domain,the distribution in the high-frequency domain has a greater correlation with fracture density.Based on the array acoustic logging data obtained by numerical simulation,and the classification of pore AR based on convolutional neural network(CNN)was established.The wavelet transform was used to transform the 1-D acoustic curve into 2-D time-frequency spectrum as the input data of CNN(fractures and moldic-pores)according to the AR.Based on the description of the pore structure of real core analysis,the formation pores are divided into two types.The accuracy of the method using actual array acoustic logging data can reach 90%.By utilizing the overall information of the acoustic wave,the method can avoid the error generated in the speed picking of the logging curve and facilitate the application to the entire carbonate reservoir.