Study On Hydrocarbon Detection Method For Carbonate Rocks In Deep Leikoupo Formation Of Western Sichuan

Carbonate reservoirs in the deep Leikoupo Formation in Western Sichuan have been the focus and difficulty of hydrocarbon exploration and development in the Sichuan Basin all the time.However,the successful detection of hydrocarbon and drilling is challenging to obtain due to the complex characteristics(e.g.,significant burial depth,weak seismic response,weaker pore fluid response,and complex gaswater reservoir relationship).Most direct hydrocarbon detection methods have unsatisfactory performance on the deep Leikoupo Formation carbonate rocks.Therefor it is necessary to develop a set of ways to improve the accuracy of hydrocarbon detection.Based on the classical gas-bearing prediction theory,the study introduces a set of methods which includes two-dimensional empirical mode algorithm(BEMD),TeagerKaiser energy operator method(TKEO,CTKEO),and deep neural network method(DNN)to evaluate the relationship between prestack data and reservoir gas-bearing.The learned relationship can enhance the accuracy of hydrocarbon detection.The main research contents and achievements are as follows:(1)Investigation and analysis of carbonate reservoir forming mechanism,reservoir characteristics,and seismic data quality of deep Leikoupo Formation in Western Sichuan,creating a direct gas-bearing detection scheme for Leikoupo formation reservoir.The favorable carbonate reservoir of the deep Leikoupo Formation in Western Sichuan is mainly the top weathering crust and the top of the third and fourth member of the Leikoupo Formation.The drilling results show that the reservoir gas-water relationship is complex,thin,and the distribution law is complicated.At the same time,the prestack data have a low signal-to-noise ratio and unequal gathers.(2)Based on the classical amplitude variation with offset(AVO)inversion theory,different domain algorithms are introduced to learn the direct relationship between prestack gather and hydrocarbon.A direct hydrocarbon detection method is formed.At present,most direct gas-bearing methods are based on AVO theory,which lays a strict theoretical foundation between amplitude variation with offset and hydrocarbon.Therefore,based on the quality of prestack gathers,a prediction method of hydrocarbon is proposed to optimize prestack gathers,enhance the sensitivity of data hydrocarbon,and directly establish the nonlinear mapping relationship between data and gas content.(3)An adaptive prestack gathers optimization method based on BEMD is proposed.According to the characteristics of prestack data,an improved BEMD algorithm based on structural morphology algorithm and cubic spline interpolation algorithm is developed.The seismic data are decomposed into components containing different characteristics quickly and efficiently,and then various features are processed accordingly.Finally,the prestack and poststack data are reconstructed and optimized.The method can adaptively optimize the prestack gather and retain the internal characteristics of the prestack gather.The actual data test in the study area verifies the effectiveness and superiority of the method.(4)A prestack gathers resolution enhancement method based on TK energy operator methods is proposed.The prestack gathers are processed by using the sensitivity and correlation of TK class operators to instantaneous signals.TKEO operator is used to increasing the longitudinal resolution of single-channel signal;CTKEO operator enhances the lateral variation characteristics of the movement.The sensitivity of prestack signal to gas content is improved by combining TK methods.The model data verification shows that the data processed by this method has a higher resolution to the gas-bearing layer and has strong pertinence to the top weathering crust reservoir of Leikoupo Formation in Western Sichuan.(5)The nonlinear mapping relationship between prestack data and gas-bearing property is established by the deep neural network,and then the gas-bearing property is detected.This method gives full play to the advantages of deep learning in classification learning and extracts deep information from the optimized data to directly characterize the gas-bearing characteristics.Marmousi2 model test shows the feasibility and accuracy of the method.(6)By combining various methods,a set of direct hydrocarbon detection methods for carbonate rocks in the deep Leikoupo Formation in Western Sichuan is formed.BEMD optimization method is used to improve the signal-to-noise ratio of prestack gathers and enhance the internal characteristics of signals.The TK energy operator method is used to enhance the resolution for the gather.Finally,the deep neural network algorithm is used to establish the nonlinear mapping relationship between hydrocarbon and prestack gather to achieve direct hydrocarbon detection.A data-based and targeted direct prediction method of gas-bearing property is formed step by step.(7)Based on the actual data of carbonate rock in deep Leikoupo Formation in Western Sichuan,the gas-bearing prediction results are consistent with the current drilling results.The depth of Leikoupo Formation in the study area is more than 5000 meters.At present,there are only three wells.In the case of no apparent structural change,the reservoir differences of the three wells are significant,so it is relatively difficult to predict the gas-bearing property.This research method has obtained more accurate gas-bearing prediction results in different target layers,weathering crust,Lei3 and Lei 4.Comparing and analyzing the prediction results of AVO inversion method further shows that this research method has dramatically improved the prediction accuracy of gas-bearing property of Leikoupo Formation in Western Sichuan and provides an idea and way for direct hydrocarbon detection.