| With China's economic development is accelerating,clean energy supply short-ages become more intensified.coalbed methane(CBM)is not only a clean energy source,but also a major source of disasters for coal mine safety production.But many factors influence the selection of CBM assessment,it is more complex.Aiming at the engineering practice of Shizhuang southern block in Qinshui Basin without seismic data and abundant logging data,this paper uses logging data as input,for through logging curve preprocessing,reservoir physical property parameters and reservoir stress parameters quantitative prediction,In order to evaluate the CBM,the random forest machine learning method is used to select the sweet spot of CBM,which has certain guiding significance for the development of coalbed methane resources in the study area.First,the logging curve preprocessing.For the presence of ambient noise and system logs taken Wavelet multiresolution analysis noise removal process,removing high frequency noise curve section to improve the ratio of S/N.Assuming that similar logging exsit consistent statistical laws,the histogram correction is used to improve the lateral comparability of logging.For CBM wells with missing key logging curves,a logging reconstruction method based on principal component analysis is used to model to achieve the reconstruction of key well curves in the area.Secondly,the model for calculating the properties of CBM,the study area sequentially porosity,permeability,and effective thickness of coal quantitative prediction.Industrial analysis and binding density curve data,a model to calculate the volume of formation porosity seam;with deep and shallow bilateral fissure porosity of the resistivity curve calculation,which is converted to a normalized value of the relative permeability to characterize the permeate region rate;log response difference between using quantitative forecast seam coal and rock thickness;Construction Vg index relative gas content,gas content 3#forecast seam.The results show that the 3#coal seam has relatively high porosity;except for Well Group 634,the permeability of other wells is relatively low;the coal seam is thicker with small lateral changes,which is conducive to coalbed methane accumulation;the coalbed methane content in the area is relatively high(15-23t/m~3).Third,the K-means cluster analysis method is used to automatically divide the lithology of the logging curves,and build a 3D geological model.Combined with 3D geological model,in-situ stress boundary conditions and rock mechanics parameters,finite element simulation simulates the in-situ stress distribution of coalbed methane reservoirs.The results show that the Mises stress difference of the 3#coal seam is large,and the in-situ stress concentration difference is obvious.The difference between the SH?Sh and stress difference of SH and Sh is small in the region,which is conducive to the determination of hydraulic fracturing parameters and the stability of the wellbore.In addition,the difference of the horizontal principal stress difference coefficient in the area is relatively small,and the overall difference is less than 0.2.It is easy to form mesh joints during hydraulic fracturing,which is beneficial to the development of CBM.Finally,reservoir parameters and initial stress as input data,divided by the Bootstrap data set,determining the measurement error model parameters,model training high accuracy predictor(94%),the stability of the model evaluation model Sex.Random forest simulation was performed on the interpolation results of the data set,the difference between the simulation results and the Kriging interpolation was compared,and the sweet spot selection in the study area(634 well group)was completed in combination with the field operation conditions.And compared with the actual CBM extraction data to verify the effectiveness and reliability of this method.The master's degree thesis has 43 pictures,9 tables,and 106 references. |
PDF Full Text Request