Automatic Identification Of Lithology And Sedimentary Microfacies Using Logging Curves

Logging curve is the direct expression of underground lithologic combination characteristics and sedimentary microfacies conversion characteristics.Through the correct geological interpretation of a large amount of information rich in logging curve,the development plan of oil and gas field can be fully formulated,and it has extremely important significance for improving oil and gas recovery.With the continuous improvement of the measurement accuracy of various logging tools and equipment and the increasing of logging technology methods,the geological information contained in the logging curve has also increased geometrically,and the logging interpretation technology has gradually changed from qualitative interpretation to semi quantitative interpretation,as well as the current quantitative interpretation with high accuracy requirements.The basis of automatic lithologic division is the accurate identification of lithologic interface,and the activity function method provides a fast and simple division method.According to the characteristics of extreme value of activity curve at the place where logging curve changes violently,combined with the actual thickness range of lithologic layer in the study area and the effective screening of extreme value of activity curve,the lithologic interface identification of block 482 is completed.On the basis of accurate identification of lithologic interface,the characteristic parameters of single rock layer curves of GR,LLD and LLS logging curves with the most obvious response to lithologic division are extracted,and the neural network depth learning model is established.According to the expected output rate and prediction rate,the combination of different characteristic parameters of different curves and the selection of structural parameters of neural network model are optimized in turn.Three curve characteristic parameters of GR mean value,GR variance and LLD variance of single rock layer are selected to complete the automatic identification of five single rock layers in block 482,including fine sandstone,siltstone,silty mudstone,argillaceous siltstone and mudstone distinguish.On the basis of accurate identification of lithologic interface,the characteristic parameters of single rock layer curves of GR,LLD and LLS logging curves with the most obvious response to lithologic division are extracted,and the neural network depth learning model is established.According to the expected output rate and prediction rate,the combination of different characteristic parameters of different curves and the selection of structural parameters of neural network model are optimized in turn.Three curve characteristic parameters of GR mean value,GR variance and LLD variance of single rock layer are selected to complete the automatic identification of five single rock layers in block 482,including fine sandstone,siltstone,silty mudstone,argillaceous siltstone and mudstone distinguish.On the basis of accurate identification of lithologic interface,the characteristic parameters of single rock layer curves of GR,LLD and LLS logging curves with the most obvious response to lithologic division are extracted,and the neural network depth learning model is established.According to the expected output rate and prediction rate,the combination of different characteristic parameters of different curves and the selection of structural parameters of neural network model are optimized in turn.Three curve characteristic parameters of GR mean value,GR variance and LLD variance of single rock layer are selected to complete the automatic identification of five single rock layers in block 482,including fine sandstone,siltstone,silty mudstone,argillaceous siltstone and mudstone distinguish.