Logging Information Processing Of Volcanic Rock Reservoirs In Jinlong 2 Well Area In Junggar Basin

The volcanic reservoirs in the Jinlong 2 well area of the Junggar Basin are characterized by low-porosity and low-permeability.Complex lithology of volcanic rocks restricts the accuracy of their identification,and the complex porosity-permeability relationship is difficult to describe by simple linear models.For this reason,it is usually not effective to predict the volcanic reservoir quality parameters by conventional logging interpretation methods.Therefore,the thesis attempts to use the ensemble learning algorithm,which is effective for solving non-linear problem,for identification of lithology,explanation of the reservoir parameters and identification of fluid of the area with the evaluation of effectiveness of methods.As a case,the volcanic reservoirs in the Jinlong 2 well area of the Junggar Basin are studied in this thesis based on the logging data.Firstly,we analyze the reservoir lithology,physical properties,hydrocarbon and electrical properties and their relationships.Then,the calibrated logging data,core data and thin section data are used to identify the complex lithology and establish the porosity model of the study area by a variety of nonlinear methods.Finally,in consideration of large vertical variation of permeability and unclear pore-permeability relationship,various machine learning algorithms are used to identify different flow units in the reservoir,and conventional methods are used to explain the permeability in the flow unit.By calculating the Pearson correlation coefficients between different capillary pressure curves and clustering,the reservoirs are divided into three categories according to the clustering characteristics combined with porosity and permeability and petrophysical parameters.Depth-matching the classified sample and coring sample with logging data,and the reservoir classification model is established by ensemble algorithm with the logging property.The result shows the good applicability of the ensemble algorithm.in the study area.In detail,the random forest is used to establish the porosity model,and the gradient boosting tree is used to establish the lithology identification model,identify the flow unit and classify reservoir.The combination of these algorithms may extent to whole area for logging interpretation.