Multi-scale Classification And Evaluation Of Tight Reservoir In Fuyu Reservoir Of Bay An Chagan Area Based On Machine Learning

The fuyu oil reservoir in Bayanchagan area of Songliao basin has little work and insufficient geological understanding,and the heterogeneity of the area is strong,and the pore structure is complex.It is difficult to realize multi-scale characterization of tight reservoirs by conventional methods.Thus,on the basis of summarizing the existing geological knowledge and exploration results as well as previous studies,machine learning method is applied to the micro-macro multi-scale comprehensive evaluate the tight reservoir of Fuyu oil reservoir in this area.In this study,the reservoir quality index is introduced as the macroscopic parameter of reservoir evaluation,and based on high-pressure mercury injection technology,machine learning method including the BOX-COX conversion,grey correlation analysis,as well as the integrated clustering analysis method is applied to objectively,efficiently and accurately evaluate the micro-pore structure of reservoir,the results of grading evaluation were verified and evaluated by casting thin sections and scanning electron microscopy.According to the classification evaluation results of tight reservoir combined with logging data,and the method of support vector machine(SVM)was applied to realize the study on the plane distribution of tight reservoir in the study area,and then verified according to the distribution of sedimentary macroscopic sand body.Finally,combining with the distribution rule of faults and oil and water,the distribution rule of high quality reservoirs in the plane is predicted,so as to realize the multi-scale classification and evaluation of tight reservoirs from micro to macro.The comprehensive research results show that:(1)The permeability of the study area is 0.001?0.93mD,and the porosity is 3.7%?23.2%.The lithology is dominated by siltstone and argillaceous siltstone,and the rock types are mainly lithic sandstone and feldspar lithic sandstone.This area belongs to tight reservoir with complex pore structure and strong heterogeneity.(2)Class I reservoirs in the study area are high quality reservoirs in the study area,belonging to type of medium-compaction,strong-dissolution,weak-cementation.This type of reservoir has low tightness,good sorting,good pore-throat distribution and good connectivity.Class II reservoirs are type of strong-compaction,medium-dissolution,medium-cementation.This type of reservoir has the characteristics of medium density,medium sorting,relatively uniform pore-throat distribution and low connectivity.Class ? reservoirs have the worst physical properties,strong heterogeneity and tight properties,belonging to the type of strong-compaction,weak-dissolution,strong-cementation,uneven pore-throat distribution and poor connectivity.(3)According to the plane distribution law of tight reservoirs in each small layer of F?,F? and F? oil formation in the study area,it can be known that the high quality reservoirs developed best in the F? reservoir group in the study area,and the reservoir area was large,mainly distributed in the central and northern part of the study area,showing increasing trend followed by decrease.The development of high quality reservoirs in F? reservoir formation is worse than F? reservoir group,which is scattered in the north of the study area,and generally presents a trend of gradual decrease.There are no high quality reservoirs in the F? reservoir group;(4)The multi-scale grading evaluation method of tight reservoir based on machine learning has high accuracy,and the prediction results of tight reservoir types are basically matched with the results of scanning electron microscopy,casting thin sections and sedimentary microfacies,which is in line with the practical geological significance.