Study On Petrophysical Modeling For Fractured-cavern Carbonate Reservoirs

Carbonate rock accounts for only 20% of sedimentary rock,but carbonate rock reservoirs contain rich oil and gas resources and production worldwide,accounting for more than 50% of the total proved oil and gas reserves.Carbonate rock reservoirs with huge exploration and development potential have become increasingly important exploration targets.Due to the complex karstification and tectonism during the burial process,various reservoir spaces have been intricately developed in carbonate reservoirs,forming various types of reservoirs with different storage and infiltration capacities for oil and gas.Compared to other rock types of reservoirs,carbonate rock reservoirs have more complex lithology,mineral components,and pore structure,with strong heterogeneity.Accurately identifying carbonate reservoir types that are rich in oil and gas resources,and constructing petrophysical models that can improve the accuracy of reservoir prediction and description have become important issues in carbonate reservoir exploration and development.In view of this,this paper selects the supervised learning method of kernel Fisher discriminant analysis(KFDA)to achieve the recognition of carbonate reservoir types.By introducing local and global kernel functions and their linear combinations,a total of three kernel functions are obtained.Six types of logging curves,namely GR,CAL,DEN,AC,CNL,and RT,are used as input data to train,test,and validate the three KFDA models,By comparing the classification accuracy and generalization ability of the three models,it is found that the model based on mixed kernel function has stronger learning,classification,and generalization abilities,and proves the effectiveness and rationality of this method for identifying carbonate reservoir types.Then,aiming at the complex pore structure and structural characteristics of siliceous minerals in the fracture-cavern reservoir in the study area,the pore structure is quantified by calculating the pore aspect ratio parameter,and the structural characteristics of siliceous minerals are better described by improving the method of obtaining the elastic modulus of rock matrix.A new fracture-cavern carbonate rock physical model is constructed by combining multiple classical rock physical models,The new model,together with the classic Xu Payne model and the improved Zhang Bingming model,was applied to the reservoir in the study area,and the accuracy of the three models was verified through shear wave velocity prediction and correlation analysis.The prediction results of shear wave velocity show that compared to the other two models,the new model can well match the conventional reservoir intervals in the study area,and also can well match the siliceous mineral reservoir.The prediction results are basically consistent with the measured velocities.At the same time,the correlation analysis results also show that the new model is superior to the other two models,thereby proving the effectiveness and applicability of the new model.Finally,this paper identified fracture-cavern carbonate reservoirs with excellent reservoir performance and rich oil and gas reserves using KFDA method,and constructed a petrophysical model for this type of reservoir,achieving the following two results:1)The supervised learning method of kernel Fisher discriminant analysis is used to identify carbonate reservoir types;2)A new petrophysical model of fracture-cavern carbonate rocks is established.