Article ID | Journal | Published Year | Pages | File Type |
---|---|---|---|---|
4740511 | Journal of Applied Geophysics | 2011 | 12 Pages |
The analysis of well logging data plays key role in the exploration and development of hydrocarbon reservoirs. Various well log parameters such as porosity, gamma ray, density, transit time and resistivity, help in classification of strata and estimation of the physical, electrical and acoustical properties of the subsurface lithology. Strong and conspicuous changes in some of the log parameters associated with any particular geological stratigraphy formation are function of its composition, physical properties that help in classification. However some substrata show moderate values in respective log parameters and make difficult to identify the kind of strata, if we go by the standard variability ranges of any log parameters and visual inspection. The complexity increases further with more number of sensors involved. An attempt is made to identify the kinds of stratigraphy from well logs over Prydz bay basin, East Antarctica using fuzzy inference system. A model is built based on few data sets of known stratigraphy and further the network model is used as test model to infer the lithology of a borehole from their geophysical logs, not used in simulation. Initially the fuzzy based algorithm is trained, validated and tested on well log data and finally identifies the formation lithology of a hydrocarbon reservoir system of study area. The effectiveness of this technique is demonstrated by the analysis of the results for actual lithologs and coring data of ODP Leg 188. The fuzzy results show that the training performance equals to 82.95% while the prediction ability is 87.69%. The fuzzy results are very encouraging and the model is able to decipher even thin layer seams and other strata from geophysical logs. The result provides the significant sand formation of depth range 316.0– 341.0 m, where core recovery is incomplete.
► An advance technology fuzzy inference system is made to identify the geological stratigraphy. ► A model is built on well log data and that used as test model to infer the strata not used in simulation. ► Initially an algorithm is trained on 80% of data and tested on remaining data of a reservoir system. ► The efficacy of this method is established and results are compared with available coring samples. ► The fuzzy results confirm that the training and prediction performance are highly encouraging.