تهيه ي نرم افزار مدل پيش بيني پارامترهاي پتروفيزيكي سنگ مخزن نفتي با استفاده از روش هاي چند متغيره ي آماري و زمين آماري

چكيده انگليسي :

Developing software for predicting petrophysical parameters of oil reservoir using multivariate statistical and geostatistical method Sadegh Karimpouli s karimpouli@mi iut ac ir 25th Februery 2009 Department of Mining Engineering Isfahan University of Technology Isfahan 84156 83111 IranDegree M Sc Language FarsiDr Nader Fathianpour fathian@cc iut ac ir Dr Keyghobad ShamsAbstractThe forefront stage in any production plan in petroleum industry is exploration which has always been considered asa risky costly and time consuming task In order to undermine and allviate this risky stage and adapt suitableproduction strategies and management it is necessary to obtain information on reservoir characteristics including insitu petrophysical parameters and their spatial distribution Due to significant drilling costs and time required for anypetroleum drilling program application of valid and accurate estimation methods for petrophysical parameters isinevitable and highly demanded In the current study exploratory data from a number of 11 exploratory wells including 3 cored ones from an oil fieldin south west Iran were available The structural settings of the reservoir had been previously determined to be ananticline with a length of 12 kilometers and width of 700 meters by petroleum geologists The most importantpetrophysical parameters for reservoir characterization employed in this study were porosity water oil saturation permeability and capillary pressure The first two parameters were calculated directly from well log data and theother parameters had been measured via core samples in laboratory In order to predict the measured parameters forunsampled wells it was primarily necessary to develop test and validate mathematical models employingmultivariate regression analysis and supervised committee machine neural networks Based on permeabilitydistribution throughout resevoir it was found that there exists two permeability populations one with lowpermeability representing dolomite and shale hosted lithologies Lithology class 1 and the other one with highpermeability representing limestone and sandstone lithologies Lithology class 2 The overal fitting between estimated permeability versus measured ones on validating data was quatified through R square R correlation coefficient to be 97 72 which is considerd as appropriate Through obtaining suitablemodel using Harris and Goldsmith 2001 capillary pressure model predicting regression models with great agreement with cored sample data for displacement capillary pressure and irreducible water saturation weredeveloped Applying the newly developed mathematical models to the unsampled exploratory wells and extending theestimated petrophysical parameters using 3D geostatistical estimation methods the block 3D solid models wereconstructed for each petrophysical parameter in part of reservoir estimation space spanning well positions Statisticsfrom 3D solid models show that the average porosity coincides with sandstone and shale lithologes whereaslimestone with some minor areas covered by dolomite has shown both low and high porosity behaviour as a resultof developed joints and fractures It is also found that the contact between 50 oil and water saturated interface varies from 2800 to 2830 metersdepth at the eastern part of reservoir in contrast with the western part where it emerges in 3000 meters depth Thepermeability gets its highest values where highly jointed and fractured porosity limestones are present Lowcapillary pressure areas coincides with areas having fractured porosities with suitable pore throat size Also areaswith low permeability represent shale dolomite or compacted nonfractured limestones having low porosity Finallythe total hydrocarbon volume above 5 porosity cut off value was estimated to be 11051800 3344580 cubicmeters To facilitate using the models developed through this study an algorithm for the systematic estimationprocedure was designed and a prototype program in Matlab e

كريم پولي ،صادق

تهيه ي نرم افزار مدل پيش بيني پارامترهاي پتروفيزيكي سنگ مخزن نفتي با استفاده از روش هاي چند متغيره ي آماري و زمين آماري

رگرسيون چند متغيره , تخمين تراوايي - شبكه عصبي تركيبي