Predication And Stochastic Modeling Of Coalbed Methane Reservoir Properties In Luling Well Field

Absolutely, compared with oil and gas clasolite reservoir, reservoir stochasticmodeling is still in its preliminary stage in the process of coalbed methane(CBM)exploration and development with many fields to be deeply researched in China. Lulingcoal field in Anhui, China, was chosed to become the study example. CBM reservoirproperties predicting mathematical model,stochastic modeling,geological model applica-tion in CBM resource assessment,and geological control mechanism of CBM resourcesspatial variability by ArcGIS and Petrel were studied in details.In order to improve the interpolation precision of spatial variability of CBM reservoirproperties, a radial basis function neural network(RBFNN) was designed based on optimalclone differential evolution algorithm (OCDE). Normalized root mean square error(NRMSE) and mean relative error percent(MREP) respectively come into use forevaluating the levels of fitting and prediction precision of RBFNN and kriging for spatialinterpolation of coalbed properties.In the case study in two-dimensional space,the MREP of coalbed thickness,volatileyield (Vdaf),ash yield (Aad) and moisture content (Mad) were all less than or equal to10%when the sample size was at84.For the coalbed thickness,Aadand Mad, the NRMSE andMREP of RBFNN interpolation were fewer than that of kriging with different samplesizes,which indicates that RBFNN interpolation has higher accuracy than kriging.As anparticular case,the similar interpolation precisions appeared in Vdafdue to the samplingnormal distribution and less deviation coefficient. Our study shows that samplingdistribution and variation extent has an impact on interpolation precision.The more similarto normal distribution and less deviation coefficient in samples,the more high interpolationprecisions emerge in coal bed properties.When sample probability distribution is not innormal distribution,RBFNN interpolation has higher accuracy than kriging.In thethree-dimensional space, RBFNN interpolation precisions is the same as intwo-dimensional space and also has the impact of the sampling distribution and variationextent.Stochastic simulation method,known as estimation&simulation error(ESE),wasameliorated for CBM reservoir characterization. In the method, the unconditionalsimulation process was a vital step,which was achieved by radial basis function neuralnetwork (RBFNN) based on MATLAB and ArcGIS geographic informationsystem(GIS).The method was applied to generate fifty stochastic simulation realizations of coal bed attributes about No.8coal seam.These stochastic geological models have richerspatial variability informations than the unconditional simulation,and better reflect theuncertainty of spatial variability of CBM reservoir property.On the other hand, stochasticgeological models about volatile yield,ash yield,moisture content,gas content and porositywere made in the study based on the Petrel software platform.On this foundation of the volumetric method, a spatial analysis model andmethodology were designed to calculate CBM resources with deterministic geologicalmodel on ArcGIS geographic information system. Based on the same method,one groupsof fifty gas resources were calculated with the data obtained by using fifty randomrealizations of CBM reservoir. The resources range were20.84-25.75×10~8m~3,the averageresources were22.48×10~8m~3,and the standard deviation was0.981.Methodology waspresented for the cell-scale spatial variability of gas resources probabilities,which issubsequently applied to the case study. The spatial variability of gas resources probabilitieswas performed by the probabilities calculation in the fifty raster data layers of gasresources from fifty random realizations.The geological control mechanism of CBM resources spatial variability was studiedby mapping and spatial analysis of two-dimension and three-dimension geological modelon ArcGIS and Petrel software platform. The high CBM resources in the north of westsection of Luling well field is identical with Wang gezhuang syncline in geographicalposition,which reveals the internal syncline structure is in favour of CBM enrichment.Inthe tectonics simple central section the thickness of CBM reservoir is the most significantcontrol for CBM resources. Abundant drag folds formed by fault at the both wings of Xiaoshijia anticline play an important role in enrichment of coalbed methane in the southeastsection of the No.8CBM reservoir. Trap effect happened in roof, floor and muddyintercalation of coal seam.The dissertation research significance is embodied in some new methods presentedfor CBM reservoir characterization and geological modeling, and an application examplefor reference in CBM exploration and development.There are fifty figures, fourteen tables and one hundred and eighty-four references inthe dissertation.