Research Of Evaluation Method Of Data Space Structure In Geological Facies Modeling

In the process of Facies modeling of petroleum reservoir.It is critical to accurate characterize and represent the spatial structure features of geological outcrop,the sedimentary facies or lithoface in the geological knowledge,the mainly available methods are traditional geostatistics modeling based on the theory of regionalized variable(Such as the truncated Gaussian and sequential indicator modeling)?the modeling method based on the target(river channel simulation)?modeling method based on facies transition probability and the Multiple-point Geostatistics which is popular in recent years.However,it is always a weak link about how to evaluate the applicability of the modeling methods and the effectiveness of recovery of the data spatial structure in the process of facies modeling.The method based on Markov chain model is proposed to assess the spatial structure characters of facies modeling.This method can characterize the contact?split and mosaic spatial structure among different types through counting the self-transfer probability and cross-transfer probability amang different types in facies modeling results with different method.Compared with indicator variograms modeling,this method can be used to analyze asymmetric?unstable data space structure,and the description is more accurate;While the variogram is limited to two-point geostatistics algorithms models,it is only applicable to analyze symmetry space structure with the obvious smooth effect.In the view of the complex spatial structure in facies modeling and the multi-scale features of data spatial structure,adopt the analysis of the multi-scale spatial structure unit,put forward the clustering model based on probabilistic center in discrete image innovatively and establish the method system of analyze data spatial structure in facies modeling.This method takes the emergence probability of different type which at each spatial position in facies modeling structure unit to represent the probability of which is belong to the typical structural unit in this position,the superposition of the possibility in all position express the final possibility of belongs to this typical structure unit.Iterative clustering method to achieve the extraction and probability characterization of spatial structure unit in the prototype model(training images).Using the method of pattern recognition,analysis the characters of spatial structure unit in different facies modeling,and compare with the prototype model about the differences of the spatial structure unit distribution histogram and similarity,and evaluating the reproduce accuracy of data spatial structure in different facies modeling;this process is repeated on multiple scales,you can evaluate the reproduce effect of multi-scale spatial structure.The multi-scale spatial structure analysis of the different facies modeling indicates that sequential indicator modeling has the poor recovery effect for the complicated space structure,Markov chain modeling method based on transfer probability has the bad recovery effect for small scale space structure,but its performance is stable when characterizing the large scale structure.Multiple-point Geostatistics modeling is good at recovering the spatial structure,balance the performance on each scales,and be suitable for complicated sedimentary facies(Petrofacies)modeling.In the north of Sulige gas field,su11 block,upper palaeozoic,H8 sector,the application of modeling show that multi-scale evaluation of data spatial structure helps to optimize model parameters and to improve the modeling accuracy.