The Application Research Of Fuzzy Cluster Analysis Method In Reservoir Classification

Petroleum reservoir classification is an important part in reservoir management which is the basis of evaluating reservoirs and instituting reservoir development projects. It's the necessity of reservoir classification research. Proper development techniques based on scientific and reasonable reservoir classification is very important to enhance oil recovery efficiency and economic profits. The reservoir data has the character of large amount, many parameters and diversities. In the process of reservoir classification ago, development projects have been made by reservoir managers'experience, which have only choose one parameter in common use as standard mainly and considered other parameters. This kind of classification could not consider many parameters together scientifically, so it would make wrong reservoir development projects easily and cause the economic losing badly.Applying the cluster analysis method to reservoir classification can make the classification change from one parameter to many parameters, which can make the reservoir classification results be scientific and reasonable.However, the classical statistics cluster analysis methods need supervision, such as the threshold, the initial cluster center and the cluster number. Moreover, they are inefficient on solving the problems of individual error data in clustering samples.Therefore, this dissertation is based on classical statistics cluster analysis methods, comparing to random variables'numeric character in probability theory, which presents fuzzy random variables'statistic character and fuzzy statistics cluster analysis method.In order to classify the oil reservoir conveniently, this dissertation is based on the classical statistics dynamic cluster analysis method and fuzzy statistics cluster analysis method, creates a fuzzy multivariate statistical system platform which has the function of create new data, open data, edit data, dynamic cluster analysis, fuzzy cluster analysis and save results etc. The interface is amicable and easy to operate. It's also has the transplant ability. This platform has been used to classify oil reservoirs in an east oil field and been approbated by reservoir engineering experts. Because of the complexity of oil reservoir classification, this platform needs more researching and expansion in the future.After the comparison and cluster validity analysis, it shows that the fuzzy statistics cluster analysis method makes the influence by the abnormal data in the cluster process and the cluster results'dependence upon the initial cluster samples weaken obviously because of the deviation subordinate degree and initial cluster centers, which assures the unsupervised nature of cluster analysis efficiently.The conclusion is that in reservoir classification the fuzzy statistics cluster analysis has validity, comparing to the classical cluster analysis. It is unsupervised and can fix reservoir types scientifically, confirm development modes rationally, make decisions effectively, provide services to reservoir management and increase exploitation proceeds to the oil field.