Data Driven for Gray Relational Analysis of Recognizing Oil-bearing Characteristics in Reservoir

The paper proposed a method of data driven gray relational analysis for recognizing oil bearing characteristics in reservoir. The method follows the objective process from data to information and from information to recognition. Firstly, reduce attributes based on training data and obtain the key attributes for recognizing oil bearing characteristics (oil layer, inferior oil layer, dry layer and water layer) by fusion of genetic algorithm and fuzzy c-means. Secondly, take the center of clusters (different oil bearing formation characteristics) of training data as the reference sequence of recognizing oil bearing characteristics in reservoir. Thirdly, obtain the weight of each key attribute through relief algorithm. At last, the testing data was estimated by data driven gray relational analysis. The paper takes oilsk81 well data in Jianghan oilfield of China as training data and takes oilsk83 well data as testing data, the estimated results are the same as the real oil bearing characteristics of each layer in oilsk83 well.