Fuzzy SVM Applied In Water-flooded Zone Identifying

At present, the mainly mining method is filling the oilfield with water in most ofour country’s oilfield. Long-term water flooding exploitation method makes manyoilfield’s water content relatively high. F oilfield is in the stage of that mining isrelatively difficult and the amount of oil is small. The most important and commonmethod to judging the value of the oilfield’s mining is the water-flooded zoneidentifying. Because of the process of the water-flooded zone identifying dynamiccomplex, the difficulty of water-flooded zone classifying increases. Fitting andregression of the reservoir data, just like porosity(PORE), permeability(K), saturationinterpretation(SW) etc, is the most important and key step in the water-flooded zoneclassifying.This article uses data from coring data obtained in F oilfield exploration anddevelopment. Coring data is the most intuitive, most reliable data that reflects theunderground geological characteristics. At the same time, because of the high cost ofcoring, the slow drilling speed, the coring data quantity is less and the sample space issmall. So it conforms to the SVM problem.The purpose of this paper is combining the SVM and the fuzzy mathematicswhich already has the ripe theoretical system each. Before this, we should contrast onthe different knowledge learning algorithms. The FSVM has already applied inseveral fields, but rarely applied in the problem of the oilfield’s reservoir data. Thispaper is looking forward to establishing the reservoir data, just like porosity,permeability, saturation interpretation etc, getting the function curve and parameters,to help F oilfield identifying the water-flooded zone, getting the classification resultswith relatively high accuracy.Support Vector Machine was first proposed for the binary classification problem,and extended to multi-classification problem in practical applications. So this paper research the SVM in the aspect of classification algorithm, on this premise, toresearch in the fitting and regression of the data, especially the nonlinear regressionalgorithm of the small sample data in high dimensional space.Fuzzy SVM is the organic combination of fuzzy mathematics and SVM, by theinput data into a high dimensional feature space by a nonlinear transformation, toobtain the optimal classification hyper plane, to make the original input space to bedivided data are linearly separable, and using the concept of fuzzy sets and SVM inthe process.The key of fuzzy algorithm design is the design of membership function. Thevalues of this function must be true, accurately reflect the membership relationbelongs to the domain or the system sample. Membership function is mainly to thedistance between the center of the sample to measure the degree of membership basedon size.Although the FSVM has been applied widely in the field of pattern recognition,but in the reservoir data problems, the application of FSVM is not widely. In thispaper, fuzzy set theory is combined with SVM, and the application in reservoir data,by using the fuzzy modeling of porosity, permeability, saturation of the vectormachine, testing and actual data from the production process, to achieve better results.