Research On Automatic Recognition Of Layer Of Neutron Lifetime Logging

The neutron lifetime logging is an important logging method in the mid-late period ofOilfield development, it can provide safeguard information for determining the high aquifer’slocation and reservoir flooded status, it makes convenient to find the potential of oilfieldremaining oil in different types of oil reservoir, and find the distribution of remaining oil inorder to make or adjust to an reasonable development plan, it is significant for Oilfield toincrease production and improve the efficiency of oil exploitation.At present, the research on improving the accuracy of distinguishing the neutron lifetimelogging layer mainly focus on the well logging technology, only few research focus on theinterpretation of neutron lifetime. The interpretation method of neutron lifetime logging isrough and simple,the depth correction of logging curve is artificial, the efficiency andaccuracy of depth correction are very low; furthermore, different interpreters give differentexplanation results to the layer, so the accuracy can not be guaranteed. Improving theinterpretation accuracy and efficiency of this method becomes the urgent need of thedevelopment of Oilfield.To solve the above problems, this paper uses hierarchical feature extraction to depthcorrection for neutron lifetime logging curve by SMOTE algorithm and Support VectorMachine combining method to make further study about the method of layer automaticrecognition. The main contents are as follows:1. Aiming at the existed depth error of neutron lifetime logging measurement,according to specialist’s experience, this paper proposes the Logging curve correction methodwhich is based on the hierarchical feature extraction to measure two curves depth correctionof different stratigraphic information. Taking advantage of the characteristics of logging curvesequential to going hierarchical feature extraction and depth correction for logging curve, firstlevel matching is making categorize matching to the curve paragraph of the calibration curveand standard curve, second level matching is to match by using the similarity of the curves,and finally the curve correction is realized.2. According to the characteristics of the standard SMOTE algorithm and theshortcoming which is existed in the unbalanced data applications, proposed the improvedSMOTE algorithm which is based on unbalanced classification, improving SMOTE algorithmmainly for the disadvantages of SMOTE algorithm sample classification vague, ignoring theimportance of boundary sample and no effective treatment on noise sample, according to deal with safety points, edge points and noise points separately, achieving the purpose of makingthe sample equilibrium and preparing data for next neutron lifetime stratification.3. Comprehensive analysis for neutron lifetime layer information,combined with theexperience of determining layer by artificially, extracting the available characteristicparameters and use binary Tree Support Vector Machine to classify logging layer. Theexperiments show that the classification accuracy can meet the practical requirements, Finally,research results are comprehensively applied to the neutron lifetime layer automaticrecognition combining with the actual condition.