Study On Lithologic Identification Of Glutenite By Conventional And Imaging Logging Data

Xujiaweizi fault depression is a large fault at the northern Songliao basin. The explored reserves of natural gas which has been refered to volcanic rocks and glutamate gas reservoir of Yingcheng formation is more than 200 billion cubic meters, the next step is to explore deep natural gas, and deep glutamate segment is the most important objective reservoir in this zone. Owing to close material source, multi-water system and rapid changeful sedimentary environment, reservoir lithology is complex and logging response has irregular change. Accurate identify of lithology becomes the priority and difficulty to adjust and exploit this type of reservoirs.Comprehensively based on theories and methods of geophysics, mathematical statistics, computer.etc disciplines, the basic material of borehole micro-scanner log (FMI), convention- al logging, mud logging and core, analyze, divide and predict the lithology of deep glutamate reservoir in Xujiaweizi. Firstly the crossplot of lithology identification is manufactured by using geophysical methods. Glutenite is precisely divided and described by using high resolution and visualizational image of borehole micro-scanner log, recognitional mode is established by FMI image of different lithology which belongs to glutamate sedimentary layer.Then, combined the material of pressure coring, the sensitivity of borehole log which reflects glutenite reservoir's lithology is optimized by using the analytic method of decision tree. Three log response values of original formation resistivity, interval travel time, and natural gamma ray are optimized as a characteristic to identify the lithology glutamate reservoir, a decision tree model which is used to identify the lithology is established. At the same time, in order to solve logging lithology identification problem better, small sample of support vector machine method is introduced, based on the data of actual logging and lithologic section,support vector machine is learned and trained,a model of support vector machine which is used by identifying logging lithology is established.According to the size, the lithology of deep glutamate reservoir in Xujiaweizi is devided by seven types: sandy conglomerate, glutenite, pebbled sandstone, fine sandstone, ger llton, silty mudstones and mudstone based on core data. The identify effect of conventional crossplot method is poor, while the forecast effect of FMI's signature which has been adjust by core data is better. the coincidence rate of decision tree method reached 82.44%, the coincidence rate of support vector machine method reached 82.95%.The prediction effect of the four methods(conventional plate, FMI's signature, decision tree, support vector machine) is synthetically compared, research shows that if the data of FMI is present, the forecast effect of other methods is compared and adjusted based on FMI's signature.the coincidence rate of forecast result improve is more precise, the identify effect is better. If the data of FMI is absence, the best way to solve identify of complex lithology is effectively combined decision tree method, support vector machine method with geophysical method.