Method Research And Application Of Gas-Water Identification In Natural Gas Reservoir

It has always been a concerned problem about gas-water identification accurately in petroleum exploration and development. How to use the existing logging data to take practical technical methods and identify reservoirs accurately is of great significance for the expansion of the natural gas reserves and increasing the value of the gas reservoir exploration. Gas-water identification accurately and evaluation have an important practical significance for the exploration and development of gas fields. Carbonate reservoirs is a common reservoir. Its reservoir properties are poor. And heterogeneity, the mud intrusion, salinity level, expanded diameter of Vuggy reservoir made the carbonate fluid properties more difficult to identify. Most gas field reservoirs of the Sichuan are carbonate reservoirs which are characterized by small porosity, low permeability. For current carbonate log interpretation, there are still some difficulties for gas-water layer identification of low porosity and low permeability in using a single conventional logging interpretation. So, the research of gas-water identification and evaluation is of great significance in carbonate reservoirs. Compliance rate of artificial interpretation of the gas-water layer in domestic logging companies is very high, but they mainly rely on the rich practical experience of engineers. It is extremely unfavorable for the promotion of the interpretation of the gas-water layer. There is not human experience in existing interpretation softwares. And there is a bid difference between computer interpretation accuracy and artificial interpretation accuracy.Based on the conventional logging data, firstly the paper introduced the geological characteristics of the study area, the physical properties of reservoir, the logging curve characteristic parameter extraction, the principles of sample selection, the pretreatment methods of the sample data. Finally we determined six logging curve characteristic parameters. They are the CNL(neutron porosity), AC (acoustic time difference), RT (deep lateral resistivity), GR(natural gamma), DEN(density). And we selected21gas samples,8gas-water layer samples and21aqueous layer samples of50samples from18wells. Using the extreme value normalization method, we take the normalization processing measures to the property values of the modeling samples and finally we get the property values which are mapped to [0,1]. And then based on the existing gas-water identification and evaluation of technical results, we concluded four kinds of the most effective methods of gas-water identification. They are the stepwise discriminant analysis, probabilistic neural network, fuzzy clustering analysis, gray clustering analysis. And we add a kind of KNN (K nearest neighbor nodes) intelligent recognition method. At the same time, carefully we introduced the principle of the five kinds of gas-water identification methods, and analysised their application effect in gas-water identification of Gaoqiao. And KNN is initially applied in gas-water identification. Through analysing the effects of the application of these methods in the study area, we elected three kids of gas-water identification methods which are the stepwise discriminant analysis, probabilistic neural network and KNN (K nearest neighbor nodes) intelligent recognition method. These methods improved the coincidence rate of Reservoir Log Interpretation about computer. This provided methods and basis for later research of gas-water identification in log interpretation. It has high reference value. These methods can also be applied in neighboring blocks of Gaoqiao and the gas reservoirs which have similar reservoir characteristics.