| With the rapid increase in energy demand,the proportion of heterogeneous clastic rock reservoirs with low porosity,low permeability,and density is increasing,which has attracted great attention from scholars and oil companies at home and abroad.The pore structure of heterogeneous clastic rock reservoirs is closely related to physical properties,seepage,and electrical properties,and is an important factor affecting reservoir quality and fluid properties.There is a lack of systematic quantitative evaluation techniques for pore structure of heterogeneous clastic rock reservoirs at home and abroad,as well as effective processing methods for saturation calculation,fluid properties,oil water interface identification,and productivity evaluation.This has brought great difficulties to the precise evaluation of heterogeneous clastic rock reservoirs by logging.Nuclear magnetic resonance technology can accurately,conveniently,and effectively identify the types of fluids in micro pores,which is of great significance for the study of clastic rock reservoirs.Therefore,this thesis attempts to use nuclear magnetic resonance technology to classify and study clastic rock reservoirs.This thesis takes the Daqingzi well in Changling sag as the research area,which is located in the south of the central depression area of Songliao basin.In this thesis,95 groups of core samples obtained from the study area were used to obtain nuclear magnetic resonance experiments,and nuclear magnetic resonance T2 spectrum was obtained.The morphology of nuclear magnetic resonance T2 saturation spectrum and centrifugal spectrum was preliminarily analyzed,and the porosity and permeability of the core samples were calculated.Then calculate the arithmetic mean,geometric mean,T2 cutoff value and other parameters of nuclear magnetic resonance T2 spectrum,and find the peak,skewness,kurtosis,standard deviation,coefficient of variation and other parameters describing the characteristics of pore sorting in the research process,and analyze the correlation between each parameter and porosity and permeability.Using the above parameters to fit the NMR T2 spectrum to obtain the normal distribution curve,calculate the normal distribution related parameters and obtain their relationship with porosity and permeability.On this basis,compared with the traditional pore structure classification method,the morphological characteristics of the nuclear magnetic resonance T2 spectrum are analyzed,and the nuclear magnetic resonance spectrum of the sample is divided into six components S1(125ms)by the method of component division.Then,combined with the above parameters,it is divided into four categories by K-means clustering analysis method(single peak partial fine pore throat type,single peak partial coarse pore throat type,double peak partial fine pore throat type,double peak partial coarse pore throat type),and 0,1,2 and 3 labels are set to correspond to four categories.Finally,the support vector machine machine learning method is used as the training set for training,and then the new 54 groups of nuclear magnetic data are quantitatively and qualitatively classified and predicted for clastic rock reservoirs,and the data processing and analysis evaluation of the target layer XX59 well in the study area are carried out. |
