Theory And Technology Research On The Lithology Identification And Evaluation Of Volcanic Reservoir

As numbers of huge volcanic reservoirs have been found all over the world since 19th century 20 years, the enormous potential for exploration and development of volcanic reservoirs has aroused much attention and interest of the researchers around the world. At present,volcanic reservoir is becoming a research focus in the area of oil and gas exploration.Because of the variety of lithology and the complexity of reservoir space, as well as the poor physical properties, volcanic reservoir has more complex rock-electricity relation compared with clastic sandstone and carbonate rocks, and this brings great difficulties about the logging interpretation on volcanic reservoir. The main difficulties are:1) improving the coincidence rate of lithology identification when dealing with the volcanic reservoirs with complex lithology and components; 2) immature theory and methods for accurate identification and quantitative evaluation of fractures and vugs, which are the storage space and channel of fluid for volcanic reservoir; 3) the unreasonable interpretation model for volcanic reservoir, because in the past rocks were directly used as component to establish volume model, rather than mineral volume model, thus affecting the calculation accuracy of reservoir porosity, permeability, saturation and some other parameters.The identification of volcanic reservoir's lithology, quantitative evaluation of fractures and building the logging interpretation model are important supplement to the reservoir evaluation after clastic rocks and carbonate rocks. As the evolvement to the traditional logging data processing methods and interpretation theory, the logging evaluation techniques and interpretation methods for volcanic reservoir have significant theory meaning, and will promote the exploration and development of various volcanic reservoirs in our country.As the existing logging evaluation techniques and interpretation methods have a series of problems and difficulties in theory and data processing, systematic research has to be carried out to establish appropriate evaluation theory and interpretation method, so that we could make a breakthrough in the logging evaluation of volcanic reservoir. According to the existing problems and difficulties in logging evaluation on volcanic reservoirs, we systematically studied the theories and techniques for volcanic reservoir identification and evaluation. Based on the establishment of the logging interpretation model, appropriate evaluation theory and data processing method were established to obtain each parameter of the model. We strived to establish a general evaluation theory and data processing method for volcanic reservoirs.1. Summary of volcanic rocks logging response characteristicsBased on the domestic and foreign literatures as well as our years of research on the deep volcanic reservoir in Songliao basin (China), we systematically reviewed the logging response characteristics and response rules of volcanic strata.1) In detail, we reviewed the logging response characteristics of the typical volcanic rocks in Songliao basin, including a variety of conventional features of logging curves and the image response characteristics of formation image logging. Given the specific range of logging response parameters of volcanic rocks in Songliao basin, general rules for the response characteristics of volcanic rocks was concluded.2) We also discussed the general characteristics of volcanic reservoirs according to the core analysis of porosity, permeability and rock density, et al.2. Lithology identification of volcanic rocks:the determined mineral composition matrix in the logging interpretation modelLithology identification is the basis and difficulty of the volcanic reservoirs research. To identify the lithology of volcanic rocks through logging data, we have to establish a response relationship between logging data and geologic lithology naming data. As a result, the first thing we need to consider is how the lithology of the volcanic rocks classify in geology. Only when the lithology types and the geological classification principles are associated appropriately, accurate logging identification results can be obtained. Geologically speaking, there are four categories of volcanic rocks which are classified according to the chemical composition of the rocks. They are basic rocks, intermediate rocks, intermediate-acid rocks and acid rocks. According to the structures of volcanic rocks, however, there are three categories:volcanic lava, pyroclastic lava and pyroclastic rocks. The exact types of the volcanic rocks can be determined by comprehensively considering the information of mineral composition and structures.1) Lithology identification according to the chemical composition of four types volcanic rocks.For the information of volcanic well section from more than 30 wells in Songliao basin which had exact naming data of core slices, we made each depth point with naming data as a sampling point (totally 430 points). Six conventional logging data were read, such as GR, TH, U, K, PE and NPHI. These data are used as the basis to discuss the lithology identification of volcanic rocks.a) Cross plot method using two groups of conventional logging data:for the 430 sampling points, the cross plot of GR-TH had the best identification result which had a numerate correct rate of 93.72%.b) Statistical method using multiple groups of conventional logging data.Cluster analysis:for the 430 sampling points, case cluster (i.e. q-type cluster) was adopted, and K-means cluster analysis with given number of categories was selected. When the iteration number was 20, the numerate correct rate was less than 50%.Discriminant analysis:for the 430 sampling points,90.2% of numerate correct rate was obtained with Fisher discriminant analysis method.Principal component analysis:the numerate correct rate was 78.42%.Factor analysis:the numerate correct rate was 77.95%.Correspondence analysis:the numerate correct rate was 65.82%. c) Establishment of pattern identification method for exact sample by using multiple conventional logging data.BP neural network method:with 60 input nodes,5 output nodes,100 middle layer nodes,0.01 of learning step and 10000 times of training, the numerate correct rate was 64.42%.SOM neural network method:standardized by mean square deviation, with 2000 of iteration number and 10*10 of output layer,95.35% of numerate correct rate was obtained.Support Vector Machine method:using radial basis function as kernel function, when C=20 andγ=23, the numerate correct rate was 98.84%.d) Pattern identification of the data processed after statistical analysis.The principal component analysis, factor analysis and correspondence analysis of statistical analysis method can well extract a few independent composite indicators of the sample from multiple conventional logging data. Thus, the correlation between inputs can be eliminated, and the input dimension can be reduced. With a simplified network structure, the network convergence can be speed up, too. As a result, the neural network method and support vector machine method that based on the principal component analysis, factor analysis and correspondence analysis, can all obtained higher correct rate.e) Selection of the final result.The methods which had numerate correct rate up to 80% were selected, and the final result was obtained using voting method. In the case that the lithology types obtained by various methods were different with each other, the result obtained by Support Vector Machine method which had the highest correct identification rate was provided as the final identification result.2) Lithology identification according to the structures of three main types volcanic rocks.a) Cross plot method using conventional logging data.Through analyzing the cross plot, we found that the logging curves which were sensitive to the rock structure included NPHI, AC and RT.b) Wavelet transform method for the conventional logging data.One-dimensional continuous wavelet transform method was mainly adopted. The transform result of dbl wavelet, bior wavelet, Haar wavelet on the resistivity curves, and the transform result of dbl wavelet and Haar wavelet on the acoustic time curves, as well as the transform result of Haar wavelet on neutron porosity curves, all have indication effect on the structure classification of volcanic rocks.c) Fractal dimension method for the conventional logging data.The ranges of response values of resistivity curves, acoustic time curves and neutron curves after fractal dimension calculation was reviewed and used to identify the type of volcanic rocks with different structures.d) Identification method using image logging images.FMI image identification mode was established in term of the volcanic rocks categories which were classified according to the rock structures, and it was applied to identify volcanic rocks with different structures.e) Selection of the final result.Based on the result of FMI image identification mode and consulting the results of fractal dimension method and wavelet transform method, the final result was determined.3) Final determination of the exact types of volcanic rocks by comprehensively considering the information of mineral composition and characteristic structures (see Table 4-1-1).3. Study of obtain fracture and vug porosityThe calculation of fracture and vug porosity is conducted by studying the data of image logging.1) Using the human-computer interactive interpretation function of software to calculate fracture and vug porosity based on the data of image logging.We have analyzed many FMI image modes of various types of fractures in the deep volcanic strata of Songliao basin. Using LogView, the image logging data was decoded and mapped. The interactive interpretation function of this software provides a platform for the identification of fractures. Besides, the interactive interpretation parameter calculation function of the software could evaluate the fractures, and provide hydraulic electrical fracture aperture (FVAH), areal trace length (FVTL), corrected fracture density (FVDC) and apparent electrical fracture porosity (FVPA).2) Digital image processing method based on the image logging images.By studying the FMI image information from image logging, fractures and caves were emerged from the basic rocks using computer image processing method, i.e. image enhancement technology, edge detection and image segmentation. Then, using Hough transformation, contour extraction and tracking technology, the fractures and vugs are extracted and recognized. Finally, the quantitative parameters of fractures and caves were calculated with image analysis techniques according to some algorithm. This effectively avoided the error caused by subjective factors during human-computer interactive interpretation process.3) Evaluation method for fracture porosity based on the sampling data of image logging.For the volcanic strata with image logging data, we can directly study the sample data matrix of image logging to identify and evaluate the fracture development. Threshold values can be chosen with appropriate methods to distinguish whether the resistivity of each point in the matrix is low resistivity response (caused by filling the pores with mud) or high impedance response (caused by dense rock and such kind of materials). Then adjacent vertex search model is created to search for low resistivity response range. Thus the fracture development can be evaluated.4) With the above results, using the weighted average method to determine the final value of fracture and vug porosity.4. Establishment of logging interpretation modelThe volcanic rocks containing clay minerals can be seen as the transition of volcanic and metamorphic rocks, and are not within the scope our study which focuses on the typical volcanic rocks of Songliao basin. Thus, the established logging interpretation model contains two parts, i.e. matrixs and fractures of the rocks. The rock matrix established through lithologic identification theories and methods is called determined rock matrix, and the resultant logging interpretation model is determined matrix volume model. Density logging, neutron logging, or the combination technique of these two methods can be used to calculate the total porosity. As a result, the matrix porosity can be obtained by calculating the difference between total porosity and fracture porosity.5. Comprehensive reservoir evaluationBased on the calculation of porosity, permeability was calculated according to appropriate formula. Then water saturation was calculated according to water saturation equation. Finally, under the guidance of the regional geological data, comprehensive evaluation to the volcanic reservoirs was given based on the above logging evaluation results.