Study On Lithologic Identification Of Volcanic Rock

Previously, oil exploration and development focus on clastic rock and carbonate rock. But with the improvement of exploration difficulty, coupled with the rapid development of economy, the demand for oil and gas resources increase. It is very important to find reservoir of new type. Volcanic rock reservoir is arousing more and more attention widespread as one kind of peculiar reservoir.Lithologic identification of volcanic rock is the most important basic work in reservoir evaluation. The composition, structure of volcanic rock is complex and volatile, which leads to variety of lithology. How to improve the accuracy rate is the difficult problem of lithologic identification in volcanic reservoir evaluation. At present, lithologic identification of volcanic rock is based on the composition of volcanic rock, such as crossplot method and microelement well logging, but never use the texture and structure information.Conventional logging data mainly reflect the composition of different volcanic rocks, weakly reflect structure information. Image logging data can provide structure information of the rock. The Gray Level Co-occurrence Matrix (GLCM) method is used to extract image texture feature. By analyzing the response features of conventional features and texture features, some good features are selected to recognize lithology of volcanic rock with Support Vector Machine (SVM) method. This paper had the following parts:1. By analyzing the chemical composition and mineral composition of volcanic rock, we divide volcanic rock into basalt, andesite, tuff, and volcanic breccia.2. In data preprocessing, we focus on the depth correction and the texture features of the median filter. By comparing the GR curve of conventional logging data with image logging data, correcting value are calculated, then depth correction is finished. The texture features are filtered by five-points median filtering, then the abnormal points will be eliminated, also the data quality can be improved greatly.3. According to different lithology, the statistical analysis method is used to calculate mean, variance and analyze the response features of different volcanic rocks with core and logging data. Crossplot analysis method is used to identify and extract the features of the different lithology with conventional logging data, as well, GR, DEN, CNL, AC and RD curves are finally selected to identify lithology.4. Using the Gray Level Co-occurrence Matrix (GLCM) method to extract image texture features with image logging data. These features include ten image texture features:contrast, energy, entropy, local uniformity, relevance. Also, the log response features of different lithology are analyzed. We select six texture features, include the contrast, entropy, local uniformity from dynamic image and the contrast, entropy, local uniformity from static image, to identify lithology.5. The method used to identify lithology is Support Vector Machine (SVM) method. Three group tests are worked:comparison of different model parameters test, comparison test of different characteristics and different training sample comparison test. The results indicate that using conventional logging features and texture features, recognition accuracy is higher than using conventional logging features or texture features alone. Studying on lithologic identification for small sample with Support Vector Machine (SVM).6. At last, the volcanic rock from eight well are identified, which get a good result. There are four kinds of lithology:basalt, andesite, tuff, basaltic breccia, andesitic breccia and tuffaceous breccia. The identification rate is nearly80%.