Identification of coal structures using geophysical logging data in Qinshui Basin, China: Investigation by kernel Fisher discriminant analysis

Abstract Coal structure is closely related to the porosity and permeability of coal reservoirs, which not only affects the enrichment of coalbed methane (CBM), but also influences the hydraulic fracturing and efficient development of CBM. The accurate identification of the coal structure would be a critical issue in CBM exploration and development, and is always a challenge. Compared with traditional methods for identifying coal structure base on borehole cores or mining seam observation, geophysical logging has become the most economic and efficient technique. Several linear correlations have been established to describe the relationships between the coal structures and well logs. However, those correlations cannot accurately reflect nonlinear relations between them. Therefore, a reliable and efficient method to identify coal structure is needed. As a powerful nonlinear classifier, the kernel Fisher discriminant analysis (KFD) method has been widely used due to its strong generalization ability. In this paper, a new quantitative coal structure identification model was developed based on the KFD method by using geophysical logging data. The model was trained, tested and optimized using 178 logging data sets from 15 CBM wells in Qinshui Basin, China. The approach accounted for all the available well logging attributes, and the training data sizes and kernel parameters were analyzed to get the most appropriate model in practice. In addition, the built model was validated individually by employing logging data of a new CBM well. The results indicate that the KFD based identification model has high prediction accuracy, which can be used as a reliable method for coal structure identification. The KFD method exhibits strong capability during the modeling and generalization in the determination of nonlinear relationships, which provides an efficient way for the coal structure prediction in basic research of coal reservoirs.